POUX: Performance Optimization Strategy for Cloud Platforms based on User Experience

Size: px
Start display at page:

Download "POUX: Performance Optimization Strategy for Cloud Platforms based on User Experience"

Transcription

1 POUX: Performance Optimization Strategy for Cloud Platforms based on User Experience Zhijin Qiu, Zhongwen Guo, Yanan Sun, Yingjian Liu and Yu Wang Ocean University of China, Qingdao, Shandong, China University of North Carolina at Charlotte, Charlotte, North Carolina, USA Alcatel-Lucent Shanghai Bell, Shanghai, China Taiyuan University of Technology, Taiyuan, Shanxi, China Abstract Cloud computing has been widely used in almost every areas of our life. System virtualization is one of the key technologies in cloud computing. However, it is a great challenge to achieve a balance among the virtual machines (s) in different physical machines (PMs) by migrating overloaded s to underloaded PMs while minimizing the number of migrations. Most of existing performance optimization strategies only concern about the hardware parameters (i.e., CPU, memory and I/O, etc.) of s, but the important user experience parameters (i.e., response time and through rate, etc.) have been ignored. We propose a novel Performance Optimization strategy based on User experience (POUX) for the cloud platform. To obtain the parameters of user experience, we design a management architecture of the cloud platform, and define the standardized interfaces between s and management center. Due to the unique characteristics of load balancing in the cloud platform, performance optimization problems are often NP-hard. Therefore, we propose a heuristic for automatic performance optimization, which plays a tradeoff among hardware utilization, user experience and the number of migrations. We use the CloudSim simulator and our deployed small-scale real-world testbed to evaluate the performance of POUX. Various experimental results have indicated that our cloud-based management architecture and performance optimization strategy not only significantly reduce the number of migrations, but also ensure a better user experience. I. INTRODUCTION Cloud computing has been emerging as the dominating Internet-based computing approach. It is flexible since the shared resources of hardware and software are provided to the remote users. Infrastructure as a Service (IaaS) is an important service type in cloud computing, which provides server, storage, and network hardware for users. To supply users with stable, efficient and economical computing resources, IaaS uses the virtualization technology to consolidate the underlying hardware resources, which can realize the dynamic scaling of virtualized resources. Virtualization polymerizes the computational power of a large number of independent computers, and smartly allocates them to improve the service flexibility of the cloud platform. Providers dynamic allocate computing resources to meet user needs, and avoid waste of computing resources. At present, the typical virtualization technologies mainly include ESX, Hyper-V, Xen, and K [1] [3]. Different from the traditional computer system architecture, virtualization technology adds the virtual machine monitor (M) between the operating system and hardware to deploy and manage s resources, which creates loose couplings between and PM. Multiple s can be deployed in the same PM. Each is assigned to independent CPU, memory and I/O resources from PM as separate computing nodes. Through M, s can be flexibly migrated among different PMs, and without impacting the user experience. When a provider leases services to a user, service level agreement (SLA) is required to ensure that the provider can provide services corresponding to the user s expense. It is a great challenge that providers are not only compliance with SLA, but also reduce the waste of cloud computing resources as far as possible. The requirements of the application running on the are constantly changing. Therefore, it is necessary to dynamically adjust the configuration properties of the according to the load and user experience of the on the cloud platform. Even more, the overloaded s need to be migrated on different PM, so that the entire cloud platform achieves balancing loads, reduces costs, and meets SLAs. Overall, how to effectively manage the cloud platform based on virtualization technology to achieve the high performance is critical but challenging. The current load balancing algorithms usually only concern about the hardware parameters of s and PMs, but ignore user experience, which is a more direct fact for performances. Another issue about current performance optimization solutions is that their evaluation functions are usually simple summations of different monitoring parameters. However, many overloaded s have only one hardware parameter out of the threshold, which may not affect user experience. Thus, the existing evaluation functions may be unfair. To solve these problems, this paper tries to perform the optimization over the cloud platform based on user experience. We put forward a performance optimization strategy, POUX, to ensure high performance of the cloud platform. This strategy includes methods of evaluating overload and PM, selecting s to migrate, and finding the most suitable destination PMs. It realizes the load balancing not only among different s of the same PM, but also among different PMs. This proposed heuristic strategy is based on user experience to

2 avoid only concerning the hardware parameters. To implement POUX based on the Xen system architecture, we also design a management architecture of the cloud platform to acquire, analyze and manage the performance of PMs and s on the cloud platform. The contributions of this paper are summarized as follows. We have designed a novel management architecture of the cloud platform, and developed a standardized communication interface, which can not only obtain each PM and s utilization of CPU, Memory and I/ O through the Xen s XAPI, but also derive the response time and throughput rate of applications in each through the UAPI interface of our development. Evaluation indicators of the proposed POUX include both parameters of hardware and user experience. So even if we increase the threshold of hardware parameters, the user experience is not affected and the number of migrations can be reduced. POUX uses a heuristic method to perform the migration strategy, which considers the different resource intensities of the overloaded and the destination PM, respectively. The efficiency of performance optimization, and the robustness of the cloud platform are improved due to the reduction in the number of migrations. In the process of migration, POUX improves the rate of migration by selecting the migrated that is greater than and closest to the overloaded resource size. The rest of this paper is organized as follows. Section 2 briefly describes the related work. In Section 3, the management architecture of the cloud platform is proposed, and the performance optimization problem is defined. We put forward a heuristic strategy POUX in Section 4. Section 5 implements and validates the management architecture and POUX. Finally, Section 6 summarizes the paper with remarks on possible future work. II. RELATED WORK With the development of virtualization technology, the flexibility of computing service makes the computer performance been fully utilized, and promotes the popularization of cloud computing. At present, the major virtualization technologies include Ware, K, Hyper-V and Xen [1] [3]. Even though different technologies adopt different methods and strategies in aspect of performance management, they all can divide the resources at the server and dynamically manage them according to load fluctuation. Ware is more mature virtualization product on the market, but it is expensive, non-open source, and relevant low performance. Kernel-based Virtual Machine (K) is a full virtualization solution for Linux on x86 hardware and open-source software, but it consumes more resources at the host. Hyper-V, formerly known as Windows Server Virtualization, is a native hypervisor. It can create s on x86-64 systems running Windows, but its applicability is poor in the Linux server. Xen Project is a hypervisor using a microkernel design, which provides services that allow multiple computer operating systems to execute on the same computer hardware concurrently. The Xen Project community develops and maintains Xen Project as free and open-source software. It also supports full virtualization and paravirtualization technology, of which is its unique advantages. Citrix XenServer based on Xen virtualization technology is a virtualization operating system, rather than the software deployed on the operating system. It does not require the support of the underlying native operating system compared with traditional software, so it can take full advantage of server performance. When the PM downtime or the load is unbalanced, the needs to be migrated. To improve the convenience of migration, iscsi [4], [5] and Network File System (NFS) [6], [7] are used in the cloud platform. When performing migrations, we only need to copy a small amount of data such as memory data, instead of migrating large amount of hard disk storage data. Currently, the methods of migration include static migration and live migration. Static migration method needs to pause the before migrating, then copies the data of system state to the destination PM, and finally rebuilds the s state on the destination PM. It is simple, but it requires to shut down when is migrated [8]. Clark et al. [9] and Sun et al. [1] propose live migration, which uses a memory pre-copy strategy to migrate s between different PMs without disrupting tasks on s. The downtime is reduced to milliseconds. However, live migration leads to performance loss and energy overhead that cannot be ignored in the modern cloud platforms, especially in the case with limited network bandwidth [11]. To achieve the load balance among PMs and maintain the highest performance of PMs, many methods have been proposed to deal with PM overloading problem through migrations [12], [13]. Wood et al. [14] present Sandpiper, a system that automates the task of monitoring and detecting hotspots, determining a new mapping of physical to virtual resources, resizing s to their new allocations, and initiating any necessary migrations. Sandpiper tries to move load from the most overloaded PMs to the most underloaded PMs. It defines volume for s and PMs: volume = (1/(1 u cpu )) (1/(1 u men )) (1/(1 u net )), where u is resource utilization. It also defines a volume-to-size ratio (VSR) for each : V SR = volume/size, where size is the memory footprint of the. It then migrates the with the maximum VSR to the PM with the least volume. Weng et al. [15] propose a management framework for the virtualized cluster system, and present an automatic performance tuning strategy to balance the workload in the virtualized cluster system. They adopt an exponential function for the cost of a with a given load. The cost of j running on node i is defined as: cost[i][j] = cost c [i][j]+cost m [i][j]+cost n [i][j], where cost c, cost m, cost n represent CPU cost, memory cost and network cost, respectively. However, all previous methods statically assume equal importance for different resources, which may not be correct due to the different time-varying demands on different resources in each PM. Chne et al.

3 Xen M Xen M XAPI UAPI Xen M Management PM Acquisition Processing Execute Monitoring Engine Dom- v1,1 v1,m1 Monitoring Engine Dom- v2,1 v2,m2 Node PM p1 Node PM p2 Machine Monitor (M), Dom- and multiple s. M is located between the operating system and the hardware, and in full control of allocating the physical resources to the [15]. Dom- has special privileges, like being able to access the hardware directly. We can create, configure, and destroy s through the Domain management and control tools located in Dom-. In particular, M enables dynamic reclamation of idle resources with more s. Each node PM contains multiple s. The application and the Monitoring Engine are deployed in each. Management PM manages the cluster through the XAPI interface, and collects the application information, which is obtained by the Monitoring Engine through the UAPI interface. Monitoring Engine Dom- vn,1 vn,mn Fig. 1. Management architecture of the cloud platform. [16] propose a Resource Intensity Aware Load balancing method (RIAL), which dynamically assigns different weights to different resources according to the utilization intensity of each in the PM. Wood et al. [14] and Weng et al. [15] propose a pre-load allocation based on the prediction of the future performance requirements. However, the above methods only focus on the utilization intensity of hardware resources, but ignores the user experience in different loads. In this paper, we aim to take the user experience into migrations. A. Architecture Design III. ARCHITECTURE OVERVIEW To implement PM migration algorithm, Wood et al. [14] and Weng et al. [15] put forward their own management architecture of the cloud platform. Similarly, in order to realize the performance optimization strategy based on user experience, as shown in Figure 1, we propose a management architecture of the cloud platform based on Xen. Management PM has three functions, which are acquisition of s status information (acquisition module), analysis of the status information (processing module) and scheduling (execute module). Acquisition module obtains each s CPU, Memory and I/O information through the official XAPI interface, and the running information of the application system in each through the our custom UAPI interface, such as the system response time, the number of task requests and the number of task responses. Processing module analyzes the information of status, finds overloaded s, and selects the migrant and destination PM according to the load balancing policy. Execute module performs a migration through the XAPI interface. Node PM is a PM as a computing node installed with the operating system of XenServer, which consists of Xen Virtual Node PM pn B. Architecture Description Each the cloud platform includes multiple PMs, each containing multiple s. The PM set is defined as P = {p 1, p2, p n }. The set for each PM is defined as V = {v i,1, v i,2, v i,mi }, where m i represents the number of s that p i contains. Ri,j,k v represents the capacity of each s hardware parameters. h v i,j,k (t) represents the utilization intensity of each s each hardware resources at time t, such as h v i,j,k (t) = {v i,j.cpu(t), v i,j.mem(t), v i,j.io(t)}, where k is the type of hardware parameter. For convenience, we always use hardware parameters to represent the hardware utilization intensity parameters. Tk v is the threshold for each hardware parameter of the, usually Tk v = 7%. Simultaneously, the hardware parameters of each PM are expressed as h p (t) at time t. R p and T p k represent the capacity and threshold of the hardware parameters of each PM, respectively. The sum of the hardware parameters of the in each PM is the amount of PM s hardware parameters, namely h p (t) = mi j=1 (hv i,j,k (t) Rv i,j,k )/Rp, where i = 1, 2, n. However, hardware parameters of the PM can be directly obtained through the XAPI interface, instead of using this equation, and. At the same time, we define u v i,j,k (t) to represent the user experience parameters of the v i,j at time t, including the response time u v i,j,1 = v i,j (t).t and throughput rate u v i,j,2 = v i,j(t).r of the application system. That is v i,j (t).t = Task.getFinishTime(t) - Task.getSubmissionTime(t) and v i,j (t).r = Task.getFinishingNumber(t) - Task.getRequestNumber(t). Tk u indicates the threshold of the user experience parameters. With regard to the threshold of system response time setting, as shown in Table 1, the 2/5/1s principle is widely accepted [17]. The threshold of the throughput rate is typically set to greater than 85%. When the throughput rate is TABLE I RESPONSE TIME AND CORRESPONDING USER EXPERIENCE. Response time Within 2s 2s-5s 5s-1s More than 1s User experience Speedy Accepted Anxiety, some user try to re-click Believe system had crashed

4 low, the system will try to request the task again, which will increase the system response time. C. Problem Formulation: Performance Optimization When the hardware parameters or the user experience parameters exceed their thresholds at certain, we need to migrate applications from that to eliminate the hotspot. However, migration can also cause performance degradation, and frequent migration may cause the crash of system and the awful of user experience. So the objective of our performance optimization is to reduce the number of migrations. We use O(t) to represent overloaded s at time t, and assume that under migration strategy M, there are M(O(t)) migrations. Then the performance optimization problem based on the user experience can be described as follows. T min M(O(t)) t=1 s.t. h v i,j,k(t ) T v k, v i,j V, t h p (t ) T p k, p i P, t u v i,j,k(t ) T u k, v i,j V, t. Here to represent the number of the migrations, T is the total time period, and t is the time after migration M(O(t)) but before t + 1. The first two constraints represent that the hardware parameters thresholds are satisfied at each and PM, respectively. The third constraint is to satisfy the threshold requirements of the user experience parameters. The objective function is to minimize the number of migrations. A simplified version of this problem can be related to the problem of multiple knapsack, which is NP-complete [18]. While several works suggest that the approximate solution can be obtained on the simplified problem only with constraints on hardware parameters, we are considering migrations based on user experience. IV. PERFORMANCE OPTIMIZATION STRATEGY In this section, we propose a new heuristics, POUX, which is based on the user experience to optimize the performance of the cloud platform. The description of POUX includes the definition of overloading, local optimization strategy and global optimization strategy. We will first provide each of them in detail and then present the overall POUX strategy. A. Defining Overloading In POUX, whether the is overloaded depends on h v i,j,k (t), uv i,j,k (t), T k v and T k u. Since hv i,j,k (t) and uv i,j,k (t) are often fluctuate, we use the single moving average (SMA) to calculate arithmetic moving averages of them. For example, h v i,j,k(σ) = h v i,j,k(t) = x=t x=t η+1 hv i,j,k (x). (1) η Hereafter, we use h v i,j,k (σ) to represent the arithmetic moving average of h v i,j,k (t) at time t. η is the window size for the moving average. Similarly, h p i,j,k (t) and uv i,j,k (t) can be converted to h p i,j,k (σ) and uv i,j,k (σ), respectively. The purpose of our performance optimization is to enhance the user experience, therefore, we also take user experience parameters u v i,j,k (σ) into consideration. We define the overloading based on the following two scenarios. Scenario 1: When there are user experience parameters that exceed their threshold, i.e. n k=1 (uv i,j,k (σ) > T k u) = 1 (regardless of any value of h v i,j,k (σ)), this is overloaded. Scenario 2: When there are hardware parameters that exceed their threshold (i.e., n k=1 (hv i,j,k (σ) > T k v ) = 1) but no user experience parameters that exceed their threshold (i.e., all n k=1 (uv i,j,k (σ) > T k u ) = ), this is also overloaded. Traditionally, in order to ensure a positive user experience, Tk v is set to a relatively low value, such as 7%. However, such low threshold significantly increases the number of migrations. Since our method has a separate constraint on user experience in Scenario 1, we can set Tk v to a higher value, such as 85%. Doing so not only can guarantee a good user experience, but also reduce the number of migrations. Consequently, the overloading of a is defined as follows. overloaded = { 1 h v i,j,k (σ) > T v k or uv i,j,k (σ) > T u k otherwise. B. Selecting Optimization Strategy When s are overloaded, performance optimization (i.e., migration) is required. We define two optimization s- trategies: local optimization and global optimization. When an overloading occurs, it is necessary to select one of these optimization strategies based on the decision condition. Whether the remaining resources of current PM can satisfy the overloaded s additional requirements for resources can be defined by the following equation. (R p T p k hp (σ)) (hv i,j,k(σ) R v i,j,k T v k ) > δ. (3) Here, δ is a tolerance value. If δ =, it indicates that the remaining resources of PM just satisfy with the additional resources needed by overloaded, but this PM may be unstable after migration due to fully usage of resources at PM. Thus, the stability of the system can be ensured by adjusting the value of δ. The selection of optimization strategies under the two overloaded scenarios is described as follows. In Scenario 1 ( n k=1 (uv i,j,k (σ) > T k u ) = 1), we analyze each of the hardware parameters h v i,j,k of this overloaded and choose the type of the most intensive utilization parameter among the k parameters. Here, only one hardware parameter is picked. If this parameter of the PM where this overloaded is located satisfies Eq. (3), local optimization is selected. Otherwise, global optimization is performed. In Scenario 2 ( n k=1 (hv i,j,k (σ) > T k v ) = 1), we analyze the hardware parameters that exceed the threshold. Here, multiple hardware parameters may be picked. If the parameter(s) of the PM where this overloaded is located still satisfies Eq. (3), (2)

5 Start Get the information of hardware and user experience Y Local Optimization Select the types of intensive resource Overloaded Y Selsct Strategy (Eq.3) Perform resource allocation or migration N N Global Optimization Select migrating and destination PM CPU Mem I/O (a) CPU Mem I/O (b) CPU Mem (c) Fig. 3. Resource allocation: (a) The overloaded, where the red part represents the capacity of overloaded resources; (b) Resource allocation 1 [(a) to (b)] where multiple resources are considered at the same time. When allocating the capacity for overloaded resources, it also expands unnecessary capacity for other underloaded resources. (c) Resource allocation 2 [(a) to (c)] which only consider the specific overload resources. Only the overload resource is increased and the other resources remain unchanged. When eliminating overloaded s, the latter method migrates less resources. I/O End Fig. 2. Execution process of POUX. local optimization is selected. Otherwise global optimization is performed. In summary, as shown in Figure 2, when the is overloaded the decision of which optimization strategy is selected is based on the following rule. { local, if overloaded & Eq.(3) holds strategy = global, C. Local Optimization R v,extend i,j,k = if overloaded & Eq.(3) does not hold. (4) When the is overloaded and Eq. (3) is satisfied (i.e., PM s idle resources can satisfy the needs of overloaded resource), local optimization is selected to eliminate the overloaded s. In different overloaded scenarios, the method of optimizing resources is different. In Scenario 1, the different types of hardware parameters of the overloaded are sorted according to n k=1 (uv i,j,k (σ) > T k u ) = 1 (i.e., ordering the hardware parameters which may cause user experience overloading), then the resource type k of the most intensive hardware parameters is selected as overloaded resource. In Scenario 2 where n k=1 (hv i,j,k (σ) > T k v ) = 1, the type of overload resource k is determined. Note that it could be multiple k overloaded. After determining the type of overloaded resource, the amount of resources R v,extend that i,j,k needs to be expanded is calculated. R v ( 1 1 i,j,k T u k u v i,j,k (σ)), Scenario 1 R v ( 1 1 i,j,k T v k h v i,j,k (σ) ), Scenario 2. (5) The Management PM commands M dynamically adds the k -type resources with the amount of R v,extend to the i,j,k overloaded. In this way, the overloaded is eliminated locally. As shown in Figure 3, local optimization uses resource redistribution for specific overloaded resources, so the number of dynamically added resources is less compared with when the multi-resource parameters are considered at the same time (a) (b) (c) Fig. 4. migration: (a) The overloaded 1 and other two s on the same PM; (b) Migration strategy 1 [(a) to (b)] shows that the overload 1 is migrated to eliminate overloaded 1. The amount of resources that are migrated is Rvm1 v hv vm1 (σ); (c) Migration strategy 2 [(a) to (c)] shows that the underloaded 2 is migrated to eliminate overloaded 1. The amount of resources that are migrated is Rvm2 v hv vm2 (σ). Obviously, Rvm2 v hv vm2 (σ) < Rv vm1 hv vm1 (σ). Therefore, the migration strategy 2 is more efficient, and has lower migration costs. In addition, local optimization eliminates overloaded s without migrations among PMs, which saves migration costs. D. Global Optimization When the local resources cannot satisfy Eq. (3), the global optimization strategy is executed. Global optimization is more complex than local optimization, mainly including selecting to migrate and selecting destination PM. As shown in Figure 4, on the PM where the is overloaded, it should choose with the minimum amount of resources to migrate, which can effectively reduce migration costs. The amount of resource R mig that needs to be migrated is defined as follows. Ri,j,k v ( 1 1 T u k u v i,j,k (σ) ) Rp 1 ( h p (σ) 1 ) + δ, T p k R mig = Scenario 1 R v i,j,k ( 1 T v k 1 h v i,j,k (σ) ) Rp ( 1 h p (σ) 1 T p k ) + δ, Scenario 2. (6) Similarly to local optimization, we first find the overloaded resource s type k, and then obtain s v satisfied i,j whose k source satisfies R v > i,j,k Rmig. These s belong to

6 Number of migrations POUX Non POUX Sandpier Time(hr) Fig. 5. Number of migrations in CloudSim. the same PM as the overloaded. v satisfied i,j is sorted in ascending order by R v to obtain i,j,k vsort i,j, which include all candidates for migration s. We search for the other PMs, in turn, and find PMs p satisfied i, which meet the k resources needs (i.e, (R p T p k hp (σ)) δ > Rmig, i). Next, i p satisfied i is sorted in ascending order according to R p to obtain p sort i as the candidates for destination PMs. Meanwhile, the remaining resources R p,idle = (R p T p k hp (σ)) δ, k k of each PM are recorded. Then we try to match a in vi,j sort to a PM in p sort i in order (from the beginning to the end). If R p,idle R mig, k is satisfied (i.e. a match k is found), then the to migrate v mig and the destination PM p des are obtained. After determining to be migrated and destination PM, Management PM commands M to perform migration through XAPI. If no match is found at the end, we have to add a new PM to meet overloaded s requirements for the resources. E. Overall POUX Strategy The detail algorithm of POUX is described as Algorithm 1. POUX is executed in the processing module of Management PM. As shown in Figure 2, after POUX obtains R v,extend i,j,k, v mig and p des, the execute module of Management PM commands M module to executes local optimization or global optimization through the XAPI. V. PERFORMANCE EVALUATION In this section, we use both the CloudSim simulator and a small-scale real-world testbed deployed by us to evaluate the performance of POUX. The implementation of POUX is based on XenServer. A. Simulation Evaluation In CloudSim, each PM has 1GHz 2-core CPU, 2GB memory, and 1GB/s network bandwidth. Each has 5Hz CPU, 512MB memory, and 1Mbit/s bandwidth. Initially, we simulate 4 s over 1 PMs, and four s are allocated on each PM. Later, according to the load, the resources of s will be dynamically adjusted. We will test migration Algorithm 1: POUX Strategy Input: PM-related parameters R p, hp (t), T p k ; -related parameters R v i,j,k, hv i,j,k (t), T v k ; and user experience-related parameters u v i,j,k (t), T u k. Output: The migrating strategy of the overloaded : v mig p des 1 while overloaded = 1 in Eq. (2) do 2 if Eq. (3) is satisfied then 3 Get R v,extend in Eq. (5); i,j,k 4 Perform resource allocation; 5 else 6 if n k=1 (hv i,j,k (σ) > T k v) = 1 then 7 Get the type k of overloaded resource; 8 end 9 if n k=1 (uv i,j,k (σ) > T k u) = 1 then 1 Choose the resource type k with the highest utilization rate; 11 end 12 Compute R mig in Eq. (6); 13 Get v satisfied i,j that satisfies R v i,j,k > Rmig 14 Sort v satisfied i,j R v, and obtain vsort i,j,k i,j ; 15 Get PM p satisfied i ; in ascending order according to that satisfies (R p T p k hp (σ)) δ > Rmig 16 Sort p satisfied i R p 17 Match v sort i,j, and obtain psort v mig p des ; 18 end 19 end, i; i in ascending order according to i ; in sequence with p sort i, obtain methods of POUX, Non-POUX and Sandpier [14]. Here, Non- POUX is a simplified version of POUX without considering user experience, and Sandpier is a strategy considering multiple parameters jointly. To facilitate the test, we only verify the CPU-intensive resources, by setting the utilization intensity of Memory and I/O below the threshold. The CPU resource consumption of each in the test instance uses the realtime load data in [16]. The CPU resource s threshold is set to 7% of non-user-experience-based strategy (Non-POUX) and 9% in POUX strategy. The threshold of the response time and the throughput rate (the ratio of the amount of requests and response times) are set to 3s and 8%, respectively. The evaluation parameter acquisition interval is set to 5 minutes, and the window size of the single moving average is set to 3. The M of Xen dynamic reclaims resources for less than 3%. The experiment runs the set of CPU resource consumption instances repeatedly within 12 hours. Number of Migrations: Figure 5 compares the number of migrations for POUX, Non-POUX and Sandpier [14] at the time 4h, 8h, and 12h. We can see that the number of migrations of Sandpier is more than those of POUX and Non-

7 25 2 POUX Non POUX Number of migrations 15 1 Response Time(s) Time(min) CPU Utilization(%) Fig. 6. Number of migrations in our testbed. Fig. 7. Response time of applications. POUX, since Sandpier considers multiple parameters at the same time. Non-POUX only considers the hardware parameters, resulting in unnecessary migrations, so the number of migrations is more than that of POUX. POUX considers both hardware parameters and user experience parameters at the same time, and aims to balance for specific overloaded resources, so fewer s are migrated. Ratio Hardware Overload User Experience Overload B. Testbed Evaluation To evaluate the management architecture and proposed POUX strategy on a real testbed, we built a small cluster, which includes six identical Dell PMs. XenServer 7. is deployed on each PM. All s run on Ubuntu server 16.1, which is based on the Linux release series 4.8. Each deploys jre-8u121, Apache Tomcat and Web Services, which implement the UAPI interface. In order to achieve the rapid migration s, a server acts as NFS. A notebook with Windows 1 OS acts as Management PM. We have developed a CPU-intensive web application, which is deployed in each. Meanwhile, a web client generates continuous requests of web access for each based on Httperf [19], [2]. Number of Migrations: Figure 6 shows the accumulated number of migrations over time. As POUX takes the user experience as the leading role in the migration and the hardware parameter s threshold is set to a higher value, the number of migrations of Non-POUX is much more than that of POUX as time increases. Response Time: On a single, we make the utilizations of CPU are 2%, 25%, up to 9% by increasing the amount of requests. As shown in Figure 7, the relationship between the CPU utilization and the application s response time is obtained, through calculating the average value of the application response time. The response time is stable in its early stage, but with the increase in CPU utilization, the response time grows rapidly. From 2% to 75%, the response time only increases by about 5ms, while up to 2s form 75% to 9%. Under the condition of neglecting Throughput Rate as the migration condition, Figure 8 compares the ratio of the number of migrations caused by overloading of user Response Time(s) Fig. 8. Ratio of the number of migrations caused by overloading of user experience and hardware. experience and hardware under the response times of 1s, 3s and 6s, respectively. Throughput Rate: On a single, we also increase the amount of CPU utilization from 2% to 9% by increasing the amount of requests. Figure 9 shows that the relationship between CPU utilization and throughput rate. clearly with more utilization the throughput rate goes down. When the CPU utilization is 7%, 8% and 9%, throughput rate is 83%, 7% and 3%, respectively. Throughput Rate CPU Utilization(%) Fig. 9. Throughput rate of applications.

8 VI. CONCLUSION Virtualization technology plays a key role in improving the performance of the cloud platform and saving energy through migrations. User experience is often overlooked when s are migrated. We propose a performance optimization strategy, POUX, for the cloud platform based on user experience. To implement this strategy, we design s management architecture of the cloud platform, obtain the information of hardware parameters through the official XAPI, and define the UAPI data interface for obtaining the information of user experience parameters. The user experience parameters refer to the response time and throughput rate of the application systems. Since user experience parameters are more direct description of the user s cloud platform requirements, POUX not only considers the hardware parameters, but also includes and focuses on the user experience parameters. In the POUX strategy, when the user experience parameters exceed the threshold, we analyze the hardware parameters, and find the resource constraints of the hardware parameters as conditions of migrations. Performance optimization is then performed for specific overloaded hardware parameters. Therefore, POUX can quickly eliminate overloaded s, and reduce the number of migrations. Evaluations via both the CloudSim simulator and our deployed small-scale real-world testbed show that our management architecture of the cloud platform and performance optimization strategy POUX are efficient and feasible. We plan to further study the relationship between the number of migrations and threshold setting of user experience parameters and hardware parameters to achieve optimal tradeoffs between overheads and effectiveness. We will also perform experiments with the simultaneous presence of multiple tasks with different loads, and further improve the performance optimization strategy. ACKNOWLEDGMENT Zhijin Qiu is supported by the fellowship from the China Scholarship Council (CSC) under No This work is also partially supported by the National Natural Science Foundation of China under Nos , and [6] A. Muthitacharoen, B. Chen, and D. Mazieres, A low-bandwidth network file system, in ACM SIGOPS Operating Systems Review, vol. 35, no. 5. ACM, 21, pp [7] P. Sahni and A. Batra, Network file system, International Journal of Research, vol. 2, no. 4, pp , 215. [8] T. S. Kang, M. Tsugawa, A. Matsunaga, T. Hirofuchi, and J. A. Fortes, Design and implementation of middleware for cloud disaster recovery via virtual machine migration management, in Proceedings of the 214 IEEE/ACM 7th International Conference on Utility and Cloud Computing. IEEE Computer Society, 214, pp [9] C. Clark, K. Fraser, S. Hand, J. G. Hansen, E. Jul, C. Limpach, I. Pratt, and A. Warfield, Live migration of virtual machines, in Proceedings of the 2nd Conference on Symposium on Networked Systems Design & Implementation-Volume 2. USENIX Association, 25, pp [1] G. Sun, D. Liao, V. Anand, D. Zhao, and H. Yu, A new technique for efficient live migration of multiple virtual machines, Future Generation Computer Systems, vol. 55, pp , 216. [11] A. Strunk, Costs of virtual machine live migration: A survey, in Proc. of 212 IEEE Eighth World Congress on Services (SERVICES),. IEEE, 212, pp [12] N. J. Kansal and I. Chana, Energy-aware virtual machine migration for cloud computing-a firefly optimization approach, Journal of Grid Computing, vol. 14, no. 2, pp , 216. [13] D. Sun, J. Zhang, W. Fan, T. Wang, C. Liu, and W. Huang, SPLM: security protection of live virtual machine migration in cloud computing, in Proceedings of the 4th ACM International Workshop on Security in Cloud Computing. ACM, 216, pp [14] T. Wood, P. Shenoy, A. Venkataramani, and M. Yousif, Sandpiper: Black-box and gray-box resource management for virtual machines, Computer Networks, vol. 53, no. 17, pp , 29. [15] C. Weng, M. Li, Z. Wang, and X. Lu, Automatic performance tuning for the virtualized cluster system, in Proc. of 29th IEEE International Conference on Distributed Computing Systems, ICDCS 9. IEEE, 29, pp [16] L. Chen, H. Shen, and K. Sapra, RIAL: Resource intensity aware load balancing in clouds, in Proceedings of IEEE INFOCOM, 214. IEEE, 214, pp [17] J. A. Hoxmeier and C. DiCesare, System response time and user satisfaction: An experimental study of browser-based applications, Proceedings of AMCIS 2, p. 347, 2. [18] M. J. Magazine and M.-S. Chern, A note on approximation schemes for multidimensional knapsack problems, Mathematics of Operations Research, vol. 9, no. 2, pp , [19] D. Mosberger and T. Jin, httperf: a tool for measuring web server performance, ACM SIGMETRICS Performance Evaluation Review, vol. 26, no. 3, pp , [2] S. Rajasekaran, Z. Ni, H. S. Chawla, N. Shah, T. Wood, and E. Berger, Scalable cloud security via asynchronous virtual machine introspection, in Proc. of 8th USENIX Workshop on Hot Topics in Cloud Computing (HotCloud 16), vol. 29. USENIX Association, 216, pp REFERENCES [1] P. Padala, X. Zhu, Z. Wang, S. Singhal, K. G. Shin et al., Performance evaluation of virtualization technologies for server consolidation, HP Labs Tec. Report, 27. [2] A. J. Younge, R. Henschel, J. T. Brown, G. Von Laszewski, J. Qiu, and G. C. Fox, Analysis of virtualization technologies for high performance computing environments, in Proc. of 211 IEEE International Conference on Cloud Computing (CLOUD). IEEE, 211, pp [3] N. Regola and J.-C. Ducom, Recommendations for virtualization technologies in high performance computing, in Cloud Computing Technology and Science (CloudCom), 21 IEEE Second International Conference on. IEEE, 21, pp [4] J. Satran, K. Meth et al., Internet small computer systems interface (iscsi), IETF RFC372, 24. [5] S. Aiken, D. Grunwald, A. R. Pleszkun, and J. Willeke, A performance analysis of the iscsi protocol, in Proceedings of 2th IEEE/11th NASA Goddard Conference on Mass Storage Systems and Technologies, (MSST 23). IEEE, 23, pp

SANDPIPER: BLACK-BOX AND GRAY-BOX STRATEGIES FOR VIRTUAL MACHINE MIGRATION

SANDPIPER: BLACK-BOX AND GRAY-BOX STRATEGIES FOR VIRTUAL MACHINE MIGRATION SANDPIPER: BLACK-BOX AND GRAY-BOX STRATEGIES FOR VIRTUAL MACHINE MIGRATION Timothy Wood, Prashant Shenoy, Arun Venkataramani, and Mazin Yousif * University of Massachusetts Amherst * Intel, Portland Data

More information

RIAL: Resource Intensity Aware Load Balancing in Clouds

RIAL: Resource Intensity Aware Load Balancing in Clouds RIAL: Resource Intensity Aware Load Balancing in Clouds Liuhua Chen and Haiying Shen and Karan Sapra Dept. of Electrical and Computer Engineering Clemson University, SC, USA 1 Outline Introduction System

More information

Double Threshold Based Load Balancing Approach by Using VM Migration for the Cloud Computing Environment

Double Threshold Based Load Balancing Approach by Using VM Migration for the Cloud Computing Environment www.ijecs.in International Journal Of Engineering And Computer Science ISSN:2319-7242 Volume 4 Issue 1 January 2015, Page No. 9966-9970 Double Threshold Based Load Balancing Approach by Using VM Migration

More information

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

Distributed Autonomous Virtual Resource Management in Datacenters Using Finite- Markov Decision Process

Distributed Autonomous Virtual Resource Management in Datacenters Using Finite- Markov Decision Process Distributed Autonomous Virtual Resource Management in Datacenters Using Finite- Markov Decision Process Liuhua Chen, Haiying Shen and Karan Sapra Department of Electrical and Computer Engineering Clemson

More information

Πποχωπημένη Κατανεμημένη Υπολογιστική

Πποχωπημένη Κατανεμημένη Υπολογιστική Πποχωπημένη Κατανεμημένη Υπολογιστική ΗΥ623 Διδάζκων Δημήηριος Καηζαρός @ Τμ. ΗΜΜΥ Πανεπιστήμιο Θεσσαλίαρ Διάλεξη 3η 1 Virtualization Concepts Definitions Virtualization A layer mapping its visible interface

More information

Black-box and Gray-box Strategies for Virtual Machine Migration

Black-box and Gray-box Strategies for Virtual Machine Migration Full Review On the paper Black-box and Gray-box Strategies for Virtual Machine Migration (Time required: 7 hours) By Nikhil Ramteke Sr. No. - 07125 1. Introduction Migration is transparent to application

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

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

Department of Information Technology Sri Venkateshwara College of Engineering, Chennai, India. 1 2

Department of Information Technology Sri Venkateshwara College of Engineering, Chennai, India. 1 2 Energy-Aware Scheduling Using Workload Consolidation Techniques in Cloud Environment 1 Sridharshini V, 2 V.M.Sivagami 1 PG Scholar, 2 Associate Professor Department of Information Technology Sri Venkateshwara

More information

A Load Balancing Approach to Minimize the Resource Wastage in Cloud Computing

A Load Balancing Approach to Minimize the Resource Wastage in Cloud Computing A Load Balancing Approach to Minimize the Resource Wastage in Cloud Computing Sachin Soni 1, Praveen Yadav 2 Department of Computer Science, Oriental Institute of Science and Technology, Bhopal, India

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

Figure 1: Virtualization

Figure 1: Virtualization Volume 6, Issue 9, September 2016 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com A Profitable

More information

Improving Throughput in Cloud Storage System

Improving Throughput in Cloud Storage System Improving Throughput in Cloud Storage System Chanho Choi chchoi@dcslab.snu.ac.kr Shin-gyu Kim sgkim@dcslab.snu.ac.kr Hyeonsang Eom hseom@dcslab.snu.ac.kr Heon Y. Yeom yeom@dcslab.snu.ac.kr Abstract Because

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

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

Supplementary File: Dynamic Resource Allocation using Virtual Machines for Cloud Computing Environment

Supplementary File: Dynamic Resource Allocation using Virtual Machines for Cloud Computing Environment IEEE TRANSACTION ON PARALLEL AND DISTRIBUTED SYSTEMS(TPDS), VOL. N, NO. N, MONTH YEAR 1 Supplementary File: Dynamic Resource Allocation using Virtual Machines for Cloud Computing Environment Zhen Xiao,

More information

Dynamic Data Placement Strategy in MapReduce-styled Data Processing Platform Hua-Ci WANG 1,a,*, Cai CHEN 2,b,*, Yi LIANG 3,c

Dynamic Data Placement Strategy in MapReduce-styled Data Processing Platform Hua-Ci WANG 1,a,*, Cai CHEN 2,b,*, Yi LIANG 3,c 2016 Joint International Conference on Service Science, Management and Engineering (SSME 2016) and International Conference on Information Science and Technology (IST 2016) ISBN: 978-1-60595-379-3 Dynamic

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

Research on Load Balancing in Task Allocation Process in Heterogeneous Hadoop Cluster

Research on Load Balancing in Task Allocation Process in Heterogeneous Hadoop Cluster 2017 2 nd International Conference on Artificial Intelligence and Engineering Applications (AIEA 2017) ISBN: 978-1-60595-485-1 Research on Load Balancing in Task Allocation Process in Heterogeneous Hadoop

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

LEEN: Locality/Fairness- Aware Key Partitioning for MapReduce in the Cloud

LEEN: Locality/Fairness- Aware Key Partitioning for MapReduce in the Cloud LEEN: Locality/Fairness- Aware Key Partitioning for MapReduce in the Cloud Shadi Ibrahim, Hai Jin, Lu Lu, Song Wu, Bingsheng He*, Qi Li # Huazhong University of Science and Technology *Nanyang Technological

More information

Priority-Aware Virtual Machine Selection Algorithm in Dynamic Consolidation

Priority-Aware Virtual Machine Selection Algorithm in Dynamic Consolidation Vol. 9, No., 208 Priority-Aware Virtual Machine Selection Algorithm in Dynamic Consolidation Hanan A. Nadeem, Mai A. Fadel 3 Computer Science Department Faculty of Computing & Information Technology King

More information

Construction Scheme for Cloud Platform of NSFC Information System

Construction Scheme for Cloud Platform of NSFC Information System , pp.200-204 http://dx.doi.org/10.14257/astl.2016.138.40 Construction Scheme for Cloud Platform of NSFC Information System Jianjun Li 1, Jin Wang 1, Yuhui Zheng 2 1 Information Center, National Natural

More information

Power Consumption of Virtual Machine Live Migration in Clouds. Anusha Karur Manar Alqarni Muhannad Alghamdi

Power Consumption of Virtual Machine Live Migration in Clouds. Anusha Karur Manar Alqarni Muhannad Alghamdi Power Consumption of Virtual Machine Live Migration in Clouds Anusha Karur Manar Alqarni Muhannad Alghamdi Content Introduction Contribution Related Work Background Experiment & Result Conclusion Future

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

An Optimized Virtual Machine Migration Algorithm for Energy Efficient Data Centers

An Optimized Virtual Machine Migration Algorithm for Energy Efficient Data Centers International Journal of Engineering Science Invention (IJESI) ISSN (Online): 2319 6734, ISSN (Print): 2319 6726 Volume 8 Issue 01 Ver. II Jan 2019 PP 38-45 An Optimized Virtual Machine Migration Algorithm

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

A Cool Scheduler for Multi-Core Systems Exploiting Program Phases

A Cool Scheduler for Multi-Core Systems Exploiting Program Phases IEEE TRANSACTIONS ON COMPUTERS, VOL. 63, NO. 5, MAY 2014 1061 A Cool Scheduler for Multi-Core Systems Exploiting Program Phases Zhiming Zhang and J. Morris Chang, Senior Member, IEEE Abstract Rapid growth

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

Profiling and Understanding Virtualization Overhead in Cloud

Profiling and Understanding Virtualization Overhead in Cloud Profiling and Understanding Virtualization Overhead in Cloud Liuhua Chen, Shilkumar Patel, Haiying Shen and Zhongyi Zhou Department of Electrical and Computer Engineering Clemson University, Clemson, South

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

Power-Aware Scheduling of Virtual Machines in DVFS-enabled Clusters

Power-Aware Scheduling of Virtual Machines in DVFS-enabled Clusters Power-Aware Scheduling of Virtual Machines in DVFS-enabled Clusters Gregor von Laszewski, Lizhe Wang, Andrew J. Younge, Xi He Service Oriented Cyberinfrastructure Lab Rochester Institute of Technology,

More information

An Efficient Virtual CPU Scheduling Algorithm for Xen Hypervisor in Virtualized Environment

An Efficient Virtual CPU Scheduling Algorithm for Xen Hypervisor in Virtualized Environment An Efficient Virtual CPU Scheduling Algorithm for Xen Hypervisor in Virtualized Environment Chia-Ying Tseng 1 and Po-Chun Huang 2 Department of Computer Science and Engineering, Tatung University #40,

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

IFRAT: An IoT Field Recognition Algorithm based on Time-series Data

IFRAT: An IoT Field Recognition Algorithm based on Time-series Data IFRAT: An IoT Field Algorithm based on Time-series Data Shuai Guo, Zhongwen Guo, Zhijin Qiu, Yingjian Liu and Yu Wang Ocean University of China, Qingdao, Shandong, China University of North Carolina at

More information

A virtual machine migration Algorithm Based on Network flow balance YangYu 1, a, ZhouHua 2,b, LiuJunHui 3,c and FengYun 4,d*

A virtual machine migration Algorithm Based on Network flow balance YangYu 1, a, ZhouHua 2,b, LiuJunHui 3,c and FengYun 4,d* Advanced Materials Research Submitted: 2014-06-25 ISSN: 1662-8985, Vols. 1044-1045, pp 1028-1034 Accepted: 2014-08-11 doi:10.4028/www.scientific.net/amr.1044-1045.1028 Online: 2014-10-01 2014 Trans Tech

More information

A New Approach to Ant Colony to Load Balancing in Cloud Computing Environment

A New Approach to Ant Colony to Load Balancing in Cloud Computing Environment A New Approach to Ant Colony to Load Balancing in Cloud Computing Environment Hamid Mehdi Department of Computer Engineering, Andimeshk Branch, Islamic Azad University, Andimeshk, Iran Hamidmehdi@gmail.com

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

Efficient Migration A Leading Solution for Server Consolidation

Efficient Migration A Leading Solution for Server Consolidation Efficient Migration A Leading Solution for Server Consolidation R Suchithra, MCA Department Jain University Bangalore,India N.Rajkumar, PhD. Department of Software Engineering Ramakrishna College of Engineering

More information

PERFORMANCE CONSTRAINT AND POWER-AWARE ALLOCATION FOR USER REQUESTS IN VIRTUAL COMPUTING LAB

PERFORMANCE CONSTRAINT AND POWER-AWARE ALLOCATION FOR USER REQUESTS IN VIRTUAL COMPUTING LAB PERFORMANCE CONSTRAINT AND POWER-AWARE ALLOCATION FOR USER REQUESTS IN VIRTUAL COMPUTING LAB Nguyen Quang Hung, Nam Thoai, Nguyen Thanh Son Ho Chi Minh City University of Technology, Vietnam Corresponding

More information

Increasing Cloud Power Efficiency through Consolidation Techniques

Increasing Cloud Power Efficiency through Consolidation Techniques Increasing Cloud Power Efficiency through Consolidation Techniques Antonio Corradi, Mario Fanelli, Luca Foschini Dipartimento di Elettronica, Informatica e Sistemistica (DEIS) University of Bologna, Italy

More information

Virtualization. Michael Tsai 2018/4/16

Virtualization. Michael Tsai 2018/4/16 Virtualization Michael Tsai 2018/4/16 What is virtualization? Let s first look at a video from VMware http://www.vmware.com/tw/products/vsphere.html Problems? Low utilization Different needs DNS DHCP Web

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

Survey on MapReduce Scheduling Algorithms

Survey on MapReduce Scheduling Algorithms Survey on MapReduce Scheduling Algorithms Liya Thomas, Mtech Student, Department of CSE, SCTCE,TVM Syama R, Assistant Professor Department of CSE, SCTCE,TVM ABSTRACT MapReduce is a programming model used

More information

CHAPTER 6 ENERGY AWARE SCHEDULING ALGORITHMS IN CLOUD ENVIRONMENT

CHAPTER 6 ENERGY AWARE SCHEDULING ALGORITHMS IN CLOUD ENVIRONMENT CHAPTER 6 ENERGY AWARE SCHEDULING ALGORITHMS IN CLOUD ENVIRONMENT This chapter discusses software based scheduling and testing. DVFS (Dynamic Voltage and Frequency Scaling) [42] based experiments have

More information

RIAL: Resource Intensity Aware Load Balancing in Clouds

RIAL: Resource Intensity Aware Load Balancing in Clouds : Resource Intensity Aware Load Balancing in Clouds Liuhua Chen, Haiying Shen, Karan Sapra Department of Electrical and Computer Engineering Clemson University, Clemson, South Carolina 29634 Email: {liuhuac,shenh,ksapra}@clemson.edu

More information

CoolCloud: A Practical Dynamic Virtual Machine Placement Framework for Energy Aware Data Centers

CoolCloud: A Practical Dynamic Virtual Machine Placement Framework for Energy Aware Data Centers 2015 IEEE 8th International Conference on Cloud Computing CoolCloud: A Practical Dynamic Virtual Machine Placement Framework for Energy Aware Data Centers Zhiming Zhang Iowa State University Ames IA, USA

More information

The Memetic Algorithm for The Minimum Spanning Tree Problem with Degree and Delay Constraints

The Memetic Algorithm for The Minimum Spanning Tree Problem with Degree and Delay Constraints The Memetic Algorithm for The Minimum Spanning Tree Problem with Degree and Delay Constraints Minying Sun*,Hua Wang* *Department of Computer Science and Technology, Shandong University, China Abstract

More information

Load Balancing Algorithm over a Distributed Cloud Network

Load Balancing Algorithm over a Distributed Cloud Network Load Balancing Algorithm over a Distributed Cloud Network Priyank Singhal Student, Computer Department Sumiran Shah Student, Computer Department Pranit Kalantri Student, Electronics Department Abstract

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

Unit 5: Distributed, Real-Time, and Multimedia Systems

Unit 5: Distributed, Real-Time, and Multimedia Systems Unit 5: Distributed, Real-Time, and Multimedia Systems Unit Overview Unit 5 provides an extension to the core topics of operating systems. It introduces distributed systems and special-purpose operating

More information

PROFILING BASED REDUCE MEMORY PROVISIONING FOR IMPROVING THE PERFORMANCE IN HADOOP

PROFILING BASED REDUCE MEMORY PROVISIONING FOR IMPROVING THE PERFORMANCE IN HADOOP ISSN: 0976-2876 (Print) ISSN: 2250-0138 (Online) PROFILING BASED REDUCE MEMORY PROVISIONING FOR IMPROVING THE PERFORMANCE IN HADOOP T. S. NISHA a1 AND K. SATYANARAYAN REDDY b a Department of CSE, Cambridge

More information

Chapter 3 Virtualization Model for Cloud Computing Environment

Chapter 3 Virtualization Model for Cloud Computing Environment Chapter 3 Virtualization Model for Cloud Computing Environment This chapter introduces the concept of virtualization in Cloud Computing Environment along with need of virtualization, components and characteristics

More information

EMC Business Continuity for Microsoft Applications

EMC Business Continuity for Microsoft Applications EMC Business Continuity for Microsoft Applications Enabled by EMC Celerra, EMC MirrorView/A, EMC Celerra Replicator, VMware Site Recovery Manager, and VMware vsphere 4 Copyright 2009 EMC Corporation. All

More information

ENERGY EFFICIENT VIRTUAL MACHINE INTEGRATION IN CLOUD COMPUTING

ENERGY EFFICIENT VIRTUAL MACHINE INTEGRATION IN CLOUD COMPUTING ENERGY EFFICIENT VIRTUAL MACHINE INTEGRATION IN CLOUD COMPUTING Mrs. Shweta Agarwal Assistant Professor, Dept. of MCA St. Aloysius Institute of Technology, Jabalpur(India) ABSTRACT In the present study,

More information

Computer Networks 53 (2009) Contents lists available at ScienceDirect. Computer Networks. journal homepage:

Computer Networks 53 (2009) Contents lists available at ScienceDirect. Computer Networks. journal homepage: Computer Networks 53 (29) 2923 2938 Contents lists available at ScienceDirect Computer Networks journal homepage: www.elsevier.com/locate/comnet Sandpiper: Black-box and gray-box resource management for

More information

Bio-Inspired Techniques for the Efficient Migration of Virtual Machine for Load Balancing In Cloud Computing

Bio-Inspired Techniques for the Efficient Migration of Virtual Machine for Load Balancing In Cloud Computing Volume 118 No. 24 2018 ISSN: 1314-3395 (on-line version) url: http://www.acadpubl.eu/hub/ http://www.acadpubl.eu/hub/ Bio-Inspired Techniques for the Efficient Migration of Virtual Machine for Load Balancing

More information

A QoS Load Balancing Scheduling Algorithm in Cloud Environment

A QoS Load Balancing Scheduling Algorithm in Cloud Environment A QoS Load Balancing Scheduling Algorithm in Cloud Environment Sana J. Shaikh *1, Prof. S.B.Rathod #2 * Master in Computer Engineering, Computer Department, SAE, Pune University, Pune, India # Master in

More information

The Design and Implementation of Disaster Recovery in Dual-active Cloud Center

The Design and Implementation of Disaster Recovery in Dual-active Cloud Center International Conference on Information Sciences, Machinery, Materials and Energy (ICISMME 2015) The Design and Implementation of Disaster Recovery in Dual-active Cloud Center Xiao Chen 1, a, Longjun Zhang

More information

MAPREDUCE FOR BIG DATA PROCESSING BASED ON NETWORK TRAFFIC PERFORMANCE Rajeshwari Adrakatti

MAPREDUCE FOR BIG DATA PROCESSING BASED ON NETWORK TRAFFIC PERFORMANCE Rajeshwari Adrakatti International Journal of Computer Engineering and Applications, ICCSTAR-2016, Special Issue, May.16 MAPREDUCE FOR BIG DATA PROCESSING BASED ON NETWORK TRAFFIC PERFORMANCE Rajeshwari Adrakatti 1 Department

More information

Hybrid Auto-scaling of Multi-tier Web Applications: A Case of Using Amazon Public Cloud

Hybrid Auto-scaling of Multi-tier Web Applications: A Case of Using Amazon Public Cloud Hybrid Auto-scaling of Multi-tier Web Applications: A Case of Using Amazon Public Cloud Abid Nisar, Waheed Iqbal, Fawaz S. Bokhari, and Faisal Bukhari Punjab University College of Information and Technology,Lahore

More information

Dynamic Resource Allocation on Virtual Machines

Dynamic Resource Allocation on Virtual Machines Dynamic Resource Allocation on Virtual Machines Naveena Anumala VIT University, Chennai 600048 anumala.naveena2015@vit.ac.in Guide: Dr. R. Kumar VIT University, Chennai -600048 kumar.rangasamy@vit.ac.in

More information

A Novel Energy Efficient Algorithm for Cloud Resource Management. Jing SiYuan. Received April 2013; revised April 2013

A Novel Energy Efficient Algorithm for Cloud Resource Management. Jing SiYuan. Received April 2013; revised April 2013 International Journal of Knowledge www.iklp.org and Language Processing KLP International c2013 ISSN 2191-2734 Volume 4, Number 2, 2013 pp.12-22 A Novel Energy Efficient Algorithm for Cloud Resource Management

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

A Novel Self-Adaptive VM Consolidation Strategy Using Dynamic Multi-Thresholds in IaaS Clouds

A Novel Self-Adaptive VM Consolidation Strategy Using Dynamic Multi-Thresholds in IaaS Clouds future internet Article A Novel Self-Adaptive VM Consolidation Strategy Using Dynamic Multi-Thresholds in IaaS Clouds Lei Xie 1,2, *, Shengbo Chen 1,2, Wenfeng Shen 1,3 and Huaikou Miao 1,2 1 School Computer

More information

Self-Adaptive Consolidation of Virtual Machines For Energy-Efficiency in the Cloud

Self-Adaptive Consolidation of Virtual Machines For Energy-Efficiency in the Cloud Self-Adaptive Consolidation of Virtual Machines For Energy-Efficiency in the Cloud Guozhong Li, Yaqiu Jiang,Wutong Yang, Chaojie Huang School of Information and Software Engineering University of Electronic

More information

Parallels Virtuozzo Containers

Parallels Virtuozzo Containers Parallels Virtuozzo Containers White Paper Deploying Application and OS Virtualization Together: Citrix and Parallels Virtuozzo Containers www.parallels.com Version 1.0 Table of Contents The Virtualization

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

A Model-based Application Autonomic Manager with Fine Granular Bandwidth Control

A Model-based Application Autonomic Manager with Fine Granular Bandwidth Control A Model-based Application Autonomic Manager with Fine Granular Bandwidth Control Nasim Beigi-Mohammadi, Mark Shtern, and Marin Litoiu Department of Computer Science, York University, Canada Email: {nbm,

More information

Virtual Machine Migration in an Over-committed Cloud

Virtual Machine Migration in an Over-committed Cloud Virtual Machine Migration in an Over-committed Cloud Xiangliang Zhang, Zon-Yin Shae, Shuai Zheng, and Hani Jamjoom King Abdullah University of Science and Technology (KAUST), Saudi Arabia IBM T. J. Watson

More information

Consolidation Assessment Final Report

Consolidation Assessment Final Report Consolidation Assessment Final Report January 2009 The foundation for a lasting relationship starts with a connection. 1.800.800.0014 biz.pcconnection.com Table of Contents Executive Overview..............................................

More information

The only open-source type-1 hypervisor

The only open-source type-1 hypervisor Monika Danikáková What is Xen? The only open-source type-1 hypervisor For Unix and Unix-like OS Linux, NetBSD and OpenSolaris From ancient greek term Xenos (ξένος), guest-friends Developed by the University

More information

Comparing Function of OpenStack and VMware. Li-qiu JIANG and Da-wei XIAO

Comparing Function of OpenStack and VMware. Li-qiu JIANG and Da-wei XIAO 2017 2nd International Conference on Wireless Communication and Network Engineering (WCNE 2017) ISBN: 978-1-60595-531-5 Comparing Function of OpenStack and VMware Li-qiu JIANG and Da-wei XIAO Institute

More information

Migration Management in Sensor-Cloud Networks

Migration Management in Sensor-Cloud Networks Journal of Computer and Knowledge Engineering, Vol. 1, No.2. 2018. DOI: 10.22067/cke.v1i2.57729 Migration Management in Sensor-Cloud Networks Farahnaz Farazestanian Abstract: Placement of virtual sensors

More information

Adaptive replica consistency policy for Kafka

Adaptive replica consistency policy for Kafka Adaptive replica consistency policy for Kafka Zonghuai Guo 1,2,*, Shiwang Ding 1,2 1 Chongqing University of Posts and Telecommunications, 400065, Nan'an District, Chongqing, P.R.China 2 Chongqing Mobile

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

The Performance Analysis of a Service Deployment System Based on the Centralized Storage

The Performance Analysis of a Service Deployment System Based on the Centralized Storage The Performance Analysis of a Service Deployment System Based on the Centralized Storage Zhu Xu Dong School of Computer Science and Information Engineering Zhejiang Gongshang University 310018 Hangzhou,

More information

Energy Efficient in Cloud Computing

Energy Efficient in Cloud Computing Energy Efficient in Cloud Computing Christoph Aschberger Franziska Halbrainer May 24, 2013 1 of 25 Introduction Energy consumption by Google 2011: 2,675,898 MWh. We found that we use roughly as much electricity

More information

High Performance Computing Cloud - a PaaS Perspective

High Performance Computing Cloud - a PaaS Perspective a PaaS Perspective Supercomputer Education and Research Center Indian Institute of Science, Bangalore November 2, 2015 Overview Cloud computing is emerging as a latest compute technology Properties of

More information

VMware vsphere with ESX 4.1 and vcenter 4.1

VMware vsphere with ESX 4.1 and vcenter 4.1 QWERTYUIOP{ Overview VMware vsphere with ESX 4.1 and vcenter 4.1 This powerful 5-day class is an intense introduction to virtualization using VMware s vsphere 4.1 including VMware ESX 4.1 and vcenter.

More information

MIDDLEWARE FOR AUTONOMOUS RECONFIGURATION OF VIRTUAL MACHINES

MIDDLEWARE FOR AUTONOMOUS RECONFIGURATION OF VIRTUAL MACHINES MIDDLEWARE FOR AUTONOMOUS RECONFIGURATION OF VIRTUAL MACHINES By Qingling Wang A Thesis Submitted to the Graduate Faculty of Rensselaer Polytechnic Institute in Partial Fulfillment of the Requirements

More information

Energy-Aware Dynamic Load Balancing of Virtual Machines (VMs) in Cloud Data Center with Adaptive Threshold (AT) based Migration

Energy-Aware Dynamic Load Balancing of Virtual Machines (VMs) in Cloud Data Center with Adaptive Threshold (AT) based Migration Khushbu Maurya et al, International Journal of Computer Science and Mobile Computing, Vol.4 Issue.12, December- 215, pg. 1-7 Available Online at www.ijcsmc.com International Journal of Computer Science

More information

Experimental Study of Virtual Machine Migration in Support of Reservation of Cluster Resources

Experimental Study of Virtual Machine Migration in Support of Reservation of Cluster Resources Experimental Study of Virtual Machine Migration in Support of Reservation of Cluster Resources Ming Zhao, Renato J. Figueiredo Advanced Computing and Information Systems (ACIS) Electrical and Computer

More information

A Hybrid Approach to CAM-Based Longest Prefix Matching for IP Route Lookup

A Hybrid Approach to CAM-Based Longest Prefix Matching for IP Route Lookup A Hybrid Approach to CAM-Based Longest Prefix Matching for IP Route Lookup Yan Sun and Min Sik Kim School of Electrical Engineering and Computer Science Washington State University Pullman, Washington

More information

VMware vsphere 4. The Best Platform for Building Cloud Infrastructures

VMware vsphere 4. The Best Platform for Building Cloud Infrastructures Table of Contents Get the efficiency and low cost of cloud computing with uncompromising control over service levels and with the freedom of choice................ 3 Key Benefits........................................................

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

VM Migration Planning in Software-Defined Data Center Networks

VM Migration Planning in Software-Defined Data Center Networks 206 IEEE 8th International Conference on High Performance Computing and Communications; IEEE 4th International Conference on Smart City; IEEE 2nd International Conference on Data Science and Systems Migration

More information

VMware and Xen Hypervisor Performance Comparisons in Thick and Thin Provisioned Environments

VMware and Xen Hypervisor Performance Comparisons in Thick and Thin Provisioned Environments VMware and Hypervisor Performance Comparisons in Thick and Thin Provisioned Environments Devanathan Nandhagopal, Nithin Mohan, Saimanojkumaar Ravichandran, Shilp Malpani Devanathan.Nandhagopal@Colorado.edu,

More information

ProRenaTa: Proactive and Reactive Tuning to Scale a Distributed Storage System

ProRenaTa: Proactive and Reactive Tuning to Scale a Distributed Storage System ProRenaTa: Proactive and Reactive Tuning to Scale a Distributed Storage System Ying Liu, Navaneeth Rameshan, Enric Monte, Vladimir Vlassov, and Leandro Navarro Ying Liu; Rameshan, N.; Monte, E.; Vlassov,

More information

Real-time Protection for Microsoft Hyper-V

Real-time Protection for Microsoft Hyper-V Real-time Protection for Microsoft Hyper-V Introduction Computer virtualization has come a long way in a very short time, triggered primarily by the rapid rate of customer adoption. Moving resources to

More information

An Optimized Time Series based Two phase strategy Pre-Copy Algorithm for Live Virtual Machine Migration

An Optimized Time Series based Two phase strategy Pre-Copy Algorithm for Live Virtual Machine Migration An Optimized Time Series based Two phase strategy Pre-Copy Algorithm for Live Virtual Machine Migration Ruchi Tailwal P.G. Scholar, Graphic Era Hill University Dehradun, India Avita Katal Assistant Professor,

More information

Deploying virtualisation in a production grid

Deploying virtualisation in a production grid Deploying virtualisation in a production grid Stephen Childs Trinity College Dublin & Grid-Ireland TERENA NRENs and Grids workshop 2 nd September 2008 www.eu-egee.org EGEE and glite are registered trademarks

More information

Preliminary Research on Distributed Cluster Monitoring of G/S Model

Preliminary Research on Distributed Cluster Monitoring of G/S Model Available online at www.sciencedirect.com Physics Procedia 25 (2012 ) 860 867 2012 International Conference on Solid State Devices and Materials Science Preliminary Research on Distributed Cluster Monitoring

More information

Dynamic task scheduling in cloud computing based on Naïve Bayesian classifier

Dynamic task scheduling in cloud computing based on Naïve Bayesian classifier Dynamic task scheduling in cloud computing based on Naïve Bayesian classifier Seyed Morteza Babamir Department of Computer Engineering University of Kashan Kashan, Iran e-mail: babamir@kashanu.ac.ir Fatemeh

More information

Microsoft Office SharePoint Server 2007

Microsoft Office SharePoint Server 2007 Microsoft Office SharePoint Server 2007 Enabled by EMC Celerra Unified Storage and Microsoft Hyper-V Reference Architecture Copyright 2010 EMC Corporation. All rights reserved. Published May, 2010 EMC

More information

Consolidating Complementary VMs with Spatial/Temporalawareness

Consolidating Complementary VMs with Spatial/Temporalawareness Consolidating Complementary VMs with Spatial/Temporalawareness in Cloud Datacenters Liuhua Chen and Haiying Shen Dept. of Electrical and Computer Engineering Clemson University, SC, USA 1 Outline Introduction

More information

Estimation of the Cost of VM Migration

Estimation of the Cost of VM Migration Estimation of the Cost of VM Migration Waltenegus Dargie Chair of Computer Networks Faculty of Computer Science Technical University of Dresden 01062 Dresden, Germany Email: waltenegus.dargie@tu-dresden.de

More information

QADR with Energy Consumption for DIA in Cloud

QADR with Energy Consumption for DIA in Cloud 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. 4, April 2014,

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