Академический Документы
Профессиональный Документы
Культура Документы
January 2010
White Paper: Discovering Improved Data Protection for the Virtual World: Focus on NetApp
Contents
Introduction.............................................................................................................................................3 Considerations .........................................................................................................................................5 NetApp Data Protection ...........................................................................................................................7
SnapManager for Virtual Infrastructure (SMVI) ....................................................................................................... 8 SnapManager for Hyper-V ........................................................................................................................................ 8 Application-Aware SnapManagers ........................................................................................................................... 8 SnapVault and Open Systems SnapVault (OSSV)...................................................................................................... 8 SnapMirror................................................................................................................................................................ 8 MetroCluster ............................................................................................................................................................ 9 FAS Data Deduplication ............................................................................................................................................ 9
All trademark names are property of their respective companies. Information contained in this publication has been obtained by sources The Enterprise Strategy Group (ESG) considers to be reliable but is not warranted by ESG. This publication may contain opinions of ESG, which are subject to change from time to time. This publication is copyrighted by The Enterprise Strategy Group, Inc. Any reproduction or redistribution of this publication, in whole or in part, whether in hard-copy format, electronically, or otherwise to persons not authorized to receive it, without the express consent of the Enterprise Strategy Group, Inc., is in violation of U.S. copyright law and will be subject to an action for civil damages and, if applicable, criminal prosecution. Should you have any questions, please contact ESG Client Relations at (508) 482-0188. This ESG White Paper was developed with the assistance and funding of NetApp.
White Paper: Discovering Improved Data Protection for the Virtual World: Focus on NetApp
Introduction
The primary driver for early adopters of server virtualization has been consolidation. A single physical server can host multiple virtual machines, all sharing the hosts physical resources. Hardware consolidation and improved resource utilization introduce cost savings and operational efficiencies. However, as organizations gain experience with the technology, new advantages emerge. Encapsulation and portability in server virtualization environments aid in availability. Encapsulating the workload into a single file containing the operating system, applications, and data enables mobility. Portability enables live, running virtual machines to be dynamically migrated from one physical server to another for load balancing, high availability, and maximum resource utilization. This aids in dealing with plannedand, importantly, unplanned downtime. One of the components needed to enable seamless workload migration is networked storage. The majority of server virtualization deployments utilize networked storage (and, in many cases, virtualization deployments are the catalysts for new networked storage implementations). ESG research found that most survey respondents implemented some form of networked storage to support their server virtualization infrastructure. Figure 1 shows the types of networked storage survey respondents cited as in use to support virtual server environments.1 Figure 1. Current Use of Storage Technologies to Support Virtual Environments What storage technologies are you currently using to support your organizations virtual server environment? Which would you consider to be the primary storage technology used to support your virtual server environment? (Percent of respondents, N=318)
iSCSI SAN Fibre Channel SAN DAS NAS Don't know Other
0%
73% 52%
All storage technologies supporting virtual server environment Primary storage technology supporting virtual server environment
38% 37%
80%
Oftentimes, migrating and deploying new applications on a virtualized platform creates an opportunity to rearchitect existing IT infrastructure. Just as in the aforementioned example where key virtualization features unlocked the implementation of networked storage, server virtualization requires the integration of multiple pieces of IT infrastructure (servers, network, and storage) in order to maintain its value throughout the entire infrastructure stack. Re-architecting doesnt end with infrastructure components. Processes will be impacted and, therefore, must be re-architected too. One example is data protectionthe components and processes used to make local and/or
1
Source: ESG Research Report, 2009 iSCSI Market Adoption Update, December 2009.
White Paper: Discovering Improved Data Protection for the Virtual World: Focus on NetApp
remote copies of primary data to be used if the primary copy is not available. In a virtual world, the encapsulation and portability of workloads allow organizations to create copies of virtual machines that can be replicated for local operational recovery and remote disaster recovery (DR)without reinstalling or reconfiguring the operating system or applications. In fact, in recent ESG research, survey respondents noted virtual machine mobility and easier, more cost-effective disaster recovery as top reasons for implementing networked storage in virtualized server environments (see Figure 2). Furthermore, nearly 40% of respondents not currently replicating virtual machines for remote DR planned to.2 Figure 2. Reasons Why End-users Will Increase the Use of Networked Storage With Server Virtualization Why do you expect that you will increase your usage of networked storage for storing virtual machines & associated data? (Percent of respondents, N = 181, multiple responses accepted)
Enables mobility of virtual machines between physical servers Easier and more cost-effective disaster recovery Provides increased uptime and availability Makes it easier to upgrade physical server infrastructure as needed Ability to store multiple copies of virtual machine images for high availability Ability to boot virtual machines from networked storage It is impractical to consolidate multiple servers and still use DAS
0% 10% 20% 30% 40%
Backup, recovery, and disaster recovery are all top considerations for server virtualization implementations. ESG polled IT professionals regarding their top IT priorities in 2009 with respect to server virtualization initiatives. As shown in Figure 3, replication of virtual machine workloads for disaster recovery and improvements in virtual machine backup and recovery ranked very high.3 As organizations mature in their use of server virtualization technology, they can begin to move beyond the early consolidation and capital cost savings benefits to achieve more automation of routine IT tasks, improved provisioning processes, greater operational efficiency, and better service levels for data protection and recovery processes. The automation and process improvements introduced via server virtualization are most often responsible for overall improvements in IT service levels.
2 3
Source: ESG Research Report, The Impact of Server Virtualization on Storage, December 2007. Source: ESG Research Report, Data Center Spending Priorities, January 2009.
White Paper: Discovering Improved Data Protection for the Virtual World: Focus on NetApp
Figure 3. Top Server Virtualization Initiatives Which of the following would you consider to be your organizations top server virtualization initiatives for 2009? (Percent of respondents, N=327)
Consolidate more physical servers onto virtualization platforms Expand number of applications running on virtual machines Make use of virtual machine replication for disaster recovery Improve backup and recovery of virtual machines Improve operational processes for managing virtual environments Move more applications from test/development to production environment Implement virtual machine mobility / HA (high availability) functionality Integrate virtual environments into existing management software frameworks Deploy a storage virtualization solution to support virtual server environment Purchase third-party management software for virtual environments
0% 10%
39% 38% 31% 24% 21% 21% 18% 17% 16% 12%
20% 30% 40% 50%
Considerations
Several things must be taken into account before embarking on the virtualization journeyespecially in the area of data protection. Best practices dictate that before virtualizingand, in most cases, re-architecting the environmentthe implications of the move from physical to virtual server infrastructure be considered. Will it increase or decrease the volume of data and/or files? Many organizations note a net growth in total storage volume after server virtualization is implemented. ESG found the difference in average capacity between those using and not using server virtualization to be pretty compelling (see Figure 4).4 An increase in the volume of data could result in more time required for backup processes, stressing often already-stressed backup windows. An increase in storage volume also impacts capacity requirements. Leveraging data deduplication technology in response to this side effect can optimize the environment. What repercussions exist when multiple virtual machines are sharing common physical resources? In a non-virtualized environment, physical servers benefit from dedicated and often underutilized resources. Conversely, virtual machines running on a physical server share available resources, improving resource utilization. Simultaneous resource-intensive processes occurring on one physical host could cause resource contentionpotentially triggering an automated migration storm (if automated migration is enabled) or, worst case, impacting the pool of application workloads sharing the common physical resources and causing performance issues. For example, simultaneous backup jobs kicking off on virtual machines sharing physical infrastructure is a common problem.
4
Source: ESG Research Report, 2009 iSCSI Market Adoption Update, December 2009.
White Paper: Discovering Improved Data Protection for the Virtual World: Focus on NetApp
Will my existing data protection methods still work, and are they optimized? To protect virtual machines and the applications that run within them, there are two strategies: protect the data files or protect the virtual disks. Traditional approaches are aimed at protecting the data files within the virtual machine and are often plagued with inefficiencies. If the data protection approach uses agent technology to facilitate file- or application-level data protection, what changes must be made to accommodate the virtual paradigm? Virtual disk methods are aimed at protecting the virtual machine files (.vmdk in VMware and .vhd in Microsoft virtual server environments) as a whole and cannot be targeted directly for backup without first quiescing the virtual machine. The latter approach may or may not capture an applicationconsistent image (the application is placed into a quiescent state which will commit in-memory transactions and then halt writes to the database and log files, guaranteeing that the application can be recovered quickly to the last committed transaction)critical for database and e-mail applications. Is a crashconsistent state acceptable or is an application-consistent image required? Are there features in the virtualization platform and/or networked storage environment that should be pursued in conjunction with or in place of existing strategies? The virtual server plan will take into account the underlying virtualization platform and any of its available basic or advanced features. Taking advantage of live migration, high availability, fault tolerance, snapshot, or backup features available in or with the hypervisor may mitigate the need to implement certain technologies in use in the physical world. One example is that a live migration feature may take the place of redundancy technologies to thwart physical hardware loss. If the underlying hardware suffers an outage, the live migration capability can move the VM to another host and operations seamlessly continue. Does the storage platform fully support all of the hypervisors advanced features? For example, a new feature of Windows Server 2008 r2 Hyper-V is Cluster Shared Volumes. In addition, advanced features of the storage platform such as snapshot, replication, and deduplication may be optimized for protecting workloads in virtualized environments. What protection requirements/recovery objectives apply? The recovery objectives for workloads will vary. Therefore, the protection requirements will vary, too. Virtual machines sharing common physical resources may leverage different data protection strategiesand should leverage the appropriate data protection strategy for the recovery requirements. Its not uncommon to see different data protection methods, frequencies of copies, and even layers of protection applied to virtual machines sharing physical resources. Will backup schedules be affected? Depending on the virtual machine density per physical host, it may not be a good idea to schedule backup jobs for multiple virtual machines at one time. In addition to the aforementioned resource contention, if you are thinking about using vStorage APIs, be sure to consider that the increased load on the server can cause post-backup deletion of VMware snapshots to time out and fail, suspending all subsequent backup jobs for the virtual machine until the snapshot is manually removed. Its also worth noting that vMotion migrations in VMware could cause unexpected backup schedule conflicts (and associated application performance degradation) as well. Will application-specific backup and recovery be required? Virtual-machine-level backups provide rapid backup and recovery (similar to bare metal recovery); however, item-level recovery for application components may not be available. In addition to the aforementioned issue of application-consistent vs. crash-consistent recovery, an application-specific approach should be implemented if granular recovery of application components (mailboxes in Microsoft Exchange, for example) is required. Can more than data be protected? Either the virtualization platform or third-party data protection solutions may offer features to preserve copies of more than just data. For example, VMware snapshots preserve the state of the virtual machine, allowing it to be returned to its point-in-time state (memory, disk, and configuration settings). It is recommended that an organizations data protection strategy be reviewed in the context of the new virtualization plan prior to going into production. To illustrate this concept, consider a new house or an extension on a new house being architected and built. In designing the new structure, the builder will include the proper
White Paper: Discovering Improved Data Protection for the Virtual World: Focus on NetApp
foundation to support it, building code requirements, and how it will be used by those residing in it. Skipping this step could result in a building that is not structurally sound, is not compliant, or doesnt fit the requirements of its users. Amending plans before the structure is built will be less costly, less disruptive, and less time consuming than trying to repair or reconfigure the structure after its been built. All facets of the data protection strategyincluding backup/recovery, availability, disaster recovery, compliance mandates, and moreshould not only be planned for, but optimized for the virtual environment prior to virtualizing. This will enable cost savings, future scalability, and the ability to meet or exceed service levels. Figure 4. The Impact of Server Virtualization on Overall Volume of Storage Capacity Total installed capacity associated with disk-based storage systems, by usage of server virtualization. (Percent of respondents)
Yes, we are currently using server virtualization (N=318) No, we are not currently using server virtualization (N=75)
25%
23% 19% 16% 15% 13% 10% 9% 9% 10% 7% 12% 11% 13% 12%
20%
15%
12%
10%
5%
4%
4% 1%
0%
250 TB to 499 TB
500 TB to 999 TB
1 PB or more
White Paper: Discovering Improved Data Protection for the Virtual World: Focus on NetApp
Application-Aware SnapManagers
NetApp offers SnapManager solutions for specific applications running in virtual machinesincluding Microsoft Exchange, SQL, and SharePoint; Oracle; and SAPthat automate backup, recovery, and verification of databases. Backups can occur while applications are online and recovery can occur at full or granular levels.
SnapMirror
SnapMirror, as its name implies, replicates the local snapshot copy from the primary to one or more secondary storage systems (1:1, 1:many, many:1, and cascading configurations are supported) over a LAN or WAN. Blocklevel updates and network compression reduce bandwidth as well as time requirements. Should production systems experience an interruption, the mirror copies can be mounted to recover from the failure. Synchronization can be synchronous, asynchronous, or semi-synchronous, allowing organizations to mirror data to meet specific DR needs.
White Paper: Discovering Improved Data Protection for the Virtual World: Focus on NetApp
MetroCluster
MetroCluster combines synchronous mirroring with array-based clustering to protect against data loss and enable continuous availability. It has the flexibility to work across a variety of clustering configurations: within a data center, across a data center, on a campus, or in a metropolitan area. Designed to address both planned and unplanned downtime, MetroCluster enables automatic failover of any single component failure and single command failover, which can be scripted for automation, of complete failure of storage node (controller and disk shelves) or the site itself. Hardware and software upgrades, including disk-shelf replacement, also can be performed non-disruptively to mitigate downtime of customers business. As a clustered solution, MetroCluster is independent of the OS and application so management is simplified and failoverwithout the need for host clusteringis transparent to the application. Mirroring at the aggregate level also simplifies administration whereby any changes (such as adding a volume) made on the primary array are automatically reflected on the secondary array. MetroCluster also allows users to read from both arrays in a round robin fashion, improving read performance especially for read-intensive environments. The combination of MetroCluster and server virtualization technologies provides an end-to-end server to storage availability solution for customers looking to virtualize their data center. MetroCluster can be combined with asynchronous SnapMirror to meet SLAs that demand continuous availability and long distance disaster recovery