Server virtualization is having a significant impact on networked storage and is even driving some IT organizations to implement networked storage for the first time. Server virtualization solutions offer functionality that relies on, and requires, underlying networked storage. As such, the data storage infrastructure is a key building block in deploying server virtualization solutions.
Storing virtual machine images on networked storage enables the mobility of virtual machines between physical servers for load balancing, high availability and maximum utilization of resources. Multiple copies of virtual machines can also be created for replication and disaster recovery purposes, and can be used to perform bare-metal recoveries more easily.
There are five issues to consider when implementing server virtualization solutions:
- Capacity planning: In the physical world, the operating system and applications are traditionally stored on local disks in the physical server. But in virtualized environments, the virtual machine images, including the guest operating system, applications and associated data, are all stored on networked storage. Server consolidation efforts require careful capacity planning as more servers are virtualized within a company.
Another driver for implementing server virtualization is its ability to provide business continuance and disaster recovery solutions. This means that multiple copies (and snapshots) of virtual machines are created and stored at both the primary and secondary data centers. Virtual servers will have a significant impact on storage capacity, which will largely depend on the servers and applications targeted for the virtualized infrastructure. Start by collecting current capacity utilization data on your physical servers and be sure to include the operating system, applications and all the data.
- Storage tiers: Not all virtual machines and applications require the same level of performance and availability. This is especially true as the virtual machine and its associated data move from test/development to production. Configure two to three different tiers of storage based on price, performance and protection. Determine how critical the virtual machine is and place it on the appropriate tier. Ask yourself what the availability and performance requirements of the virtual machine and associated applications are. Be sure they're on a storage tier that matches your criteria. Keep in mind that the virtual machine can move between tiers of storage over its lifecycle.
- Zoning: The physical servers that are running the virtualization software and using networked storage for load balancing and failover must all be in the same zone. In large-scale deployments, creating one zone may not be practical or possible. In these cases, it makes sense to logically create multiple zones based on functionality. Check with your storage system vendor and follow its recommended best practices.
- LUN creation: When creating logical unit numbers (LUN) on the storage system, you need to determine if you want many small LUNs or fewer, larger LUNs. Having fewer large LUNs will be easier to manage and give the virtual server administrator greater flexibility. Creating multiple smaller LUNs can help avoid I/O contention and provide granular RAID characteristics that closely match application requirements. There is no simple way to decide this. It may require running the applications to see if disk performance is acceptable and tweaking the size of the LUN and RAID level to optimize performance. Remember, you can use cold migration to maintain virtual machines images when recreating the LUN.
- Networked storage administration: Traditional storage resource management tools will only have visibility to the Fibre Channel host bus adapter (HBA) on the host operating system (e.g. VMware ESX Server). They cannot directly access the virtual machine operating system that uses the storage. The virtual machines will be playing hide-and-seek from storage area network (SAN) administration tools. You'll have to rely on the server virtualization's management tools to manage the individual virtual machines.
N_Port ID Virtualization (NPIV), a technology that virtualizes world wide names (WWN) and assigns a unique WWN to each virtual machine, will allow storage management tools to work again. But NPIV is not yet supported across all server virtualization solutions. Nor are all HBAs compatible with NPIV. Note: ISCSI implementations will not have this limitation, since each virtual machine has a unique IP address. If you are currently using SAN administration tools, ask your vendor if it supports your server virtualization solution. While you're at it, ask them if it can maintain the relationship between the guest operating system and the storage in a virtualized environment.
The current adoption rate of server virtualization solutions across all industries, regardless of company size, is phenomenal. Networked storage is a fundamental piece of the virtualized environment and requires careful consideration during the implementation process. Server virtualization is a major catalyst for universal adoption of storage networks. There are hundreds of thousands of storage networks, but the opportunity is for millions. We have not reached this level of adoption in storage networking because the inherent complexity and cost currently outweighs the perceived value. Drivers, such as data growth, compliance, consolidation and data protection, have not provided reason enough for storage networking to be universally adopted. Server virtualization will play a big role in changing this dynamic, and the use of virtual machines will also drive further adoption of disaster recovery based on its ease of recovery and cost effectiveness.
About the author: Mark Bowker is an analyst with Enterprise Strategy Group in Milford, Mass. In this role, he supports ESG's Intelligent Information Management practice. Mark is also focused on server virtualization technologies and Content Aware Storage systems.
This was first published in September 2007