LAS VEGAS—VMware is planning to “radically” change storage management and provisioning in vSphere environments, removing the need to set up LUNs, RAID groups and NAS mount points. In VMworld preview sessions this week, a VMware
VMware Principal Engineer Satyam Vaghani said the new APIs would use tools such as I/O Demultiplexers, Capacity Pools and VM Volumes. A VM Volume is a virtual machine disk file (VMDK) stored natively inside a Fibre Channel or Ethernet SAN.
Under the scenario laid out by Vaghani and VMware Storage Product Manager Vijay Ramachandran, virtual machine administrators and storage administrators would share management responsibilities through a common interface and process.
The presenters said VMware developed the APIs with EMC, NetApp, Dell, IBM and Hewlett-Packard, but they began the session with a disclaimer that none of those vendors has committed to support the APIs in their arrays. Vaghani declined to give a possible time frame for the integration of these APIs into arrays. However, a source from one of VMware’s storage partners said he expected the APIs to be part of vSphere 6.
Vaghani summed up his session by saying: “This is a radical new type of storage system. It’s probably the biggest disruption of NAS and SAN that you’ll ever see.”
Ramachandran admitted that many customers consider VAAI and VASA point solutions that aren’t granular enough and lack a general framework for future arrays.
“Customers want granularity of management,” he said. “Data management in vSphere is different than data management on storage arrays” because vSphere operates at the VM and VMDK level while storage arrays operate at the LUN and RAID group level.
He said a key requirement is that “data management should be exactly the same between storage arrays and VMDKs.”
No more SAN-vs.-NAS distinction?
Vaghani said the new APIs will enable vSphere to do application profile-based provisioning. “The goal is to enable storage systems to naturally store a VMDK as a distinct entity and provide VMDK with granular data services,” he said.
“We’re introducing VM Volume-based enterprise storage systems. VM Volumes mark the end of the SAN-versus-NAS debate.”
That’s because the new features operate the same way on block and file storage. The I/O Demultiplexer (demux) carries I/O from the host to the storage system and can front thousands of VM Volumes. Unlike a LUN, the demux holds no data. Capacity would be managed through Capacity Pools (CPs), which can span arrays and include profiles that allocate physical space and a set of data services. A CP profile could specify the frequency of snapshots and replication for the pool, for example. The CP is out of the data path, unlike a LUN or mount point.
The VMware admin would carve out a VM Volume and set quality-of-service parameters for performance and services. The storage admin would assign a default set of profiles to each CP that the VM admin can tweak on a VM Volume basis.
“Storage administrators can delegate without losing control of storage management,” Vaghani said. “For the first time, these two guys are on the same page.”
vSphere 5 has a VM Storage Profiles feature, but setting them up involves LUNs.
Vaghani conducted several demos using storage arrays from EMC, NetApp, Dell EqualLogic and IBM (XIV) to show the creation of CPs and profiles, how the relationship between the CP and an I/O demux works, how storage admins can define profiles on the array, and how a VM can be cloned on a VM Volume basis.
Like with anything new, there are unanswered questions about the new management tools. In a blog about the vStorage APIs tech preview sessions, EMC vSpecialist Chief Technology Officer Scott Lowe posted several questions. They include: How will IOPS requirements be handled, how granular will the policy attributes be, and are VMware’s promises even feasible?
Helping VM administrators manage storage is clearly a major goal for VMware, which walks a tightrope between complementing and competing with the features of its storage array partners—including its parent company, EMC. VMware CEO Paul Maritz said the goal is to complement.
“The biggest problem when you look at data centers is you need two teams of people,” Maritz said Monday during a press question-and-answer session. “One team is responsible for managing the VM side of things and provisioning and allocating resources, and another team is responsible for provisioning and allocating storage.
“A lot of our effort is going into creating interfaces that allow us to talk to storage solutions so we can tell them what we think is going on, and they can tell us what they think is going on.”
This article was previously published on SearchVirtualStorage.com.