Combining server and storage virtualization creates new opportunities to leverage these technologies that are not available when they are used separately.
Contents |
Executive Summary
Virtualization offers tremendous benefits by leveraging the advantages of running large physical servers in a central location with those of logical distributed computing in multiple locations across the network. Storage virtualization can bring the original promise of SANs to reality by creating large storage pools and then virtually allocating capacity as needed. Combining them into a fully virtualized data center multiplies their value.
Introduction
The rapid proliferation of server virtualization technologies, and especially VMware, has been driven by the huge operational advantages of consolidating many small servers on a few large boxes while maintaining the totally separate operating environments of the originals and allowing dynamic allocation of resources.
This solution both dramatically reduces support and maintenance and creates new opportunities to leverage the power of the server and storage infrastructure. Adding storage virtualization multiplies the flexibility and value of the total solution beyond the sum of its parts.
How Storage Virtualization Can Enhance VMware
Easy Provisioning of Volumes to Virtual Machines
In general, the value of virtualization grows with the number of servers being managed. Typically, therefore, virtualized environments support tens or even hundreds of logical machines on a very few, large physical boxes. The ability to virtualize the storage to these virtual servers quickly becomes very attractive. In an environment with hundreds of virtual machines, each requiring (say) about ten volumes, the total number of storage volumes required would be in the thousands. A SAN-based volume manager on the back-end of the virtualized environment allows users to create volumes for each of the virtual machines quickly and efficiently. It eliminates the need to deal with LUN management at the array level. Because each volume is provisioned at the virtualization layer, they can be mirrored, replicated, snapshotted, and even mounted into standard servers without the VMware operating system if needed.
Storage virtualization provides the flexibility to rapidly allocate capacity and allocate thousands of volumes on demand. Because adding and removing virtual machines and applications is easy, the environment becomes very dynamic. To fully realize that flexibility, however, storage needs to be provisioned and reallocated with equal speed and simplicity. Storage virtualization enables this flexibility.
Test Environments
Test Environments are one of the killer applications for virtual servers. A storage virtualization technology that supports low-capacity, point-in-time snapshots can increase this advantage by enabling the rapid creation of multiple snapshot copies of production volumes and their assignment to virtual test environments.
Additionally, snapshots can reduce data preparation time before each test. So, testers can be assigned real “live” data within seconds and then take snapshots of the data throughout the testing process. Should a multi-stage test fail at say, stage 13, the tester could go back to the snapshot taken at the beginning of the stage and run the test again, eliminating the need to repeat the 12 previous tests. Additionally, since the real failure may have occurred earlier in the testing process, the user could go back to previous snapshots taken at each stage and view the data to determine the root cause of the failure.
All of these features significantly reduce the time needed for testing and increase the productivity of the testing team. Bringing a product to market quicker or isolating a software bug quicker can improve the profitability of a company.
Enhanced Backup
Having a virtual environment with hundreds of virtual machines can create a complicated, expensive backup proposition. Snapshot functionality obviates the need to install backup agents on every virtual machine. The backup can be done by creating snapshot copies for every virtual server and then assigning the copies to a virtual machine with the dedicated role of backup server. In this manner, the backup server is the only virtual machine that needs the backup software. When dealing with hundreds of virtual servers, this can reduce the cost of backup licenses considerably.
With capacity growing at exponential rates and processing hours becoming more and more important, backup windows are becoming non-existent. Simply stated, there is too much data to backup during off-hours. By comparison, snapshots can be taken at any time without taking the application offline. This creates a zero-window for backup. For many users, this solves the “shrinking backup” window problem.
Snapshots can add another significant benefit to the overall backup strategy. It is possible to keep point-in-time snapshots online for extended periods of time. If the data needs to be restored, the restoration can be done in seconds from a point-in-time snapshot rather than in hours off tape, reducing the recovery time objective (RTO)from hours to seconds.
One issue in using snapshots is ensuring point-in-time consistency between volumes. This often requires that the application be suspended for a time while(for example) database buffers are flushed and all I/O completed. More complex solutions such as the use of consistency groups should be evaluated. Snapshots will not usually eliminate the requirement to take a full backup of the systems on a regular basis.
Consolidation of Servers at the Disaster Recovery Site
Today, remote mirroring and disaster recovery is a requirement even for very small companies due to regulation, corporate policies, or simply common sense. Large enterprises typically have the resources to spend on necessary communication lines, equipment, software, and training for disaster recovery. However, small and medium-sized businesses do not always have these resources, leaving the company exposed to regional disasters. This is another area where the combination of storage and server virtualization software can enable an affordable solution for disaster recovery.
Over the life of a remote mirror implementation, the most expensive component is the communication lines between the source and target locations. A solution that can remotely mirror using a snapshot-based technique, where only the differences between the snapshots are transmitted, avoids the need for very expensive communication lines between the locations. It is possible to use T1 or T3 lines that often cost hundreds of dollars per month, rather than higher bandwidth lines that can cost tens of thousands of dollars per month.
A remote mirror is an insurance policy to make systems available should a regional disaster render your primary site unavailable. However, there is no need to dedicate resources until a failure occurs. A virtual server gives users the flexibility to assign just very small amount of resources required to accomplish the mirror to the remote site. With snapshot-based remote replication, those resources are minimal because of the reduction of the amount of data being transmitted to the remote site. Should a failure occur at the primary site, server virtualization could then be used to assign more virtual machines to support the production workload. Virtualization of the remote resources allows them to be used for other purposes such as testing while they are in "standby" waiting for a failure, while guaranteeing that they will be instantly available when that primary site failure happens. This is true even if the primary site uses “physical” servers.
Thus the powerful combination of storage and server virtualization, therefore, allows users to build a disaster recovery site at a fraction of the cost. While large enterprises use expensive communication lines (e.g., OC-3, OC-12), high-end arrays in both the local and remote sites, and dedicated servers in the remote site, virtualization allows the use of inexpensive communication lines, inexpensive storage arrays at the remote sites and the use of virtual servers instead of physical servers at the remote location.
Topology
Essentially, server and storage virtualization work in very similar ways. They pool the physical resources and assign resources to applications as needed. The benefits of virtualizing any resources are efficiency, utilization and ease of management. By combining the two solutions, a fully virtualized environment can be created, giving the user the flexibility to assign or re-assign resources as necessary.
Storage Virtualization Solutions with VMware
Products such as LSI StoreAge SVM™ when combined with VMware provide a good set of resource virtualization function, which will allow organizations to realize many of these new benefits.
Action Item: Virtualization technology is becoming increasingly prominent in IT environments. Both server and storage virtualization offer unique benefits. However, it is not until the two technologies are combined that users can realize the full benefits of the solution. The ability to scale both server and storage resources as needed is a tremendous benefit. Additionally, the combination of the two technologies creates new opportunities for value. To the extent possible, IT executives should ensure that server and storage virtualization projects are strongly linked and architected at the same time.
Footnotes: