vSphere storage features
The most significant change to storage management in vSphere 4 is the addition of thin provisioning for virtual disks and hot extension for virtual logical unit numbers (LUNs). These features allow for expansion of virtual disks on the fly.
Thin provisioning lets server admins allocate storage as needed, similar to the way array-based thin provisioning works today, but administered through the VMware vCenter management system at the server level. At last fall's VMworld user conference, VMware officials had spoken about linking VMware's thin provisioning with array-based thin provisioning, but that integration isn't available in this release.
"Long term, we may open APIs so storage arrays have visibility as to where virtual machines reside in volumes, but today we don't have that," said Jon Bock, VMware senior product marketing manager.
The new hot-extension feature lets administrators expand LUNs at the Virtual Machine File System (VMFS) layer without downtime. "So when the storage array expands a LUN, the VMFS volume can grow – it makes things less painful if you run out of space," Bock said.
To support these new provisioning features, vSphere 4 will also provide a virtual machine's view of storage systems, as well as monitoring and alerting on virtual machine storage consumption for server admins. VMware Storage VMotion, which migrates virtual machines on back-end storage (if necessary) to support server migrations with VMware VMotion, can now be deployed using NFS, iSCSI or Fibre Channel protocols and be managed through the vCenter GUI. Previously, Storage VMotion required a command-line interface.
While VMware offers its own basic multipathing features for virtual machines, APIs in vSphere 4 add more advanced multipathing software offered by storage vendors, such as VMware parent company EMC Corp.'s PowerPath. VMware's multipathing offers basic load balancing and failover, Bock said, but specialized multipathing programs can also perform more advanced load balancing with visibility into application queues and faster failovers.
VMware will support 10 Gigabit Ethernet (10 GbE) iSCSI devices with vSphere 4. VMware's SCSI drivers can now be "paravirtualized" or aware that they're virtual systems communicating with the ESX layer within a box. This can create a fast path for storage traffic because the SCSI driver doesn't act as if it has to go out to an external storage switch, Bock said.
New data protection offering and integration
vSphere 4 will include a new data protection product and a new approach to integration for existing third-party backup tools.
VMware is looking to compete better with Microsoft's Hyper-V, which can offer low-end customers built-in snapshots for virtual machines with DPM. VMware's new DataRecovery product is based on the same vStorage API it has extended to third-party data protection vendors, and offers similar capabilities. It takes a snapshot of a virtual machine and then does a granular backup from that snapshot to remove the processing load from the production machine. DataRecovery can also do file-level recovery from a VMDK backup.
Bock said VMware isn't looking to compete with third-party backup offerings, but is aiming at smaller customers without a good backup tool. DataRecovery will be limited to disk-based backups of virtual machine images only, and will support up to 100 virtual machines or 2 TB of storage.
For enterprise backup vendors, VMware will offer direct API-based integration for backups, eliminating the need for VMware Consolidated Backup (VCB). VCB required a customer to add another software module to the backup server, and then the customer or backup vendor—or both—had to join the backup software with VCB through scripting. VMware will now open up APIs to its backup partners, allowing backup software to query virtual machines directly. Bock said the change should simplify deployment for users.
Delicate dance with storage partners continues
VMware customer Daniel Lewis, manager of network services at the University of Southern California's Marshall School of Business in Los Angeles, said the low-end backup capabilities are wasted on an enterprise product. "I know how the business model works," he said. "They need an open architecture to some degree."
But if VMware is going to pull backup features into its software, Lewis said he wishes the vendor would go whole hog.
"A low-end backup product doesn't belong in enterprise software," he said. "If you're going to go for the low end, OK, put it in the workstation product. But if you're going to do something for the enterprise, do something meaningful."
Otherwise, "hot expansion sounds fantastic," Lewis said. "We build systems small anyway, and every so often we have to take them down to expand them."
Thin provisioning, on the other hand, "has always been one of the things that kind of scares me," Lewis said. "Unless you have really, really good monitoring and alerting, you can hurt yourself really badly. If you have that intelligence built in, that's one thing; but if it's manual and intertwined with the storage management, I don't want to deal with it."
However, "if I go to an array with thin provisioning, it means I have to have it on the entire volume—doing it at the VMware level gives me the option of doing it on a single server," he noted.
Chris Wolf, a senior analyst at Midvale, Utah-based Burton Group, said thin provisioning will be another area where VMware must balance competition and cooperation with its storage partners.
"I'm not really hearing VMware articulate where to use VMware thin provisioning vs. the array," he said. "So far, though, our clients want to let the feature bake a little bit and see what the performance is like—meanwhile, they're already using array-based thin provisioning."
vSphere 4 is expected to be generally available by the end of June.
EMC says its PowerPath/VE software has been integrated with VMware's APIs for multipathing. PowerPath/VE can automatically discover network paths and automate load balancing between them. PowerPath/VE is priced at $1,430 per server.