Optimizing Virtual Machine Live Migration without Shared Storage in Hybrid Clouds

Size: px
Start display at page:

Download "Optimizing Virtual Machine Live Migration without Shared Storage in Hybrid Clouds"

Transcription

1 2016 IEEE 18th International Conference on High Performance Computing and Communications; IEEE 14th International Conference on Smart City; IEEE 2nd International Conference on Data Science and Systems Optimizing Virtual Machine Live Migration without Shared Storage in Hybrid Clouds Shan He, Chunming Hu, Bin Shi, Tianyu Wo, Bo Li State Key Laboratory of Software Development Environment (NLSDE) School of Computer Science and Engineering Beihang University, Beijing, China, Abstract Virtual machine live migration technology allows a running VM migrates from one physical host to another with no impact on users. As the scale of distributed computing gets larger and larger, hybrid cloud, which is integrated cloud service utilizing both private and public clouds to perform distinct functions within the same organization becomes a hotspot for both academia and industry. Thus, it will be vital to migrate virtual machines in hybrid clouds. The biggest issue which VM migration in hybrid clouds faces today is how to transfer the large amount of storage data. Storage migration of virtual machine with existing mechanisms of memory migration has been studied for a long time. Unfortunately, no one can perform perfectly without modification. It is compelling to consider a new migration model and optimization methods. In this paper, we analyze the advantage and disadvantage of classical methods, and put forward a method which combines the strengths of pre-copy and post-copy migration model and we also use disk working set to optimize virtual machines storage migration. We have developed and implemented our approach to the QEMU/KVM hypervisor and run a series of experiments. The presented technique shows a reduction of up 20.5% on average of the total transfer time for most of the selected scenario. Our approach can also reduce the migration downtime in most cases. Index Terms virtual machine; live migration; storage; working set I. INTRODUCTION Nowadays cloud computing is a popular topic, more and more companies have migrated their business to cloud environments. Most of these cloud infrastructure can be divided into two categories. One is private clouds, which is a particular model of clouding computing that cloud infrastructure can be operated by a single organization, the other is public clouds in which service providers make resources available to the general public over the Internet such as EC2 [1] and Azure [3]. Recently, hybrid clouds, a combination of public and private clouds, become increasingly popular. Virtual machine live migration allows a running VM to move from one physical host to another with no impact on the availability of virtual machine, that makes it possible to be leveraged for a variety of tasks such as load balancing, system maintenance, fault tolerance, and so on. Live migration is an importance feature of virtualization technology for administrators of data centers and clusters, because it can significantly increase the mobility of data centers and manageability of clusters. In a cloud environment, live migration only transfers memory and devices state between two machines that share a common local network and storage system such as Storage Area Network (SAN) [4], Network Attached Storage (NAS) [11]. SAN is a high-speed network of storage devices which provides block-level storage that can be accessed by applications running on any networks servers. With the development of the technology of XVLAN [17], OpenFlow [20] and other virtual networks, it is possible for virtual machine migration in wide area networks (WAN) with no disruptions on network connectivity. however, it is less likely to use shared storage system in hybrid clouds for some reasons. First, existing share storage systems do not provide high efficient I/O performance in WAN as they do in LAN. Second, once a shared storage system has been implemented among several clouds, it is necessary to rely on the storage devices outside of cloud environment, which can bring security problems and difficulties of data management. Consequently, the ability to migrate VM across hybrid clouds without shared storage system is increasingly compelling. WAN VM migration is considered as hotspots of research by the academics and business.some solutions have been proposed by moving existing technologies for migrating memory and device state to transfer storage. In this work, we analyze the advantage and disadvantage of classical methods, such as pre-copy [10], post-copy [12] and hybrid copy. We propose a method migrating virtual machines storage combine the strengths of pre-copy and post-copy with disk working set which is the set of blocks that may be read later. This method can have negligible downtime, shorter total time of migration and minimal impact on the application. We present an integrated memory and storage migration system with our disk working-set. Before we introduce our approach, we analyze the disadvantage and advantage of existing mechanisms of live migration. We found that combining the advantage of pre-copy and post-copy can reduce the amount of data being transferred and have a minimal impact on application at the same time. We do a series of experiments to research disk working-set according to different applications. In some scenario, it can work well. Then, we have implemented our approach to the KVM hypervisor and do some experiments. This research is meaningful, although its performance is not always so good for some cases /16 $ IEEE DOI /HPCC-SmartCity-DSS

2 The contributions of this paper are as follows: We propose a method of live storage migration with the disk working set which is the set of blocks to be read in the post-copy stage. This method can minimize total migration time, downtime,volume of data transferred, and the impact on the application. We show the feasibility of the proposed technique by providing an implementation of this method in KVM hypervisor; We demonstrate the effectiveness of the proposed technique by running a series of benchmarks. We achieve an average improvement in the total migration time of over 15.45% with up to 29.38% for certain scenarios; Our migration solution can be applied to various networks environment, compared with the migration with IO Mirroring, has better robustness; The rest of this paper is organized as follows: Section II gives an overview of the background and our motivation. In section III, design a series of experiments to find out disk working-set and analyzes the feasibility of live storage migration with disk working set. Section IV and V contains the design and implementation of the proposed live migration framework. Section VI evaluates our techniques. Section VII introduces the related work on the live migration. Section VIII concludes the paper. II. BACKGROUND AND MOTIVATION The technologies of live migration can be divided into three migration models: the pre-copy model, the post-copy model, and the hybrid copy model. Each of the models has its advantages and limitations. This section introduces and analyze these models,after comparing theirs merits and shortcomings, gave the goals our method requires to meet. A. Pre-copy model In the pre-copy model, the target VM starts running after entire VM is copied to the target host.one of the most famous pre-copy models is the iterations pre-copy method. As it is implemented by KVM [14]with Dirty Block Tracking (DBT) [15] [16], which uses a bitmap to mark dirty blocks. The disk file is copied to the target host from beginning to end, and the memory is migrated at the same time. During the migration, the source VM keeps running. And all write operations are intercepted and trigger the operation of set the bitmap. The blocks are transmitted in several iterations. In each iteration, only the dirty blocks marked in the previous round need to be sent again. Another prevalent approach of the pre-copy is IO Mirroring [18], which will interpose on all disk write operations and sent data to the target host immediately. When all blocks have been copied, the migration ends. The IO Mirroring method just one round and synchronize dirty blocks when write operations happen. The advantage of this model is all read and write operations have good performance when the VM runs at the destination. The disadvantage is dirty blocks are transferred repeatedly, which will lead to extra traffic and the increase of the transmission and migration time. The iterations pre-copy method requires transmission in the last round to converge so that it can enter into the stop and copy stage. But for large size of disk, it takes a long time for a round to finish, while the number of blocks marked in this round is large. So it is hard to obtain a converged value. B. Post-copy model In the post-copy model, the target VM will start running as long as the basic data to support running of VM have been forwarded to the target host. After the state of the target VM is running, storage migration begins. This model includes three steps. First,live memory migration is executed until the number of dirty pages converges. Second, stop the source VM and copy rest of dirty pages to the target host. Last, start the target VM and then begin storage migration. When read operations at the destination are on the block that has not been sent to the target host, the target host will send a request to get those blocks from the source host. This operation is remote reads which incur extra delays, the application must wait until it gets the required data. So, in this model, each block is transferred at most once so that the total transmission of data for blocks is minimized. However, the impact caused by remote reads on the applications is the most serious of three migration models. C. Hybrid-copy model We migrate the blocks, not in the Disk working set, by the Post-Copy with on-demand fetching. Before the beginning of the Post-Copy stage, we start a server which listens to HTTP requests. It starts a separate TCP connection for every request. During the Post-Copy stage, the destination VM is running and the source VM is paused. Each I/O read operations will be caught by the I/O Working module, which will call the ondemanding client to get the read data if these blocks have not been in the destination. When issuing the write operations, we record the write blocks to remark the blocks are the newest which means that it wont be updated by the Block Copy module. Size of chunk requested is 1M, which is the same with the base unit. D. Efficient storage migration model Moving a VM to the another cloud requires substantial downtime and migration time when the entire disk is copied to the target host with current available technology in hybrid clouds. An efficiently live migration model should have a negligible downtime, keep the consistency of data and transparent to the user. So four kinds of measure values to evaluate the performance of the migration methods are the downtime, total migration time, the performance of VM, total data transmission. As mentioned previously, in the pre-copy model the target VM starts running after migration is over, so there is no additional performance degradation when the VM runs at the destination, and remote reads cannot happen. As table Ishows, the strength of the pre-copy model is the minimal impact on the application, but it has the largest total data transmission. The migration time increases as the transmission increases. On the contrary, the post-copy model 922

3 TABLE I: comparison of three migration model Migration model Down time Total migration time Total transmission Performance impact Pre-copy 1s most most least Post-copy least least least most Hybrid copy 1s middle middle middle has the minimum data transmission. While its shortcoming is the heavy impact on the applications because of remote reads. In an ideal situation, we expect the negligible downtime, the minimum transmission, the shortest migration time and the negligible impact on the user. However, it is impossible to achieve but we can get very close to it. The hybrid copy model is a compromise which can have both strength of the pre-copy model and the post-copy model. It is possible to seek a satisfactory balance between the migration time and the performance of virtual machines by using a hybrid copy model. (a) CPU workload s trace (b) memory workload s trace III. DISK WORKING SET AND CHARACTERISTICS This section analyzes the characteristics of disk working sets from different workloads. To understand whether the history I/O operations can be useful for optimizing storage migration, we collect and analyze various IO trajectories of diverse workloads. The idea of disk working set comes from the concept of memory working set, which defines the pages accessed in a given time interval. Analogous to the working set in memory, disk working set is the approximation of the set of blocks that the process will access in the future for a certain time. In the latter part of this section, we put forward a hybrid copy model with disk working set. A. Trace collection In order to analyze the relevant of working set characteristics and storage migration. We choose four kinds of diverse workloads, high CPU load, memory overhead, file I/O intensive and OLTP workloads which are common in web services. All the benchmark workloads we collect are simulated by sysbench benchmark [8] which is one of the most common benchmark tools. For example, the workload of high CPU load is the CPU performance test which is to verify prime numbers by doing standard division of the number by all numbers between two and square root of the number. And we can specify the maximum number to determine the running time. In our experiment, we set the maximum is 40000, which can be running for twenty minutes in VM. The configuration of VM is 1 CPU, 2GB RAM and 10GB disk. Two machines with 16GB of RAM and 8 Intel Core i7 processors are utilized to run VMs. We get the I/O trace through the virtual machine monitor (VMM) which can interpose all virtual disk read and write operations. We intercept all the I/O operations and record the read or write property, the access time, the number of sectors and quantities of sectors requested. Each 512 bytes means a block. We run all workloads in the same environment. B. Temporal locality and Spatial Locality Characteristics Temporal locality and Spatial Locality Characteristics are basic types of the principle of locality, which used to describe (c) File IO workload s trace (e) OLTP workload s trace (d) local graphs of File IO workload s trace (f) local graphs of OLTP workload s trace Fig. 1: Trace collection of four workloads simulated by sysbench. The horizontal axis shows time in milliseconds, while the vertical axis represents the index of blocks. The red means the read operations to disk, while the green dots are write operations to disk a phenomenon in which the same values and related storage locations, are frequently accessed. Temporal locality Characteristics means that the same values are likely to be accessed again in the future. Spatial locality Characteristics refers to that if a particular storage location is accessed. Then it is very likely that adjacent positions will be referenced in the future. As Figure 1shows, we record I/O traces from four kinds of workloads. The horizontal axis shows time in milliseconds, while the vertical axis represents the index of blocks. The red means the read operations to disk, while the green dots are the write operations to disk. Temporal locality and Spatial Locality Characteristics are obvious. Take the workload of high CPU as an example, even though the I/O operations of this workload are not so many, two lines appeared in the figure 923

4 1a. Suggest the same blocks and neighbors are accessed in the near future. File IO workloads have complicated traces as show in figure 1c.First, It creates a large number of test files, then produces a lot of IO operations according to the set mode. We can make a choice between the test mode from sequential reads and writes or random reads and writes or a combination. Random read and write is used. In the figure 1d, we provide a local image of the figure for showing I/O traces in details. From this local image, it is evident that read operations gather in some chunks while write operations gather in other chunks. From these figures, across all workloads, the principle of locality works well. So it is possible to take advantage of disk working set in the migration. Whats more, the working set of read operations and the working set of write operations have different impacts on the process of storage migration. One of the most important differences, In the post-copy stage, remote reads will reduce the performance of service, while write operations at the destination have no impact on the applications. However, write operations in the pre-copy stage will increase the number of dirty blocks, which it expects to be a very small number. C. A hybrid copy model with disk working set To optimizing storage migration, we propose a hybrid copy migration with disk working set. History, the first round of virtual disk copy and memory migration belongs to the pre-copy stage. During the pre-copy stage, dirty blocks are maintained by a block bitmap which will be sent to target host at the stop and copy stage. Dirty blocks are transferred in the post-copy stage. The hybrid copy model is a compromise method that cant overcome all disadvantages. The long migration time and large transfer volume are the shortcomings of pre-copy model. While the performance of virtual machine affected by remote reads is the weakness of the post-copy model. Retransmission of dirty blocks is the major reason to increase time. In the hybrid copy methods, only dirty blocks have to be copied again and all blocks are copied twice at most. So compared with the pre-copy model, it can reduce the quantity of data transmission. When compared to the post-copy model, the hybrid copy model causes fewer times of remote reads but have to carry more blocks. Most of hybrid copy methods don t make full use of the advantage that coordinates the balance between the migration time and the impact on the applications. Some optimum methods can be taken to get the best results. First of all, If the blocks that will be accessed by remote reads in the post-copy stage can be sent to the target host in the pre-copy stage, the number of remote reads can significantly decrease. It is clear that the set of blocks have to be sent in the pre-copy stage is as small as possible. The experimental result tells that the disk working set exists in the most workloads, and the set of blocks written can be different from the set of blocks to be read, what s more, these two sets may overlap a small part of books in most cases. If we make the set of blocks corresponding to read operations as the set which has to be migrated in the pre-copy stage, it can meet the requirement that the blocks of remote reads is copied in the pre-copy stage. So we transfer working set of blocks accessed by read operations in pre-copy stage and other data in the post-copy stage, we can get the best trade-offs between the total time and the impact on the applications. Sometimes a compromise is the best methods. In addition, our solution can be applied in various networks environment, compared with the migration with IO Mirroring, has better robustness. The convergence of dirty data in the pre-copy model is required to ensure it works normally. IV. DESIGN Our system provides virtual machines live migration without shared storage in the hybrid clouds. Our goals are to optimize the storage migration which is an important issue in a wide area network. We wish to present a method that can have minimal total migration time and downtime time, the least impact on the application. This section will present the architecture of our migration system and the process of the entire migration. A. Architecture overview Our system is based on the hybrid copy model and uses IO Mirroring mechanisms to sync dirty blocks. Living memory migration is implemented using an iterative pre-copy which is the most popular approach. Living storage memory is a hybrid copy method which determines blocks sent in the pre-copy stage by disk working set. First, a block bitmap for marking disk working set is generated depending on the historical data at the beginning of migration. Then, in the pre-copy stage, the migration of disk working set is prior to the memory migration. The migration of disk working set uses IO mirroring to keep consistency. In the stop and copy stage, the dirty pages and the dirty blocks in working set are transferred to the target host. Last, in the post-copy stage, the target VM starts running, then the blocks which are not included in disk working set are copied from the beginning of the disk file to the end. At the same time, if a read operation at the destination accesses to the blocks which have not yet been transferred, the target host will fetch these blocks from the source host on-demand. Figure??depicts the system architecture. The Live Migration module is responsible for the tasks of memory migration implemented by the method of iterative pre-copy. The Transport module handles memory pages and disk blocks transfers between the two hosts. The I/O Working is a multi-functional module that intercepts all I/O operations and deals with these differently according to the stage, such as update the information that will be used to generate a block bitmap of working set in the non-migration stage. During the non-migration stage, the I/O Working intercepts I/O operations for recording the historical information. In the pre-copy stage, the I/O Working interposes on all disk writes and mirrors these to the Block Copy module. The I/O Working on the destination is responsible for remote reads. The I/O Working module only works on the virtual machine whose state is running. For an example, in the postcopy stage, the source VM is stopped so that no I/O operations can be intercepted at the source host. 924

5 Fig. 2: Architecture Overview: shows the main components involved in migration The I/O Write/Read module is responsible for writing and reading data from and to the virtual disk. Two major functions of Block Copy module in the source is to copy the data into the disk buffer and synchronize the mirror. In the pre-copy stage, Block Copy module copies disk working set through the I/O Write/Read module from the beginning to end. At the same time, Block Copy module receives the disk writes from the I/O Working module. If the write operation is to a block that has been copied, Block Copy module then copied these data to the Disk Buffer. In the postcopy stage, Block Copy module proceeds linearly across the disk, copying the disk region that has not been copied to the Disk buffer. While Block Copy module at the destination is working to copy the received data to the virtual disk file. The on-demand fetching Server module and the on-demand fetching client are responsible for issuing remote read in the post-copy stage. On-demand fetching Server gets data from disk by the I/O Read/Write model, then transfers these data to the on-demand client. The on-demand client module can be called by the I/O Working module at the destination when the operation of read data is on the disk block which has not been transferred. B. Process of migration Our migration methods can make up of four stages: nonmigration stage, pre-copy stage, stop and copy stage, post-copy stage. As figure 3 shows, this section describes the work of each stage. 1) Non-migration stage a) The I/O Working module intercepts all the I/O operations and records the I/O sequence. all disks are divided into many chunks size of 1M, and number these chunks from the beginning to the end. Then we record the number of chunks instead of the number of the sector when we record the I/O sequence. We define the max size of working set 1/10 of disk size. 2) Pre-copy stage a) At the beginning of migration, a bitmap for marking disk working set is generated according to the historical records b) The Block Copy module on the source transfers blocks belong to disk working set sequentially, while receives the disk writes from the I/O Working and mirrors those blocks that have been copied; c) Until all blocks of working set have been copied, living migration module starts memory migration. Block Copy module still issues disk writes during the memory migration; d) It is the time to stop the source VM When memory migration converges and the quantity of data transmission include disk blocks and pages can be transferred with a limited downtime; 3) Stop and Copy stage a) Remaining data is copying to target host; b) Start the on-demand fetching server that issues remote reads; c) After the steps above completed, a sign represents end of this stage will be sent to the target host. The target VM starts running after receiving sign. 4) Post-copy stage a) The Block Copy module on the source host transfers blocks that have not been copied to the Disk buffer from the beginning to end; b) The I/O Working at the destination intercepts I/O operations, if the operation is read blocks that have not been received, the I/O Working module get data of blocks by the On-demand fetching client which requests those blocks from the source; 925

6 TABLE II: Five kinds of applications Application Idle OLTP File Server Compress file Kernel compile Description No application Sysbench benchmark Postmark benchmark Compressing a large file(3.7g) compiling the Linux kernel Fig. 3: Shows the migration process c) When all blocks have been copied to the target host, the migration is over; V. IMPLEMENTATION We have implemented the proposed technique in the opensource virtual machine monitor KVM, and visualizer QEMU [7], which intercepts disk I/O request. The section discusses the details and issues of our implementation. A. Disk working set Disk working set is the set of blocks that moved to the target host in the pre-copy stage. As we discuss in?? We expect that blocks will be read in the post-copy stage can be in our working set. The less the requests of remote read are, the better performance of VM is. When the target VM issue disk writes, it is safe to update blocks directly. So we only concern the disk working set of reads operations. We record the I/O read requests sequence to generate a bitmap of disk working set. To take use of spatial locality characteristics, we consider 1MB chunk instead of 512B Block as the basic unit of counting accesses. First, migration methods in KVM use 1MB as the basic unit of transferring, and work of [25] shows that 1MB chunk can have a good effect than 512B. Second, in the implement of migration in KVM, each basic unit are copied to target host with the overhead costs of a few bytes, such as the flags to tell target host what the following data represent, device name and its length,the address of block and so on. The smaller size of transmission chunk is, the greater the overhead. We record the most recent N the index of chunks in a FIFO queue. We set the size of working set is M which is one-tenth of the size of the disk. N is larger than M for a lot of duplicated chunk. At the beginning of migration, we get M different chunks from the N sequentially, then generate bitmap. B. Pre-Copy and IO Mirroring We migrate the Disk working set by the Pre-Copy method and synchronize dirty blocks with IO Mirroring. We use a bitmap to mark the disk working set which use a bit to represent a block. IO Mirroring reflects any additional changes to the source disk that take place during storage migration. Our Block Copy module is implemented to linearly across the disk and copying one disk region in working set at a time. The I/O Working on the source interposes on disk writes and sends these to the Block Copy module. We record the max value of the number of blocks that have been copied. All synchronization works can fall into three types: 1) write to a chunk that has been copied; 2) write to a chunk being copied; 3) write to a chunk that has not been copied. In the first case, Block Copy module enqueues data to the disk buffer if the chunk belongs to disk working set. In the second case, Block Copy will enqueue data when the copy of this chunk has finished. The disk buffer is implemented by a queue. In the last case, we do nothing. C. Post-Copy and on-demand fetching We migrate the blocks, not in the Disk working set, by the Post-Copy with on-demand fetching. Before the beginning of the Post-Copy stage, we start a server which listens to HTTP requests. It starts a different TCP connection for every request. During the Post-Copy stage, the destination VM is running and the source VM is paused. Each I/O read operations will be caught by the I/O Working module, which will call the ondemanding client to get the read data if these blocks have not been in the destination. When issuing the write operations, we record the write blocks to remark the blocks are the newest which means that it wont be updated by the Block Copy module. Size of chunk for on-demand fetching is 1M, which is the same with the basic unit. VI. EVALUATION We implemented IO Mirroring and our prototype implemented in the KVM and QEMU. All experiments are performed on two identical hosts with a 16GB memory of RAM and 8 Intel Core i7 processors. The configuration of VM for migration is 1CPU, 2 GB RAM and 40GB disk. For lacking a standard benchmark suite for live migration of virtual machines, we have selected several general application scenarios listed in II. We use five workloads, an idle VM, OLTP simulation using sysbench benchmark, File server simulation using postmark [13], compressing a large file and compiling the Linux kernel. The OLTP workload used sysbench to generate to IO pattern that simulates an OLTP workload with random access. We use postmark benchmark to simulation File Server workload with a 30% write and 70% read. The Postmark is a file-system benchmark to measure performance for the applications such as , netnews, and web-based commerce. A. Downtime All live migrations are expected to achieve minimal downtime which has minimal impact on the service. In general, 926

7 Fig. 4: The downtime of migration Fig. 5: The total migration time of five workloads Fig. 6: The amount of data transferred the max downtime is one second which is a common value set for downtime. We migrate the VM with different types of workload. From Figure 4, we can know that our prototype can decrease downtime on poly-type our selected application workloads. Our disk working set is set of blocks corresponding to the I/O read operations so that there would be less dirty blocks that have to be transferred to the target host during the migration. In the stop and copy stage, dirty blocks that have to be copied repeatedly are produced by write operations. The size of the intersection between the set of blocks corresponding to read operations and the set of blocks corresponding to write operations is smaller, the lower the number of dirty blocks synchronized is. However, when the dirty blocks are so much that the downtime is one second that is the maximum value. B. Migration time Migration time is the total time of virtual machine migration. Figure 5 shows the total migration time of virtual machines with five different workloads. Our results show that our migration times are less than the method of IO Mirroring in five workloads. But Two places in our implementation have extra time. One place is when a bitmap to mark disk working set generates at the beginning of migration. It will take some time to read the records and initialize the bitmap. The other is that some seconds will be consumed for starting the ondemand fetching service. As figure 6shows that we can reduce the transmission.only the block which belongs to the working set and is modified during the pre-copy stage has to be transferred for many times. The other blocks are only transferred for one time. C. Performance overhead Performance overhead should be minimized during migration. However, the evaluation of performance overhead during migration is quite complicated, because the overhead is influenced by the comprehensive influence of downtime, total migration time and overhead incurred during different stages of data transfer. On the one hand, our method incurs more overhead than IO Mirroring because we use more ondemand fetching whose response time is long. On the other hand, our method performs better since that it incurs shorter downtime and total migration. In this case, we demonstrate two experiments to evaluate the overhead. First, we run the same workload tasks with different migration method and compare their elapsed time. As is shown in Figure 7a, our method brings 5.4% more overhead than IO Mirroring with file compression workload. Compared with the promoted metrics mentioned above, it is acceptable. Second, we execute the IO intensive workload both when IO intercept on and off. Figure 7b shows that IO intercept only incurs 7.3% overhead. When the workload has less IO operations, the overhead will be even much lower. (a) Compress file during(b) Compress file when live VM migration IO intercept on and off Fig. 7: File compression time D. Limitation and Future work The experimental results show that our migration method reduces the downtime and migration time in some situations. In all experiments, downtime is reduced and the transmission is generally decreased compared with other methods. As for performance overhead, as the increase of hit rate, the overhead reduce. There is still space for improving. In future, we will validate our method in real-world applications. VII. RELATED WORK Virtual machine migration is no longer a new issue for the reason that living migration technologies in LAN are mature enough. And it has been implemented in all major hypervisors including Xen [9], KVM, Microsoft Hyper-V [6] and VMware ESX [5]. The major approach of these systems is a pre-copy that iteratively copies memory pages from the source host to destination host. The live blocks migration also has been 927

8 implemented in most major hypervisors. But, up to now, no methods have been widely accepted as one of the best methods for migration entire virtual machines. Live storage migration can be divided into three kinds of migration models. The major approaches are dirty block tracking and IO mirroring, both two we have introduced. Especially, IO mirroring is implemented in Microsoft Hyper- V, VMware ESX, and Xen with DRBD [24]. However, live storage migration of virtual machines using IO mirroring does not apply in KVM. In general, live virtual machine migration considers three respects, memory and device state, storage and network. So studies of live migration over WAN focused on either storage or network concerns. With the development of network technologies, the technologies of tunneling and network virtualization are alleviating the networking limitations that previously prevented moving VM across WAN. Such as Cisco OTV [2], VXLAN and SDN [21]. CloudNet [22] [23] put forward a cloud computing platform that seamlessly connects resources at multiple data center, and a centralized virtual private network. There are so many studies about pre-copy based storage migration. For example, Bradford [10] implements a disk block level iterative pre-copy method in Xen. It can support live storage migration in the WAN. zheng [25] had explored some optimal improvements on the per-copy model to migrate storage over WAN. And XvMotion [19] do some improvements in storage migration implemented by IO Mirroring in VMware ESX. The common optimization methods including compression, de-duplication, change the transport order and so on. XvMotion is deployed in VMware ESX using IO Mirroring and integrates support for memory and storage migration over the local and wide area. CloudNet addresses many of shortcomings of migration in Xen with DRBD. CloudNet use synchronous disk replication and XvMotion use asynchronous replication. VIII. CONCLUSION In this paper, we present a novel method to optimize the migration in hybrid clouds. We combine the strengths of pre-copy and post-copy migration model and also use disk working set to optimize virtual machines storage migration. We have designed and implemented our approach to the QEMU/KVM hypervisor and ran a series of experiments. The presented technique shows a reduction of up 20.5% on average of the total transfer time in most of the selected for some certain poly-type scenario. Our approach can also reduce the migration downtime in most cases. IX. ACKNOWLEDEMENT The work in this paper has been supported in part by the China 863 program under Grant No. 2013AA01A213. REFERENCES [3] Microsoft azure: Cloud computing platform and services. microsoft.com/. [4] Storage area network. [5] vsphere hypervisor. vsphere-hypervisor.html. [6] Windows server 2008 r2 hyper-v live migration. com/en-us/download/details.aspx?id=12601, [7] Qemu [8] K. A. Sysbench: a system performance benchmark. sourceforge.net. [9] P. Barham, B. Dragovic, K. Fraser, S. Hand, T. Harris, A. Ho, R. Neugebauer, I. Pratt, and A. Warfield. Xen and the art of virtualization. In ACM SIGOPS Operating Systems Review, volume 37, pages ACM, [10] R. Bradford, E. Kotsovinos, A. Feldmann, and H. Schiöberg. Live widearea migration of virtual machines including local persistent state. In Proceedings of the 3rd international conference on Virtual execution environments, pages ACM, [11] G. A. Gibson and R. Van Meter. Network attached storage architecture. Communications of the ACM, 43(11):37 45, [12] T. Hirofuchi, H. Ogawa, H. Nakada, S. Itoh, and S. Sekiguchi. A live storage migration mechanism over wan for relocatable virtual machine services on clouds. In Proceedings of the th IEEE/ACM International Symposium on Cluster Computing and the Grid, pages IEEE Computer Society, [13] J. Katcher. Postmark: A new file system benchmark. Technical report, Technical Report TR3022, Network Appliance, [14] A. Kivity, Y. Kamay, D. Laor, U. Lublin, and A. Liguori. kvm: the linux virtual machine monitor. In Proceedings of the Linux symposium, volume 1, pages , [15] Y. Luo. A three-phase algorithm for whole-system live migration of virtual machines. In Proceedings of the 2007 Asian technology information program s (ATIP s) 3rd workshop on High performance computing in China: solution approaches to impediments for high performance computing, pages ACM, [16] Y. Luo, B. Zhang, X. Wang, Z. Wang, Y. Sun, and H. Chen. Live and incremental whole-system migration of virtual machines using blockbitmap. In 2008 IEEE International Conference on Cluster Computing, pages IEEE, [17] M. Mahalingam, D. Dutt, K. Duda, P. Agarwal, L. Kreeger, T. Sridhar, M. Bursell, and C. Wright. Virtual extensible local area network (vxlan): A framework for overlaying virtualized layer 2 networks over layer 3 networks. Technical report, [18] A. Mashtizadeh, E. Celebi, T. Garfinkel, M. Cai, et al. The design and evolution of live storage migration in vmware esx. In USENIX ATC, volume 11, pages 1 14, [19] A. J. Mashtizadeh, M. Cai, G. Tarasuk-Levin, R. Koller, T. Garfinkel, and S. Setty. Xvmotion: Unified virtual machine migration over long distance. In 2014 USENIX Annual Technical Conference (USENIX ATC 14), pages , [20] N. McKeown, T. Anderson, H. Balakrishnan, G. Parulkar, L. Peterson, J. Rexford, S. Shenker, and J. Turner. Openflow: enabling innovation in campus networks. ACM SIGCOMM Computer Communication Review, 38(2):69 74, [21] M.-K. Shin, K.-H. Nam, and H.-J. Kim. Software-defined networking (sdn): A reference architecture and open apis. In 2012 International Conference on ICT Convergence (ICTC), pages IEEE, [22] T. Wood, K. Ramakrishnan, P. Shenoy, and J. Van der Merwe. Cloudnet: dynamic pooling of cloud resources by live wan migration of virtual machines. In ACM Sigplan Notices, volume 46, pages ACM, [23] T. Wood, K. Ramakrishnan, P. Shenoy, J. Van der Merwe, J. Hwang, G. Liu, and L. Chaufournier. Cloudnet: Dynamic pooling of cloud resources by live wan migration of virtual machines. IEEE/ACM Transactions on Networking (TON), 23(5): , [24] H. Xu, L. Xing, and R. Robidoux. Drbd: Dynamic reliability block diagrams for system reliability modelling. International Journal of Computers and Applications, 31(2): , [25] J. Zheng, T. S. E. Ng, and K. Sripanidkulchai. Workload-aware live storage migration for clouds. In ACM Sigplan Notices, volume 46, pages ACM, [1] Amazon elastic compute cloud. ls. [2] Cisco active network abstraction

Live Virtual Machine Migration with Efficient Working Set Prediction

Live Virtual Machine Migration with Efficient Working Set Prediction 2011 International Conference on Network and Electronics Engineering IPCSIT vol.11 (2011) (2011) IACSIT Press, Singapore Live Virtual Machine Migration with Efficient Working Set Prediction Ei Phyu Zaw

More information

Columbia University COMS W Virtual Machine Migration

Columbia University COMS W Virtual Machine Migration E-Migrate RESEARCH Columbia University COMS W6998-6 Virtual Machine Migration Kay Sripanidkulchai, IBM T.J. Watson Research Center October 13, 2010 Migration Technologies and Process Steps Virtualization

More information

CloudAP: Improving the QoS of Mobile Applications with Efficient VM Migration

CloudAP: Improving the QoS of Mobile Applications with Efficient VM Migration CloudAP: Improving the QoS of Mobile Applications with Efficient VM Migration Renyu Yang, Ph.D. Student School of Computing, Beihang University yangry@act.buaa.edu.cn In IEEE HPCC, Zhangjiajie, China,

More information

CloudNet: Dynamic Pooling of Cloud Resources by Live WAN Migration of Virtual Machines

CloudNet: Dynamic Pooling of Cloud Resources by Live WAN Migration of Virtual Machines CloudNet: Dynamic Pooling of Cloud Resources by Live WAN Migration of Virtual Machines Timothy Wood, Prashant Shenoy University of Massachusetts Amherst K.K. Ramakrishnan, and Jacobus Van der Merwe AT&T

More information

2012 Seventh International Conference on P2P, Parallel, Grid, Cloud and Internet Computing

2012 Seventh International Conference on P2P, Parallel, Grid, Cloud and Internet Computing 2012 Seventh International Conference on P2P, Parallel, Grid, Cloud and Internet Computing Present the Challenges in Eucalyptus Cloud Infrastructure for Implementing Virtual Machine Migration Technique

More information

COMS: Customer Oriented Migration Service

COMS: Customer Oriented Migration Service Boise State University ScholarWorks Computer Science Faculty Publications and Presentations Department of Computer Science 1-1-217 COMS: Customer Oriented Migration Service Kai Huang Boise State University

More information

Service Level Management for Iterative Pre-Copy Live Migration

Service Level Management for Iterative Pre-Copy Live Migration Service Level Management for Iterative Pre-Copy Live Migration Thomas Treutner, Helmut Hlavacs Research Group Entertainment Computing University of Vienna, Austria Email: firstname.lastname@univie.ac.at

More information

MODELING OF CPU USAGE FOR VIRTUALIZED APPLICATION

MODELING OF CPU USAGE FOR VIRTUALIZED APPLICATION e-issn 2455 1392 Volume 2 Issue 4, April 2016 pp. 644-651 Scientific Journal Impact Factor : 3.468 http://www.ijcter.com MODELING OF CPU USAGE FOR VIRTUALIZED APPLICATION Lochan.B 1, Divyashree B A 2 1

More information

Live Virtual Machine Migration with Bandwidth Dynamic Assignment

Live Virtual Machine Migration with Bandwidth Dynamic Assignment Advances in Engineering Research (AER), volume 130 5th International Conference on Frontiers of Manufacturing Science and Measuring Technology (FMSMT 2017) Live Virtual Machine Migration with Bandwidth

More information

RPFF: A Remote Page-fault Filter for Post-copy Live Migration

RPFF: A Remote Page-fault Filter for Post-copy Live Migration RPFF: A Remote Page-fault Filter for Post-copy Live Migration Kui Su, Wenzhi Chen, Guoxi Li, Zonghui Wang College of Computer Science Zhejiang University, Hangzhou, 3127, China Email: {sukuias12, chenwz,

More information

University of Alberta. Zhu Pang. Master of Science. Department of Computing Science

University of Alberta. Zhu Pang. Master of Science. Department of Computing Science University of Alberta HIGH PERFORMANCE LIVE MIGRATION OVER LOW-BANDWIDTH, HIGH-DELAY NETWORK WITH LOSS PREVENTION by Zhu Pang A thesis submitted to the Faculty of Graduate Studies and Research in partial

More information

CBase: A New Paradigm for Fast Virtual Machine Migration across Data Centers

CBase: A New Paradigm for Fast Virtual Machine Migration across Data Centers 2017 17th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing CBase: A New Paradigm for Fast Virtual Machine Migration across Data Centers Fei Zhang 1,2,3, Xiaoming Fu 2, and Ramin Yahyapour

More information

Improving CPU Performance of Xen Hypervisor in Virtualized Environment

Improving CPU Performance of Xen Hypervisor in Virtualized Environment ISSN: 2393-8528 Contents lists available at www.ijicse.in International Journal of Innovative Computer Science & Engineering Volume 5 Issue 3; May-June 2018; Page No. 14-19 Improving CPU Performance of

More information

Lecture 1: January 22

Lecture 1: January 22 CMPSCI 677 Distributed and Operating Systems Spring 2018 Lecture 1: January 22 Lecturer: Prashant Shenoy Scribe: Bin Wang 1.1 Introduction to the course The lecture started by outlining the administrative

More information

Research on Availability of Virtual Machine Hot Standby based on Double Shadow Page Tables

Research on Availability of Virtual Machine Hot Standby based on Double Shadow Page Tables International Conference on Computer, Networks and Communication Engineering (ICCNCE 2013) Research on Availability of Virtual Machine Hot Standby based on Double Shadow Page Tables Zhiyun Zheng, Huiling

More information

VM Migration, Containers (Lecture 12, cs262a)

VM Migration, Containers (Lecture 12, cs262a) VM Migration, Containers (Lecture 12, cs262a) Ali Ghodsi and Ion Stoica, UC Berkeley February 28, 2018 (Based in part on http://web.eecs.umich.edu/~mosharaf/slides/eecs582/w16/021516-junchenglivemigration.pptx)

More information

An Integration and Load Balancing in Data Centers Using Virtualization

An Integration and Load Balancing in Data Centers Using Virtualization An Integration and Load Balancing in Data Centers Using Virtualization USHA BELLAD #1 and JALAJA G *2 # Student M.Tech, CSE, B N M Institute of Technology, Bengaluru, India * Associate Professor, CSE,

More information

A three phase optimization method for precopy based VM live migration

A three phase optimization method for precopy based VM live migration DOI 10.1186/s40064-016-2642-2 RESEARCH Open Access A three phase optimization method for precopy based VM live migration Sangeeta Sharma * and Meenu Chawla *Correspondence: sanjsharma29@gmail.com Department

More information

Virtual machine architecture and KVM analysis D 陳彥霖 B 郭宗倫

Virtual machine architecture and KVM analysis D 陳彥霖 B 郭宗倫 Virtual machine architecture and KVM analysis D97942011 陳彥霖 B96902030 郭宗倫 Virtual machine monitor serves as an interface between hardware and software; no matter what kind of hardware under, software can

More information

Keywords: disk throughput, virtual machine, I/O scheduling, performance evaluation

Keywords: disk throughput, virtual machine, I/O scheduling, performance evaluation Simple and practical disk performance evaluation method in virtual machine environments Teruyuki Baba Atsuhiro Tanaka System Platforms Research Laboratories, NEC Corporation 1753, Shimonumabe, Nakahara-Ku,

More information

Nutanix Tech Note. Virtualizing Microsoft Applications on Web-Scale Infrastructure

Nutanix Tech Note. Virtualizing Microsoft Applications on Web-Scale Infrastructure Nutanix Tech Note Virtualizing Microsoft Applications on Web-Scale Infrastructure The increase in virtualization of critical applications has brought significant attention to compute and storage infrastructure.

More information

XvMotion: Unified Virtual Machine Migration over Long Distance

XvMotion: Unified Virtual Machine Migration over Long Distance XvMotion: Unified Virtual Machine Migration over Long Distance Ali José Mashtizadeh, Stanford University; Min Cai, Gabriel Tarasuk-Levin, and Ricardo Koller, VMware, Inc.; Tal Garfinkel; Sreekanth Setty,

More information

Converged Platforms and Solutions. Business Update and Portfolio Overview

Converged Platforms and Solutions. Business Update and Portfolio Overview Converged Platforms and Solutions Business Update and Portfolio Overview IT Drivers In Next 5 Years SCALE SCALE 30,000+ physical servers 500,000+ virtual servers Current tools won t work at this scale

More information

vsan Remote Office Deployment January 09, 2018

vsan Remote Office Deployment January 09, 2018 January 09, 2018 1 1. vsan Remote Office Deployment 1.1.Solution Overview Table of Contents 2 1. vsan Remote Office Deployment 3 1.1 Solution Overview Native vsphere Storage for Remote and Branch Offices

More information

Enhanced Live Migration of Virtual Machine Using Comparison of Modified and Unmodified Pages

Enhanced Live Migration of Virtual Machine Using Comparison of Modified and Unmodified Pages Available Online at www.ijcsmc.com International Journal of Computer Science and Mobile Computing A Monthly Journal of Computer Science and Information Technology IJCSMC, Vol. 3, Issue. 2, February 2014,

More information

Construction and Application of Cloud Data Center in University

Construction and Application of Cloud Data Center in University International Conference on Logistics Engineering, Management and Computer Science (LEMCS 2014) Construction and Application of Cloud Data Center in University Hong Chai Institute of Railway Technology,

More information

Virtualization. Part 1 Concepts & XEN

Virtualization. Part 1 Concepts & XEN Part 1 Concepts & XEN Concepts References and Sources James Smith, Ravi Nair, The Architectures of Virtual Machines, IEEE Computer, May 2005, pp. 32-38. Mendel Rosenblum, Tal Garfinkel, Virtual Machine

More information

Dynamic Translator-Based Virtualization

Dynamic Translator-Based Virtualization Dynamic Translator-Based Virtualization Yuki Kinebuchi 1,HidenariKoshimae 1,ShuichiOikawa 2, and Tatsuo Nakajima 1 1 Department of Computer Science, Waseda University {yukikine, hide, tatsuo}@dcl.info.waseda.ac.jp

More information

An Experimental Study of Load Balancing of OpenNebula Open-Source Cloud Computing Platform

An Experimental Study of Load Balancing of OpenNebula Open-Source Cloud Computing Platform An Experimental Study of Load Balancing of OpenNebula Open-Source Cloud Computing Platform A B M Moniruzzaman, StudentMember, IEEE Kawser Wazed Nafi Syed Akther Hossain, Member, IEEE & ACM Abstract Cloud

More information

VMware vsphere Clusters in Security Zones

VMware vsphere Clusters in Security Zones SOLUTION OVERVIEW VMware vsan VMware vsphere Clusters in Security Zones A security zone, also referred to as a DMZ," is a sub-network that is designed to provide tightly controlled connectivity to an organization

More information

Native vsphere Storage for Remote and Branch Offices

Native vsphere Storage for Remote and Branch Offices SOLUTION OVERVIEW VMware vsan Remote Office Deployment Native vsphere Storage for Remote and Branch Offices VMware vsan is the industry-leading software powering Hyper-Converged Infrastructure (HCI) solutions.

More information

Evaluate the Performance and Scalability of Image Deployment in Virtual Data Center

Evaluate the Performance and Scalability of Image Deployment in Virtual Data Center Evaluate the Performance and Scalability of Image Deployment in Virtual Data Center Kejiang Ye, Xiaohong Jiang, Qinming He, Xing Li, and Jianhai Chen College of Computer Science, Zhejiang University, Zheda

More information

The vsphere 6.0 Advantages Over Hyper- V

The vsphere 6.0 Advantages Over Hyper- V The Advantages Over Hyper- V The most trusted and complete virtualization platform SDDC Competitive Marketing 2015 Q2 VMware.com/go/PartnerCompete 2015 VMware Inc. All rights reserved. v3b The Most Trusted

More information

CSE 120 Principles of Operating Systems

CSE 120 Principles of Operating Systems CSE 120 Principles of Operating Systems Spring 2018 Lecture 16: Virtual Machine Monitors Geoffrey M. Voelker Virtual Machine Monitors 2 Virtual Machine Monitors Virtual Machine Monitors (VMMs) are a hot

More information

Lecture 1: January 23

Lecture 1: January 23 CMPSCI 677 Distributed and Operating Systems Spring 2019 Lecture 1: January 23 Lecturer: Prashant Shenoy Scribe: Jonathan Westin (2019), Bin Wang (2018) 1.1 Introduction to the course The lecture started

More information

vsan Security Zone Deployment First Published On: Last Updated On:

vsan Security Zone Deployment First Published On: Last Updated On: First Published On: 06-14-2017 Last Updated On: 11-20-2017 1 1. vsan Security Zone Deployment 1.1.Solution Overview Table of Contents 2 1. vsan Security Zone Deployment 3 1.1 Solution Overview VMware vsphere

More information

Enhancing the Performance of High Availability Lightweight Live Migration

Enhancing the Performance of High Availability Lightweight Live Migration Enhancing the Performance of High Availability Lightweight Live Migration Peng Lu 1, Binoy Ravindran 1, and Changsoo Kim 2 1 ECE Dept., Virginia Tech {lvpeng,binoy}@vt.edu 2 ETRI, Daejeon, South Korea

More information

On the Interplay between Network Traffic and Energy Consumption in Virtualized Environment: An Empirical Study

On the Interplay between Network Traffic and Energy Consumption in Virtualized Environment: An Empirical Study 14 IEEE International Conference on Cloud Computing On the Interplay between Network Traffic and Energy Consumption in Virtualized Environment: An Empirical Study Chi Xu Simon Fraser University British

More information

A Rank-based VM Consolidation Method for Power Saving in Datacenters

A Rank-based VM Consolidation Method for Power Saving in Datacenters Regular Paper A Rank-based VM Consolidation Method for Power Saving in Datacenters Shingo Takeda 1 and Toshinori Takemura 2 In this paper, we propose a simple but flexible virtual machine consolidation

More information

Baoping Wang School of software, Nanyang Normal University, Nanyang , Henan, China

Baoping Wang School of software, Nanyang Normal University, Nanyang , Henan, China doi:10.21311/001.39.7.41 Implementation of Cache Schedule Strategy in Solid-state Disk Baoping Wang School of software, Nanyang Normal University, Nanyang 473061, Henan, China Chao Yin* School of Information

More information

Xen and the Art of Virtualization

Xen and the Art of Virtualization Xen and the Art of Virtualization Paul Barham,, Boris Dragovic, Keir Fraser, Steven Hand, Tim Harris, Alex Ho, Rolf Neugebauer,, Ian Pratt, Andrew Warfield University of Cambridge Computer Laboratory Presented

More information

Virtual Machines Disco and Xen (Lecture 10, cs262a) Ion Stoica & Ali Ghodsi UC Berkeley February 26, 2018

Virtual Machines Disco and Xen (Lecture 10, cs262a) Ion Stoica & Ali Ghodsi UC Berkeley February 26, 2018 Virtual Machines Disco and Xen (Lecture 10, cs262a) Ion Stoica & Ali Ghodsi UC Berkeley February 26, 2018 Today s Papers Disco: Running Commodity Operating Systems on Scalable Multiprocessors, Edouard

More information

Datacenter replication solution with quasardb

Datacenter replication solution with quasardb Datacenter replication solution with quasardb Technical positioning paper April 2017 Release v1.3 www.quasardb.net Contact: sales@quasardb.net Quasardb A datacenter survival guide quasardb INTRODUCTION

More information

Carbonite Availability. Technical overview

Carbonite Availability. Technical overview Carbonite Availability Technical overview Table of contents Executive summary The availability imperative...3 True real-time replication More efficient and better protection... 4 Robust protection Reliably

More information

Cooperative VM Migration for a virtualized HPC Cluster with VMM-bypass I/O devices

Cooperative VM Migration for a virtualized HPC Cluster with VMM-bypass I/O devices Cooperative VM Migration for a virtualized HPC Cluster with VMM-bypass I/O devices Ryousei Takano, Hidemoto Nakada, Takahiro Hirofuchi, Yoshio Tanaka, and Tomohiro Kudoh Information Technology Research

More information

Research and Design of Crypto Card Virtualization Framework Lei SUN, Ze-wu WANG and Rui-chen SUN

Research and Design of Crypto Card Virtualization Framework Lei SUN, Ze-wu WANG and Rui-chen SUN 2016 International Conference on Wireless Communication and Network Engineering (WCNE 2016) ISBN: 978-1-60595-403-5 Research and Design of Crypto Card Virtualization Framework Lei SUN, Ze-wu WANG and Rui-chen

More information

Hop Onset Network: Adequate Stationing Schema for Large Scale Cloud- Applications

Hop Onset Network: Adequate Stationing Schema for Large Scale Cloud- Applications www.ijecs.in International Journal Of Engineering And Computer Science ISSN:2319-7242 Volume 6 Issue 10 October 2017, Page No. 22616-22626 Index Copernicus value (2015): 58.10 DOI: 10.18535/ijecs/v6i10.11

More information

Remote Software Service for Mobile Clients leveraging Cloud Computing

Remote Software Service for Mobile Clients leveraging Cloud Computing Remote Software Service for Mobile Clients leveraging Cloud Computing Dr. Chunming Hu (hucm@act.buaa.edu.cn ) In Cooperate with Dr. Jianxin Li, Mr. Weiren Yu Associate Professor, School of Computer Science

More information

What s New in VMware vsphere 4.1 Performance. VMware vsphere 4.1

What s New in VMware vsphere 4.1 Performance. VMware vsphere 4.1 What s New in VMware vsphere 4.1 Performance VMware vsphere 4.1 T E C H N I C A L W H I T E P A P E R Table of Contents Scalability enhancements....................................................................

More information

Hedvig as backup target for Veeam

Hedvig as backup target for Veeam Hedvig as backup target for Veeam Solution Whitepaper Version 1.0 April 2018 Table of contents Executive overview... 3 Introduction... 3 Solution components... 4 Hedvig... 4 Hedvig Virtual Disk (vdisk)...

More information

Counting Sort for the Live Migration of Virtual Machines

Counting Sort for the Live Migration of Virtual Machines Counting Sort for the Live Migration of Virtual Machines QingXin Zou 123 ZhiYu Hao 3 Xu Cui 4 XiaoChun Yun 13 YongZheng Zhang 3 1 Institute of Computing Technology, Chinese Academy of Sciences, Beijing,

More information

Using MySQL in a Virtualized Environment. Scott Seighman Systems Engineer Sun Microsystems

Using MySQL in a Virtualized Environment. Scott Seighman Systems Engineer Sun Microsystems Using MySQL in a Virtualized Environment Scott Seighman Systems Engineer Sun Microsystems 1 Agenda Virtualization Overview > Why Use Virtualization > Options > Considerations MySQL & Virtualization Best

More information

[This is not an article, chapter, of conference paper!]

[This is not an article, chapter, of conference paper!] http://www.diva-portal.org [This is not an article, chapter, of conference paper!] Performance Comparison between Scaling of Virtual Machines and Containers using Cassandra NoSQL Database Sogand Shirinbab,

More information

Nutanix White Paper. Hyper-Converged Infrastructure for Enterprise Applications. Version 1.0 March Enterprise Applications on Nutanix

Nutanix White Paper. Hyper-Converged Infrastructure for Enterprise Applications. Version 1.0 March Enterprise Applications on Nutanix Nutanix White Paper Hyper-Converged Infrastructure for Enterprise Applications Version 1.0 March 2015 1 The Journey to Hyper-Converged Infrastructure The combination of hyper-convergence and web-scale

More information

The Load Balancing Research of SDN based on Ant Colony Algorithm with Job Classification Wucai Lin1,a, Lichen Zhang2,b

The Load Balancing Research of SDN based on Ant Colony Algorithm with Job Classification Wucai Lin1,a, Lichen Zhang2,b 2nd Workshop on Advanced Research and Technology in Industry Applications (WARTIA 2016) The Load Balancing Research of SDN based on Ant Colony Algorithm with Job Classification Wucai Lin1,a, Lichen Zhang2,b

More information

Cisco Virtualized Workload Mobility Introduction

Cisco Virtualized Workload Mobility Introduction CHAPTER 1 The ability to move workloads between physical locations within the virtualized Data Center (one or more physical Data Centers used to share IT assets and resources) has been a goal of progressive

More information

Virtual WAN Optimization Controllers

Virtual WAN Optimization Controllers Virtual WAN Optimization Controllers vwan Virtual WAN Optimization Controllers accelerate applications, speed data transfers and reduce bandwidth costs using a combination of application, network and protocol

More information

A SDN-like Loss Recovery Solution in Application Layer Multicast Wenqing Lei 1, Cheng Ma 1, Xinchang Zhang 2, a, Lu Wang 2

A SDN-like Loss Recovery Solution in Application Layer Multicast Wenqing Lei 1, Cheng Ma 1, Xinchang Zhang 2, a, Lu Wang 2 5th International Conference on Information Engineering for Mechanics and Materials (ICIMM 2015) A SDN-like Loss Recovery Solution in Application Layer Multicast Wenqing Lei 1, Cheng Ma 1, Xinchang Zhang

More information

Implementation and Performance Evaluation of RAPID-Cache under Linux

Implementation and Performance Evaluation of RAPID-Cache under Linux Implementation and Performance Evaluation of RAPID-Cache under Linux Ming Zhang, Xubin He, and Qing Yang Department of Electrical and Computer Engineering, University of Rhode Island, Kingston, RI 2881

More information

Difference Engine: Harnessing Memory Redundancy in Virtual Machines (D. Gupta et all) Presented by: Konrad Go uchowski

Difference Engine: Harnessing Memory Redundancy in Virtual Machines (D. Gupta et all) Presented by: Konrad Go uchowski Difference Engine: Harnessing Memory Redundancy in Virtual Machines (D. Gupta et all) Presented by: Konrad Go uchowski What is Virtual machine monitor (VMM)? Guest OS Guest OS Guest OS Virtual machine

More information

VERITAS Volume Replicator. Successful Replication and Disaster Recovery

VERITAS Volume Replicator. Successful Replication and Disaster Recovery VERITAS Volume Replicator Successful Replication and Disaster Recovery V E R I T A S W H I T E P A P E R Table of Contents Introduction.................................................................................1

More information

OS Virtualization. Why Virtualize? Introduction. Virtualization Basics 12/10/2012. Motivation. Types of Virtualization.

OS Virtualization. Why Virtualize? Introduction. Virtualization Basics 12/10/2012. Motivation. Types of Virtualization. Virtualization Basics Motivation OS Virtualization CSC 456 Final Presentation Brandon D. Shroyer Types of Virtualization Process virtualization (Java) System virtualization (classic, hosted) Emulation

More information

How it can help your organisation

How it can help your organisation How it can help your organisation History Types of Virtualisation & Hypervisors Virtualisation Features Why Virtualisation? Virtualising Oracle Performance Licensing Support Cloud 1998 VMware founded by

More information

Virtual WAN Optimization Controllers

Virtual WAN Optimization Controllers acel E RA VA DATAS HEET Virtual WAN Optimization Controllers acelera VA Virtual WAN Optimization Controllers accelerate applications, speed data transfers and reduce bandwidth costs using a combination

More information

Research on Heterogeneous Communication Network for Power Distribution Automation

Research on Heterogeneous Communication Network for Power Distribution Automation 3rd International Conference on Material, Mechanical and Manufacturing Engineering (IC3ME 2015) Research on Heterogeneous Communication Network for Power Distribution Automation Qiang YU 1,a*, Hui HUANG

More information

Chapter 5. Large and Fast: Exploiting Memory Hierarchy

Chapter 5. Large and Fast: Exploiting Memory Hierarchy Chapter 5 Large and Fast: Exploiting Memory Hierarchy Principle of Locality Programs access a small proportion of their address space at any time Temporal locality Items accessed recently are likely to

More information

An Efficient Method by Using Prediction to Reduce Times in Live Migration of Virtual Machine

An Efficient Method by Using Prediction to Reduce Times in Live Migration of Virtual Machine An Efficient Method by Using Prediction to Reduce imes in Live Migration of Virtual Machine Boseob Kim, and ungchun Kim Computer cience and Engineering ept, ogang Univ, eoul, outh Korea Abstract - Cloud

More information

On Selecting the Right Optimizations for Virtual Machine Migration

On Selecting the Right Optimizations for Virtual Machine Migration On Selecting the Right Optimizations for Virtual Machine Migration Senthil Nathan, Umesh Bellur, and Purushottam Kulkarni Department of Computer Science and Engineering, IIT Bombay {cendhu,umesh,puru}@cse.iitb.ac.in

More information

Understanding Data Locality in VMware vsan First Published On: Last Updated On:

Understanding Data Locality in VMware vsan First Published On: Last Updated On: Understanding Data Locality in VMware vsan First Published On: 07-20-2016 Last Updated On: 09-30-2016 1 Table of Contents 1. Understanding Data Locality in VMware vsan 1.1.Introduction 1.2.vSAN Design

More information

Virtualization of the MS Exchange Server Environment

Virtualization of the MS Exchange Server Environment MS Exchange Server Acceleration Maximizing Users in a Virtualized Environment with Flash-Powered Consolidation Allon Cohen, PhD OCZ Technology Group Introduction Microsoft (MS) Exchange Server is one of

More information

The Continuity of Out-of-band Remote Management Across Virtual Machine Migration in Clouds

The Continuity of Out-of-band Remote Management Across Virtual Machine Migration in Clouds The Continuity of Out-of-band Remote Management Across Virtual Machine Migration in Clouds Sho Kawahara Department of Creative Informatics Kyushu Institute of Technology Fukuoka, Japan kawasho@ksl.ci.kyutech.ac.jp

More information

ABSTRACT I. INTRODUCTION

ABSTRACT I. INTRODUCTION International Journal of Scientific Research in Computer Science, Engineering and Information Technology 2018 IJSRCSEIT Volume 3 Issue 3 ISS: 2456-3307 Hadoop Periodic Jobs Using Data Blocks to Achieve

More information

Virtual Security Gateway Overview

Virtual Security Gateway Overview This chapter contains the following sections: Information About the Cisco Virtual Security Gateway, page 1 Cisco Virtual Security Gateway Configuration for the Network, page 10 Feature History for Overview,

More information

A Hybrid Architecture for Video Transmission

A Hybrid Architecture for Video Transmission 2017 Asia-Pacific Engineering and Technology Conference (APETC 2017) ISBN: 978-1-60595-443-1 A Hybrid Architecture for Video Transmission Qian Huang, Xiaoqi Wang, Xiaodan Du and Feng Ye ABSTRACT With the

More information

CLOUD COMPUTING IT0530. G.JEYA BHARATHI Asst.Prof.(O.G) Department of IT SRM University

CLOUD COMPUTING IT0530. G.JEYA BHARATHI Asst.Prof.(O.G) Department of IT SRM University CLOUD COMPUTING IT0530 G.JEYA BHARATHI Asst.Prof.(O.G) Department of IT SRM University What is virtualization? Virtualization is way to run multiple operating systems and user applications on the same

More information

HSG-LM: Hybrid-Copy Speculative Guest OS Live Migration without Hypervisor

HSG-LM: Hybrid-Copy Speculative Guest OS Live Migration without Hypervisor HSG-LM: Hybrid-Copy Speculative Guest OS Live Migration without Hypervisor Peng Lu, Antonio Barbalace, Binoy Ravindran Department of Electrical and Computer Engineering, Virginia Tech, USA lvpeng, antoniob,

More information

Storage access optimization with virtual machine migration during execution of parallel data processing on a virtual machine PC cluster

Storage access optimization with virtual machine migration during execution of parallel data processing on a virtual machine PC cluster Storage access optimization with virtual machine migration during execution of parallel data processing on a virtual machine PC cluster Shiori Toyoshima Ochanomizu University 2 1 1, Otsuka, Bunkyo-ku Tokyo

More information

Nested Virtualization and Server Consolidation

Nested Virtualization and Server Consolidation Nested Virtualization and Server Consolidation Vara Varavithya Department of Electrical Engineering, KMUTNB varavithya@gmail.com 1 Outline Virtualization & Background Nested Virtualization Hybrid-Nested

More information

SURVEY PAPER ON CLOUD COMPUTING

SURVEY PAPER ON CLOUD COMPUTING SURVEY PAPER ON CLOUD COMPUTING Kalpana Tiwari 1, Er. Sachin Chaudhary 2, Er. Kumar Shanu 3 1,2,3 Department of Computer Science and Engineering Bhagwant Institute of Technology, Muzaffarnagar, Uttar Pradesh

More information

COMPUTER ARCHITECTURE. Virtualization and Memory Hierarchy

COMPUTER ARCHITECTURE. Virtualization and Memory Hierarchy COMPUTER ARCHITECTURE Virtualization and Memory Hierarchy 2 Contents Virtual memory. Policies and strategies. Page tables. Virtual machines. Requirements of virtual machines and ISA support. Virtual machines:

More information

SaaSaMe Transport Workload Snapshot Export for. Alibaba Cloud

SaaSaMe Transport Workload Snapshot Export for. Alibaba Cloud SaaSaMe Transport Workload Snapshot Export for Alibaba Cloud Contents About This Document... 3 Revision History... 3 Workload Snapshot Export for Alibaba Cloud... 4 Workload Snapshot Export Feature...

More information

Two-Level Cooperation in Autonomic Cloud Resource Management

Two-Level Cooperation in Autonomic Cloud Resource Management Two-Level Cooperation in Autonomic Cloud Resource Management Giang Son Tran a, Alain Tchana b, Laurent Broto a, Daniel Hagimont a a ENSEEIHT University of Toulouse, Toulouse, France Email: {giang.tran,

More information

GFS: The Google File System

GFS: The Google File System GFS: The Google File System Brad Karp UCL Computer Science CS GZ03 / M030 24 th October 2014 Motivating Application: Google Crawl the whole web Store it all on one big disk Process users searches on one

More information

Online Optimization of VM Deployment in IaaS Cloud

Online Optimization of VM Deployment in IaaS Cloud Online Optimization of VM Deployment in IaaS Cloud Pei Fan, Zhenbang Chen, Ji Wang School of Computer Science National University of Defense Technology Changsha, 4173, P.R.China {peifan,zbchen}@nudt.edu.cn,

More information

Domain Level Page Sharing in Xen Virtual Machine Systems

Domain Level Page Sharing in Xen Virtual Machine Systems Domain Level Page Sharing in Xen Virtual Machine Systems Myeongjae Jeon, Euiseong Seo, Junghyun Kim, and Joonwon Lee CS Division, Korea Advanced Institute of Science and Technology {mjjeon,ses,joon}@calabkaistackr

More information

VMware vsphere Beginner s Guide

VMware vsphere Beginner s Guide The latest version of VMware s virtualization platform, vsphere 5, builds on the already solid foundation of. With the growth of cloud computing and the move from ESX to ESXi, it s imperative for IT pros

More information

Storage Optimization with Oracle Database 11g

Storage Optimization with Oracle Database 11g Storage Optimization with Oracle Database 11g Terabytes of Data Reduce Storage Costs by Factor of 10x Data Growth Continues to Outpace Budget Growth Rate of Database Growth 1000 800 600 400 200 1998 2000

More information

Acceleration Performance Tests for IBM Rational ClearCase Remote Client (CCRC)

Acceleration Performance Tests for IBM Rational ClearCase Remote Client (CCRC) Acceleration Performance Tests for IBM Rational ClearCase Remote Client (CCRC) Conducted by Certeon for IBM Ready for Rational CCRC Certification November 28 PAGE 1 Table of Contents Certeon Acceleration

More information

Live Migration of Virtualized Edge Networks: Analytical Modeling and Performance Evaluation

Live Migration of Virtualized Edge Networks: Analytical Modeling and Performance Evaluation Live Migration of Virtualized Edge Networks: Analytical Modeling and Performance Evaluation Walter Cerroni, Franco Callegati DEI University of Bologna, Italy Outline Motivations Virtualized edge networks

More information

StarWind Virtual SAN Free

StarWind Virtual SAN Free #1 HyperConverged Appliance for SMB and ROBO StarWind Virtual SAN Free Value Proposition For virtualization admins, IT geeks, bloggers, students or those who look to build and maintain virtualization infrastructure

More information

A Comparison of File. D. Roselli, J. R. Lorch, T. E. Anderson Proc USENIX Annual Technical Conference

A Comparison of File. D. Roselli, J. R. Lorch, T. E. Anderson Proc USENIX Annual Technical Conference A Comparison of File System Workloads D. Roselli, J. R. Lorch, T. E. Anderson Proc. 2000 USENIX Annual Technical Conference File System Performance Integral component of overall system performance Optimised

More information

VFS Interceptor: Dynamically Tracing File System Operations in real. environments

VFS Interceptor: Dynamically Tracing File System Operations in real. environments VFS Interceptor: Dynamically Tracing File System Operations in real environments Yang Wang, Jiwu Shu, Wei Xue, Mao Xue Department of Computer Science and Technology, Tsinghua University iodine01@mails.tsinghua.edu.cn,

More information

Research Article TSMC: A Novel Approach for Live Virtual Machine Migration

Research Article TSMC: A Novel Approach for Live Virtual Machine Migration Applied Mathematics, Article ID 297127, 7 pages http://dx.doi.org/1.1155/214/297127 Research Article TSMC: A Novel Approach for Live Virtual Machine Migration Jiaxing Song, Weidong Liu, Feiran Yin, and

More information

EaSync: A Transparent File Synchronization Service across Multiple Machines

EaSync: A Transparent File Synchronization Service across Multiple Machines EaSync: A Transparent File Synchronization Service across Multiple Machines Huajian Mao 1,2, Hang Zhang 1,2, Xianqiang Bao 1,2, Nong Xiao 1,2, Weisong Shi 3, and Yutong Lu 1,2 1 State Key Laboratory of

More information

Paperspace. Architecture Overview. 20 Jay St. Suite 312 Brooklyn, NY Technical Whitepaper

Paperspace. Architecture Overview. 20 Jay St. Suite 312 Brooklyn, NY Technical Whitepaper Architecture Overview Copyright 2016 Paperspace, Co. All Rights Reserved June - 1-2017 Technical Whitepaper Paperspace Whitepaper: Architecture Overview Content 1. Overview 3 2. Virtualization 3 Xen Hypervisor

More information

Centralization of Network using Openflow Protocol

Centralization of Network using Openflow Protocol Indian Journal of Science and Technology, Vol 8(S2), 165 170, January 2015 ISSN (Print) : 0974-6846 ISSN (Online) : 0974-5645 DOI : 10.17485/ijst/2015/v8iS2/61217 Centralization of Network using Openflow

More information

An Experimental Cloud Resource Broker System for Virtual Application Control with VM Allocation Scheme

An Experimental Cloud Resource Broker System for Virtual Application Control with VM Allocation Scheme An Experimental Cloud Resource Broker System for Virtual Application Control with VM Allocation Scheme Seong-Hwan Kim 1, Dong-Ki Kang 1, Ye Ren 1, Yong-Sung Park 1, Kyung-No Joo 1, Chan-Hyun Youn 1, YongSuk

More information

IN this letter we focus on OpenFlow-based network state

IN this letter we focus on OpenFlow-based network state 1 On the Impact of Networ State Collection on the Performance of SDN Applications Mohamed Aslan, and Ashraf Matrawy Carleton University, ON, Canada Abstract Intelligent and autonomous SDN applications

More information

Split Migration of Large Memory Virtual Machines

Split Migration of Large Memory Virtual Machines Split Migration of Large Memory Virtual Machines Masato Suetake Kyushu Institute of Technology masato@ksl.ci.kyutech.ac.jp Hazuki Kizu Kyushu Institute of Technology hazuki@ksl.ci.kyutech.ac.jp Kenichi

More information

CSCI 8530 Advanced Operating Systems. Part 19 Virtualization

CSCI 8530 Advanced Operating Systems. Part 19 Virtualization CSCI 8530 Advanced Operating Systems Part 19 Virtualization Virtualization This is a very old idea It appears in many different forms A variety of commercial products exist The idea has become hot again

More information