ATLANTA -- As users expand Hyper-V deployments, they’re encountering new wrinkles in storage integration and disaster recovery, both of which were hot topics among Microsoft
CSVs top of mind
Janssen Jones, associate director of AIT infrastructure at Indiana University, put Cluster Shared Volumes (CSVs) through their paces and presented the lessons he learned in a session on optimizing Hyper-V environments.
CSVs, which first became available in Windows Server 2008 R2, allow multiple physical or virtual machines (VMs) to access the same logical unit number (LUN), and support for them in Hyper-V means each VM no longer needs its own LUN.
According to Jones, attaching too few virtual servers per CSV led to intermittent troubles with network connectivity. When network administrators were performing maintenance, the CSVs weren’t always able to connect to Active Directory, which acts as a kind of permissions broker through which VMs access CSVs. This Active Directory issue can also occur with more VMs per CSV, but the effect is less widespread with some consolidation of workloads on each volume, Jones said.
However, Jones cautioned, too many VMs per CSV can create backup contention, especially with snapshot-based backup processes such as Microsoft’s Data Protection Manager. In that case, too many machines snapping system image and data files on the same LUN at the same time can weaken backup performance.
Jones said the happy medium for the university’s 200 Hyper-V VMs on 16 physical servers is to provision one CSV per host, which in his environment works out to about 15 VMs per CSV. However, as host servers grow larger and accommodate more VMs, each host may end up with two CSVs, he said.
Also, Jones said he would like to see the ability to do storage live migration of virtual hard disks (VHDs) between disks. “I can replicate CSVs from one SAN to another, but I can’t move a VHD between two separate CSVs. You have to bring a VM offline to do that.”
Getting SANs, DR up to speed
Meanwhile, Hyper-V still has some work to do to get its SAN integration on par with VMware vSphere’s. Scott Ladewig, a network and operations manager of information services at the Olin Business School at Washington University in St. Louis, said his Xiotech SAN works well for both ESX and Hyper-V workloads, but some of the finer points of integration aren’t there yet with Hyper-V. For example, Ladewig said he’d like to use Xiotech’s Volume Shadow Copy Service snapshot writer with Windows Server Core, which isn’t yet supported.
“They’ll build it in in the next version,” he said.
And users said they’d like Microsoft to offer multi-site high availability and disaster recovery without the need for third-party products, particularly SAN-based replication.
Joseph Henrich, assistant vice president of technology for a community bank in the Midwest, said he deployed Hyper-V primarily for better availability and recoverability of servers without needing like hardware at primary and secondary sites. Today it is possible using third-party products to create stretched clusters using Hyper-V, but Henrich said he’d like to be able to keep multiple sites synced without requiring expensive hardware.
Jones echoed this, saying he’d like to see “a better replication story” from Microsoft when it comes to virtualization, preferably included in his Windows license and not part of a separate storage system.
“I don’t want to pay more,” he said. “We’re paying more for storage [systems] than anything else, and a lot of what you’re buying there is the ability to do things like replicate.”
Beth Pariseau is a senior news writer for SearchServerVirtualization.com. Write to her at firstname.lastname@example.org.