The new vSphere Storage Appliance has serious limits. Is the EMC relationship keeping VMware from realizing its full potential? As I mentioned in my wrap-up of VMworld 2011, one of the shiny new VMware products to hit the shelves this year is the vSphere Storage Appliance, or VSA. The VSA delivers SAN-like features to small businesses that may not have the budget, staffing, or technical sophistication to purchase a real SAN. But the initial release of the VSA is marred by a number of serious deployment and scalability limitations. The real story behind VMware’s introduction of the VSA has less to do with the VSA itself and more with VMware’s relationship with its partners and its parent company, EMC. What is the VSA? The vSphere Storage Appliance is a cluster of virtual machine appliances that offer local storage as a unified, highly redundant datastore. A VSA cluster can currently span either two or three vSphere hosts, and it uses network-based synchronous mirroring (RAID1) combined with local RAID10 on the hosts to ensure data availability and to deliver reasonable performance. The vSphere cluster hosts are then able to connect to the VSA cluster’s datastores via NFS — effectively allowing the hosts to see and share each other’s storage in real time. In a two-host deployment, the first appliance operates a datastore roughly half the size of its local storage while the other half is used to store a mirror image of the second host’s datastore, thereby distributing the storage load across both hosts. When a third host is added to the picture, there are three total datastores with each host backing up one of the other two in a round-robin fashion. Although the network-based mirroring introduces a degree of latency, a VMware engineer told me informally that a three-host cluster using a total of 24 15k SAS drives can typically push around 3,000 IOPS in real life — generally enough for most small-business customers. There’s another trade-off: The combination of RAID10 on the local hosts and mirroring in between the hosts is fantastically inefficient from a capacity perspective, effectively resulting in about a quarter of the physical storage being available to store virtual machines. For that reduced capacity, small-business customers get a SAN-like solution much less costly and complex than an actual SAN. There are other significant limitations. First, the VSA can be deployed only in a greenfield environment (if you have existing virtual machines on local storage, you have to move them somewhere else prior to implementing the VSA). Worse, the VSA cannot be scaled once it’s installed. If you install with two nodes, you cannot scale to three nodes afterward, nor can you add more physical disks to your hosts and expand the footprint of the VSA. Speaking from personal experience, those two limitations alone would prevent me from wanting to implement VSA in any environment that might experience unexpected growth — which, these days, is pretty much anywhere. In other words, the VSA is a potentially great product with a couple of show-stopping problems. If VMware fixes them, the VSA could become the de facto answer for any small business deploying vSphere. Its simplicity, the fact that it’s managed entirely from within vCenter, and its low cost compared to all but the cheapest entry-level SAN/NAS could make it a very compelling option. VMware’s motivation At first glance, the VSA seems like a fairly logical step for VMware. Over the last few years VMware has migrated high-end features such as High Availability and vMotion (which both require shared storage) down to its entry-level vSphere Essentials Plus bundle. VMware allows its customer base to use those sought-after features without buying and deploying a SAN — seems pretty cut and dried. It’s anything but simple. To begin with, HP, one of VMware’s closest and most supportive partners, has been selling its own VSA (yes, even the same acronym) since it acquired LeftHand Networks back in 2008. HP’s P4000 Virtual SAN Appliance is much more feature-rich than VMware’s offering in a number of different ways. It doesn’t have any of the scalability or deployment issues that the 1.0 release of VMware’s VSA has. Plus, it’s capable of array-based replication to other P4000 SANs and provides an easy VSA-to-physical P4000 migration path. If there’s already a good product that fits the niche and it’s sold by a partner, why wouldn’t VMware promote it rather than developing its own product? For one thing, VMware is largely owned by EMC — a direct and occasionally vicious competitor of HP’s. I’d wager that hell would need to freeze over before an EMC property would promote a chunk of HP storage tech. Second, the management of VMware’s VSA is built entirely into VMware vCenter. Plug-ins allow you to manage P4000 arrays from within vCenter, but they are not the only way to manage it, which means that VMware doesn’t really control the environment. Third, HP’s VSA is also compatible with Microsoft Hyper-V — a fact that might not make VMware terribly comfortable. Even if VMware just thought it could build a better mousetrap (and isn’t quite there yet), there are other interesting aspects to this story. The primary goal of the VSA is to allow small businesses to avoid implementing a SAN and still have access to shared storage features. That means EMC and others hawking entry-level SANs will probably lose business if the VSA develops into a solid, full-featured product. Would EMC really be happy about the VSA eating into entry-level VNXe sales? Putting it all together The vSphere Storage Appliance is a neat idea that needs work. But if VMware keeps at it and introduces some deployment and expansion flexibility — as it has said it will — the VSA could become a real asset to small businesses and gain popularity in ROBO deployments in larger enterprises. VMware continues to evolve into becoming a real storage company. As it does so, the relationship with EMC will become a bigger problem. Whether or not EMC actually pushes VMware in certain architectural directions, or customers or partners just think so, the perception of a conflict of interest may be damaging. More to the point, as VMware’s virtualization competitors edge toward feature parity with vSphere, VMware cannot afford to be constrained by EMC’s agenda. Maybe it’s time VMware spun off — all the way off — from EMC. This article, “vSphere Storage Appliance: The fruit of a conflict of interest,” originally appeared at InfoWorld.com. Read more of Matt Prigge’s Information Overload blog and follow the latest developments in storage at InfoWorld.com. For the latest business technology news, follow InfoWorld.com on Twitter. Software DevelopmentTechnology Industry