EonStor GS Family Best Practices Guide White Paper Version: 1.1 Updated: Apr., 2018 Abstract: This guide provides recommendations of best practices for installation and configuration to meet customer performance and availability requirements.
Contents Contents... 2 Summary... 4 Audience... 4 Terminology... 4 EonStor GS Configuration Architecture... 5 Propose Solutions... 6 Requirements... 6 Product family for application type... 6 Capacity... 6 Performance... 6 Drive Type... 7 Front-end connectivity... 7 RAID Protection Level... 7 LUN & file system number... 8 Rack Unit... 8 Storage Configuration Principle... 9 Select Product Series... 9 Logical Drive (LD)... 9 Calculating Usable Capacity... 9 Calculating Performance... 10 Form Factors of the Storage Systems and Expansion Enclosures... 10 Controller Cache Memory... 10 First Installation... 11 Basic Settings... 11 (1) Access to EonOne User Interface... 11 (2) Update Firmware... 11 (3) Time Settings... 11 (4) Licensing... 11 (5) Channel and Network Service Settings... 12 (6) Export System Configuration Information... 12 Configure a Pool... 12 (1) Create a Pool... 12 (2) Add Logical Drives (LDs) into a Pool... 13 (3) Add a Spare Drive... 13 Volume Consideration... 13 Volume Type... 13 Mapping & Cabling for Block-level Volumes... 14 High Availability Configuration for Block-level Volumes... 15 Cabling & Trunking for File-level Volumes... 15 Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 2 / 20
High Availability Configuration for File Sharing... 16 Replication... 18 Disaster Recovery Solution... 18 Example... 19 Disaster Recovery Solution for Database HA Architecture... 19 Conclusion... 20 References... 20 Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 3 / 20
Summary The purpose of this practice is to focus on how to propose and configure the maximum performance and high availability architecture with Infortrend storage system. Especially for redundant controller model, the performance and high availability plays an important role when we deploy the solution. The guideline covers the most of general cases when using redundant model. We strongly recommend following these principle to deploy Infortrend product. However, some exception cases are not addressed and covered in this guide. Therefore, this guide is not absolute correct for every situation. In a mixed business environment, some practices of the third party application should be taken into consideration that will be mentioned in the other specific application guide. Audience This guide is intended for the Infotrend partners, customers and employees who are proposing and configuring the Infortrend system. We assume the audience is familiar with Infortrend GS storage operation. Terminology Storage controller A storage node provides the computing resource to service the I/O between storage and hosts. There are processor, memory, cache backup module and connectivity on it. Storage system An enclosure includes controllers, fans and power supplies. Expansion enclosure An enclosure with drive slots that is used to expand the additional drives for storage system. Asymmetric Logic Unit Access (ALUA) The paths from both controllers to any given LUN are available but only one of the controller which the LUN assigned to can process the active/optimized path. RAID (Redundant Array of Independent Disks) Storage virtualization technology that integrate multiple disks together to provide better performance and data redundancy. Logical Drive The basic unit of RAID group created by controller. Pool the integration of logical drives where volumes can create. Block-level volume A block-level storage device that can be accessed via block-level protocols. File-level volume A storage device with file system that can be accessed via file system shared protocols. LUN (Logical Unit Number) - A block-level storage device that showed on host side. File system A mechanism to control how data is stored and retrieved via file system protocol such as CIFS, NFP & FTP. Shared folder A location for shared storage access via file system protocol. Cloud gateway A volume-based cloud feature that can backup volumes to or recovery volumes from the public cloud. SyncCloud A folder-based cloud feature that synchronize multiple folders of Infortrend storages with a cloud folder. Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 4 / 20
EonStor GS Configuration Architecture EonStor GS series is an enterprise-class storage system which provides an efficient and cost-effective way to allocate storage resources to meet the capacity, availability, and performance requirements of different applications. Basically, users can create a pool on the basis of multiple logical drives (LDs) with different RAID protection levels and a certain number of drives. Moreover, these pools can consist of one or more types of disks, providing the flexibility for deployment. A volume comprising of multiple disk sectors can be created on a pool, and the users can choose whether to enable the file system. The Infortrend s proprietary firmware in the EonStor GS storage systems provides an alternative configuration between the block and file level volume according to the application requirements. Therefore, users can create a storage pool and use it for block-level and file-level purposes. Furthermore, since the EonStor GS series is a unified storage system, users can configure different types of volumes in the same pool at the same time. For the channel connections, Ethernet ports (onboard and expansion I/O host boards) on the EonStor GS series provide services for the file (CIFS, NFS and AFP) and the block (iscsi) traffic. Thus, users can enjoy system resiliency by switching the I/O type between block-level and file-level via the web-based management suite- EonOne. EonStor GS configuration architecture Besides consolidating SAN and NAS storage configurations, EonStor GS series is also complemented with two proprietary cloud backup solutions Cloud gateway and SyncCloud. Cloud gateway is a pool-based but volume enabled backup/restore solution. To implement the solution, a pool is connected to a bucket of cloud provider via RESTful API standard. user can decide whether to enable the cloud gateway backup function for a volume. With abundant features, there are three types of backup modes that can be selected - Cloud Tier mode, Cloud Cache mode and Cloud Fully Cache mode. cloud gateway backup features are also available for block volumes and file volumes. On the other hand, SyncCloud is a folder-based cloud backup feature that allows folders in an EonStor GS storage system to be synchronized with folders/buckets on cloud storage provided by public cloud services, such as Amazon S3, Aliyun, Microsoft Azure and OpenStack. Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 5 / 20
Propose Solutions Infortrend products offer a fairly complete product line and flexible modular design that provides accurate and efficient solutions for customer needs. This section we will list some customer requirements information and illustrate solution design steps to meet those requirements. Requirements Product family for application type EonStor GS 3025A series is an all-flash array and unified storage which also integrates file-level, block-level and cloud-integrated features. It is empowered by lightning performance and extreme low latency (up to 700K IOPS within 0.5 ms latency) to meet any high density workloads. Moreover, it is installed with optimized firmware designs to extend SSD lifespan with latency-sensitive applications. EonStor GS Family is a hybrid storage that consolidates file-level, block-level and cloud-integrated features. This product is your best choice if you need file sharing requirement in your application, especially with cloud services. EonStor DS Family is hybrid storage with unprecedented performance in pure block-level environment applications for structured data. For GS Family, there are also Turbo versions with better performance for file-level (NAS) applications. GS Family: http://www.infortrend.com/tw/products/gs DS Family: http://www.infortrend.com/tw/products/ds Capacity Capacity is one of the most important factors when choosing a RAID protection level. Generally, customers will provide the capacity requirements for raw capacity or usable capacity in advance. The main difference is that the relevant factors of usable capacity are the RAID protection and reserved space overhead, which is a critical issue for capacity-sensitive applications. Performance Currently, Infortrend provides the information of the following two performance tests. You can find them in the product introduction on our official website. IOPS workloads by simulating small-block random access. Bandwidth workloads by simulating large-block sequential access. These standard simulation testing and performance results are used by the industry to evaluate the maximum achievable performance with a specific configuration. However, the performance may have fluctuated values depending on the workload type and environment. In order to provide accurate recommendations what the customers need, we prefer to be informed (the type of servers and applications, read/write behavior ratio, data block size, etc.) as more as possible. Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 6 / 20
Drive Type Inevitably, there is always a trade-off between performance, capacity and cost for different drive types. Hence, the drive type will be a key element that affects the expected workload. SSD drives are for IOPS-demand and latency-sensitive applications. Especially, the EonStor GS 3025A all-flash array is boosted by leveraging SSD drives, leading outstanding performance results. Infortrend products support both SAS and SATA interface SSDs. However, the price of the SSD is usually the highest of all disk types. 10K SAS drives are for applications with medium bandwidth workloads. NL-SAS drives are suitable for large-capacity applications. SATA is commonly used for SMB capacity-oriented for cost effective purposes. Infortrend also provides MUX board solution so that users can configure SATA disks in the redundant model. Front-end connectivity Infortrend provides a modular design host board that simplifies system maintenance and upgrades. The converged host board offers users the resiliency to switch between 8Gb/s FC, 16Gb/s FC, 10Gb/s iscsi and FCoE connectivity. Please note that the front-end connectivity bandwidth may affect your system performance output. For example, 4x 16Gb/s FC channels can provide up to (4 * 16Gb/s) / 8 = 8GB/s performance throughput. Please make sure the bandwidth of front-end channels can meet the demands of your application workload. The EonStor GS supports multiple protocols (NFS, FTP, and / or SMB) for file-level NAS applications. In order to achieve better performance, it is recommended to enable the Jumbo Frame function. RAID Protection Level Infortrend provides multiple RAID protection levels. Basically, the selection of the RAID protection levels is based on performance, capacity and data critical levels. RAID 1 provides the best availability and the best IOPS performance for random write access behavior with small block size workloads. However, its capacity utilization is the lowest since half of them will be used for data protection. RAID 1 is useful when read IOPS or reliability is considered more important than write performance or storage capacity. RAID 5 has the highest usable capacity, but the lowest availability. Its performance is between RAID1 and RAID6 for general-purpose workload. RAID 6 has more parity drives than RAID 5. It provides medium-high protection and storage capacity, but with the lowest overall performance. Infortrend recommends that you use RAID 6 with hot spares for general-purpose workloads. RAID 6 has the higher availability than RAID 5 and better capacity usage than RAID 1. In general, since Infortrend storage systems can provide outstanding performance, they can still meet the performance requirements with RAID 6 protection level. Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 7 / 20
LUN & file system number For block-level, Infortrend s proprietary controller assignment mechanism supports Asymmetric Logical Unit Access (ALUA) and Symmetric Active-Active for redundant models. However, the selection of these mechanisms is a trade-off between manageability and performance. If more than two pools are deployed in the configuration, we recommend that you use ALUA to reach the optimal performance. For the controller assignments, each pool is assigned to a controller. The Symmetric Active-Active mode allows a single pool to be assigned to two storage controllers simultaneously while providing high availability architecture. For a single LUN, unlike ALUA mode, Symmetric Active-Active can use both two controllers instead of only one controller. The performance of single LUN can be increased to 180% of ALUA. For more detail about Symmetric Active-Active, please refer to the application note. Generally, the host operating system supports disk striping function to combine multiple volumes into a single LUN. For example, the Striped Volume in Windows server 2012 R2. The performance impact of the ALUA mode will be less than the Symmetric Active-Active mode. For file-level volumes, they do not support Symmetric Active-Active mode pools. In order to optimize the overall system performance, it is recommended to create at least 2 pools with each pool assigned to a controller. Moreover, assign each controller to the host server so that the front-end workload of both controllers can be load balanced. Rack Unit Commonly, IT administrators concern about the accommodation space for the storage system since the rack space in their environment is usually limited. Therefore, the rack unit has become one of the limitations. Before we deploy the storage systems, the ratio of performance and capacity density should be put into consideration. For example, the performance/density of a SAS HDD is more than a NL-SAS disk. To achieve the same performance, customers may need more rack unit space when using NL-SAS disks. Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 8 / 20
Storage Configuration Principle In this section, we will provide detailed configuration guidelines based on customer s requirements of the previous sections. Infortrend and its partners can propose appropriate solutions by following the principles. Select Product Series Infortrend has complete product lines for business scales in each family, which can be distinguished by performance, the supporting drive number and memory cache capabilities. According to the RAID level and I/O workload type of customer s application, we can choose the appropriate product series based on our performance report. For example, if the user needs the performance requirements: 4K random read 200K IOPS for block-level, it is suggested to select GS 2000 series since GS 1000 series may not meet the requirements. Logical Drive (LD) A storage pool can contain multiple logical drives (LDs) that can achieve stable performance and risk spreading. To optimize the system deployment, we recommend that an LD in the 2U12 and 3U16 models should be configured with the same number of drives as the enclosure bay number, including up to 16 disks with a hot spare drive. For models with larger bay numbers, we recommend configuring the LD as a 12-drive unit, including a hot spare drive. Form factor Recommendation 2U12bay 11 disks per LD + 1 hot spare 3U16bay 15 disks per LD + 1 hot spare 2U24bay / 4U24bay 11 disks per LD + 1 hot spare 2U25bay 12 disks per LD + 1 enclosure spare 2U12bay + 4U60bay expansion enclosure 11 disks per LD + 1 hot spare 4U60bay + 4U60bay expansion enclosure 14 disks per LD + 1 hot spare For redundant models, it is recommended that the number of LDs be even. Therefore, the LDs can be evenly distributed in even pools and load balancing is achieved by assigning them to two controllers. We recommend that the configurations of the LDs (e.g. the RAID level, disk number per LD and stripe size) in the same pool are identical. Calculating Usable Capacity As we mentioned in the previous section, the usable capacity is considered with RAID protection and reserved system overhead. Usable capacity = Capacity per disk x (Disk number RAID parity Hot spare) x LD number x 0.9 (overhead) For example, we used 32x 6TB NL-SAS drives with 2 RAID-6 LDs and 2 hot spare drives. The usable capacity will be: 6TB x (16 2 1) x 2 x 0.9 =140 TB. The Reserved Space" for data service is the space in a pool where it conserves data such as snapshot images, file system configurations and metadata. It is recommended to reserve at least 30% of the pool's space for these data services for future needs. By default, the system will send notifications when the reserved space is less than 30% of capacity. Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 9 / 20
Calculating Performance The minimum number of drives to meet performance requirements can be estimated via Infortrend performance report. Infortrend performance report includes different I/O workload simulation tests with multiple RAID levels. To meet your requirement of performance and capacity, please contact Infortrend or its partners for the best configuration. Form Factors of the Storage Systems and Expansion Enclosures Infortrend provides multiple form factors of storage devices and expansion enclosures. Depending on the number of drives and rack unit limits, there are a variety of combinations of products to choose from. For limited rack unit purposes, Infortrend offers 2U25 and 4U60 high-density expansion enclosures models to achieve immense capacity requirements. For example, if the requirement includes 64 NL-SAS drives with 6U rack unit limits. Infortrend can easily meet the requirements with a 2U12 bay RAID device and a 4U60 high-density expansion enclosure. Controller Cache Memory To achieve the maximum performance, we suggest installing at least 16GB memory cache per controller (32GB total). As more data services are enabled, more cache memory may be required. Consequently, we recommend that the cache memory is to be mounted symmetrically to the DIMM slots since the controller is dual-channel designed. For example, there are 4 DIMM slots on an EonStor GS 3000 controller. The performance of dual 16GB DIMMs memory is better than that of a single 32GB DIMM. However, the performance of dual 16GB DIMMs and four 8GB DIMMs are the same. Please remember to install the blue DIMM slots first. Note that if you install different sizes of memory on a dual DIMM slot, the controller may not boot up properly. The default write cache policy of Infortrend s storage systems is mirrored write-back cache policy. For each write request, the data is written to the write cache on the primary controller and then synchronized to the write cache on the secondary controller. After the synchronization is complete, the acknowledgement signal will be sent back directly to the host server without waiting after the data is written to the storage device. Due to the write cache data is always mirrored and protected between redundant controllers, the policy has enabled the storage system to provide low latency and high throughput for write-intensive applications. However, users can still switch the write cache policy to write-through cache manually. Moreover, the system will also automatically switch from write-back to write-through cache based on the trigger settings. The mechanism of write-back cache The mechanism of write-through cache Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 10 / 20
First Installation Here we list the basic settings that required for the first installation in this section. For more information on advanced features and parameters, please refer to the Software user manual. Basic Settings (1) Access to EonOne User Interface Default IP address The default IP address of the EonStor GS is 10.10.1.1. Please connect the storage system to the server or PC via EonOne software suite and set them into the same network domain. If there is a firewall in the environment, you need to open the corresponding port to access the management port (please refer to the Software user manual). Assign IP address via console If you cannot access the EonOne interface through the default IP address, go to the Console interface (Baud Rate: 38400) via RS-232 cable. The default login password is set to admin, and you can operate via the Enter / Esc and direction keys. Go to View and edit Configuration parameters Communication Parameters Internet Protocol (TCP/IP) lan0 Manually set the IP address of the management port. (2) Update Firmware Since our firmware version will continue to be enhanced, it is recommended that you update the firmware to the latest version at the first startup. We strongly recommend that you turn the storage system into shutdown mode before the first firmware upgrade since the latest firmware may contain huge enhancements. After upgrading the firmware for the first time, please remember to press the reset controller button to ensure that the upgrade process is complete. (3) Time Settings It s recommended to complete time zone and time settings when the system initialization so that we can obtain the exact event log of the system. The time synchronization is required especially for the features such as cloud functions, joint AD server, and other scheduling triggered features. (4) Licensing If you purchased an advanced license, please activate the license at the first installation. A Standard License is free for all users and is preloaded on your devices. Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 11 / 20
(5) Channel and Network Service Settings The default channel mode of the storage system is set to block-level. Change the channel mode to your desired channel type according to your applications. If the channel type is changed, the system must be restarted to activate the settings. For file-level protocols, you should enable the network service of the protocol on EonOne before sharing a folder (some protocols are enabled by default). (6) Export System Configuration Information Since Infortrend storage systems do not need extra drive spaces as system disk, we recommend you to export the system configuration information to preserve the current system status after all settings are completed. You may import this configuration file after you replace the hardware component or restore the system to the default process. Configure a Pool (1) Create a Pool Create the first pool based on your deployment plan. You can specify the LDs you want to create. The default stripe size is typically set to 128K for general purposes, and you can modify it for specific I/O workload. For example, it is more ideal to set the SPC-2 benchmark tool to a 512K stripe size. It is recommended that you set the stripe size to 256K for Microsoft Exchange and Autodesk solutions. Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 12 / 20
(2) Add Logical Drives (LDs) into a Pool Add more than one LD into the pool. It s recommended that configurations of all LDs (e.g. drive number, RAID level and the stripe size) within a pool are the same. If you are planning to expand your storage in the future, you must consider this principle. For example, a RAID 6 LD consists of 10 drives with 128K stripe size. You should expand the capacity with the same type of LDs to get a better performance. To reach cost-effective solution, automated storage tiering feature may be used. A hybrid pool can include SSD and HDD (SAS or NL-SAS drives) LDs within a pool. Users can get the benefits of performance from SSD LDs and capacity from HDD LDs. For more detail about automated storage tiering feature, kindly refer to the application note. (3) Add a Spare Drive There are three types of spare drive. A local spare drive is dedicated to a logical drive and can be used to replace any member drives, even across subsystem enclosures. However, it cannot be used to replace a drive in other LDs even if the LD is in the same enclosure. A global spare drive is not specified to a specific logical drive. Even more, it can be used to replace any disk in the system. We recommend that you use a local or enclosure spare for each logical drive or enclosure as well as some extra global spare drives to achieve multi-level protections. Since it's better to rebuild a logical drive in the same enclosure, the global spare is only suitable for second protection. It s also recommended that you use the same drive type as LD as a spare drive. For example, a SSD spare drive is assigned to a LD consisting of SSDs. Volume Consideration Volume Type The Infortrend EonStor GS family supports multiple types of volumes and features to meet various I/O workloads demands. Basically, the volume can be divided into the block-level and the file-level. These two types of volumes can be further enabled the features below: Thin provisioning technology allows users to allocate a large amount of virtual capacity for an application server, regardless of the actual physical capacity available, which provides flexibility for space efficiency. However, the full-provisioning volume can provide higher performance than the thin provisioning volume. It is a trade-off depending on the application you are going to use. Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 13 / 20
Cloud gateway feature can be enabled if the pool has been connected to a cloud provider. Please note that if the cloud feature is enabled, thin provisioning will be automatically enabled at the same time. For more information about the cloud gateway feature, please refer to Cloud-integrated application note. A flow diagram of multiple decisions for volume creation Mapping & Cabling for Block-level Volumes We strongly recommend that you connect the front-end connection symmetrically to both controllers to ensure high availability and prevent single point of failure in the event of a controller or connection failure. For some data service features, the symmetric connection of front-end on both controllers is required. For example, a volume which is assigned to the primary controller (controller A) is LUN-mapped to the channel 4 on both controllers, and these two ports are connected to a host with MPIO enabled. MPIO mode: If the MPIO is in the Round Robin mode, the IO workload will be load-balanced via these two paths. Otherwise, the IO workload will only operate through the channel of the controller in which the volume was assigned to. Controller failover: The pool assignment of the controller A will switch to the controller B when controller A fails. Thus, the IO workload can still be available via the channels on the controller B. The failover time for ALUA or Symmetric A-A volumes is approximately 30-40 seconds. (The application failover might be longer because of the network, host server OS, and also application attributes.) Channel failover: The controller can monitor whether its channel ports are disconnected. If the channel port of the primary path fails, the IO workload will failover to the passive path via another controller. Block-level controller failover procedure Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 14 / 20
High Availability Configuration for Block-level Volumes To achieve no single point of failure, the HA configurations are required. In the following example we leverage the advantage of the EonStor GS redundant model. We recommend that you map the volume to at least two channels of a controller with MPIO enabled. For redundant models, there are at least 4 available paths in this case. We cross the cabling by configuring with dual switch between the storage and host servers to avoid a single path or network failure. In the event of a switch or cable failure, the host servers can still access via the remaining paths. Availability Solution Controller failover / Cable failover for a Symmetrically map the volume to the channel on both controllers (A & B) single channel on both controllers with MPIO enabled on the host server. Symmetrically connect the cable on both controller ports on the different switch (switch 1 & switch 2). Cable availability within a controller Map the volume to more than one channel of a controller with MPIO enabled and connect the cable on the controller ports to different switch. Power availability Plug redundant PSU to separate PDUs Block-level HA configuration example Cabling & Trunking for File-level Volumes For file-level volumes, we strongly recommend you to connect the front-end connection symmetrically to both two controllers to implement the high availability as well as load-balancing when a controller or channel fails. For some data service features, the symmetric connection of front-end on both controllers is required. Generally, the shared folder can only be accessed via the channels on the controller where its pool resides. For example, although channel 4 & channel 5 of the both controllers on EonStor GS are connected to the same switch. The shared folder of the pool can only be accessed via the channels of the controller which its pool is assigned to. Controller failover: The pool assignment of the shared folders will switch from controller A to controller B. Moreover, the channels on the controller A will also physically switch to controller B and thus the original IP address of the controller A can still be accessed to the shared folder. Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 15 / 20
Cabling failover: To avoid single point of failure, we recommend that you deploy more than one connection through a single controller and configure trunk groups across multiple ports. For the switch connection, connect the cable to redundant switch for high availability. High Availability Configuration for File Sharing Same as the previous section, the EonStor GS redundant model is used in the following example. We recommend that you connect the cables to the channels with same port number on both controllers symmetrically to the switch. In case of the controller failure, the controller will failover to the surviving controller and continues to operate I/O since it s connected to the same switch. Furthermore, we connect channels of the different controller to the different switches with IP trunking (link aggregation), so that the storage can still be accessed by the surviving cables with the same IP address if one of the channels fails. To implement the high availability, we use dual switch between the storage and host servers to avoid single path or network failure. In case of one switch or cable failure, the host servers can still access via the remaining path. Availability Solution Controller failover Symmetrically connect the cables to the channel on both controller (A & B). In case when controller failover, the IP of the channel in the controller will also failover to the channel on controller B. Cable availability on a controller Connect at least two channels of a controller to different switch with enabling trunk group. Thus, the storage system can still be accessed via the same IP address when one of the cables is disconnected. Power availability Plug redundant PSU to separate PDUs Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 16 / 20
File-level HA configuration example Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 17 / 20
Replication Disaster Recovery Solution For some critical application like database, replication to a remote site can be configured as disaster recovery solution. To deploy the disaster recovery configuration, we recommend using synchronous volume mirror or asynchronous volume mirror with a schedule. In disaster recovery solutions, there are trade-offs between response time and Recovery Point Objective (RPO). Synchronous volume mirror requires higher network bandwidth and is suitable for critical data. The host will write data to both the source and target at the same time. Synchronous volume mirror has the best RPO but much more response time. Asynchronous volume mirror has less impact on I/O latency compared to synchronous volume mirror. The data in the source volume will be replicated to the target volume at replication pair creation and a differential snapshot is sent during each synchronization. Moreover, replication supports target snapshot function to transmit the latest snapshot to the remote site for the second protection. For more detail about automated storage tiering feature, kindly refer to the application note. Disaster recovery solution with Infortrend remote replication Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 18 / 20
Example Disaster Recovery Solution for Database HA Architecture Database is always a critical and latency-intensive application. Asynchronous volume mirror is suitable for database to deploy disaster recovery solution. Both sites can contain high availability architecture with cluster host servers and a redundant storage system. Here is an example. The IT administrator needs a cost-effective and high performance storage solution for database application. In addition, to minimize the risk of downtime, the administrator also requires that the storage is capable of providing failover function for the database cluster and need a replicated site in case disaster happens. The solution and configuration Infortrend provided will be as below. Disaster recovery solution for database HA architecture For example, the customer needs 40TB with automated storage tiering for SQL server and 100TB for file server. Moreover, to accelerate the read performance, the customer requests a 1TB SSD cache pool. The solution can be proposed with 4U60bay form factor storage as below. Estimated Purpose Requirement Pool Assignment Logical Drive (LD) Drive Number Capacity SQL Server 40TB Pool 1 Controller A Tier0: SSD RAID1 8 x 400GB SSD Tier1: NL-SAS RAID5 14 x 4TB NL-SAS File Sever 100TB Pool 2 Controller B LD1: NL-SAS RAID5 15 x 4TB NL-SAS LD2: NL-SAS RAID5 15 x 4TB NL-SAS SSD Cache SSD Cache 1TB Pool Controller A&B N/A 4 x 400GB SSD 2 x 400GB SSD Global Spare N/A N/A N/A N/A 2 x 4TB NL-SAS An example of proposing solution to meet the customer s requirement 48TB 100TB 1.6TB SSD Cache N/A Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 19 / 20
Conclusion This practice provides explicit configurations and recommendations for the Infortrend storage systems in consideration of customer s requirements. By deploying the high availability configuration, users can enjoy data service features on EonStor GS without single point of failure in-mind. To ensure data integrity, Infortrend EonStor GS family provides abundant disaster recovery solutions such as local/remote volume-based replication and cloud gateway. For NAS file sharing, file replication and SyncCloud features not only simplify the deployments but provide the reliability. For more information of data services on EonStor GS, please refer to the reference list in the following section. References Category Introduction Snapshot Block replication File replication Cloud gateway SyncCloud SSD Cache Automated Strorage Tiering Symmetric Active-Active Technical document Introduction to EonStor GS/GSe EonStor GS/GSe Snapshot Feature Guide EonStor GS/GSe Block Replication Feature Guide EonStor GS/GSe File Replication Feature Guide EonStor GS/GSe Cloud-Integrated Features EonStor GS/GSe SyncCloud Feature Guide SSD Cache Automated Storage Tiering Advantages of EonStor GS Family Symmetric Active-active Controller Copyright 2017 Infortrend Technology, Inc. All rights reserved. Page 20 / 20