New products designed from the ground up to specifically serve storage for virtual servers can offer dramatic savings in terms of dollars and the time spent managing storage.
Over the past few months, Taneja Group Labs has done a considerable amount of work with products I'll call next-generation virtual infrastructure storage, including Hewlett-Packard's StoreVirtual VSA products, Tintri's VMstore and VMware's Virtual SAN (VSAN). Some of them are described as software-defined storage, while others are more akin to virtualization-specific storage, but they all represent an emerging class of storage that appears poised to fundamentally alter virtual infrastructure administration.
Each product has proven intriguingly useful in making virtualization storage a bit more agile and much more automated, even though each approaches the issue from a different angle. StoreVirtual VSA is deeply injected into the virtual infrastructure and has a uniquely versatile approach to scalability: basically, you can run it anywhere and keep adding more to scale or adapt. In a recent report, we recognized that Tintri's virtual-infrastructure-dedicated VMstore is so deeply integrated with virtual machine (VM) and VM-specific tasks that it makes the act of managing virtual storage almost disappear -- to the degree that we called it a 60x reduction in the time administrators spend on storage management tasks in the virtual infrastructure. Meanwhile, VMware's VSAN is wrapped in a logical organizational and configuration layer that highly automates ongoing management, what VMware calls Storage Policy-Based Management (SPBM).
Cutting storage costs in virtual environments
These are just a few of the products coming to market, but the makings of a sea change are clearly evident in these new approaches to virtualization storage, which merits serious evaluation. Why? Because virtualization-specific storage may have a bigger impact on storage costs than anything we've seen, and storage costs are a really big deal in the data center scheme of things.
Based on our testing and conversations with vendors, Taneja Group uses a rule of thumb that midrange storage still costs the typical organization $3 to $5 per gigabyte (GB) to purchase, and then $5 to $8 per GB annually to operate. Over the life of a typical storage array, operating that storage will cost many times more than the original purchase price, with the vast majority of that cost the result of the time and effort spent managing the system. Since there are still non-virtual systems in almost every data center, virtualization-specific storage may not replace all of a data center's storage but it will replace enough to radically reduce storage costs.
New virtual storage systems bring fundamental change
Let's examine four ways these new storage approaches can change storage management.
1. Speedy configuration and reconfiguration. Integrating storage with the virtual infrastructure makes it more agile and more easily adaptable to changing virtual workload requirements. Such storage systems can typically be deployed and then drastically altered (for performance, protection or availability) from within the virtual infrastructure, and without the need to reconfigure a separate storage layer.
2. More automation. If storage is integrated closely enough with the virtual workload, it's also often highly automated. This can practically eliminate configuration and provisioning and, in effect, "pre-organize" the storage infrastructure. This can make it easier to deploy VMs or copies of VMs with the right level of performance, availability and protection, and then monitor and confirm how storage is provisioned. This is at the heart of VMware's VSAN SPBM, but similar capabilities are springing up elsewhere. This is also why Fibre Channel storage vendors have long anticipated VMware's long-promised Virtual Volume (vVol) functionality that's beginning to look a bit like vaporware. According to VMware, vVol will change a storage system's focus from LUNs or volumes to its own VMDK file architecture.
3. Easier management. We're also seeing how this concept of automation and storage policies can be extremely powerful in the continued management of the storage infrastructure, and in ways that may not be obvious. For example, consider how an application needs to move from test/development into production today. With virtualization-specific storage, simply changing a policy could alter the performance and availability configuration for 10, 20 or even 100 different VMs all at once; no cloning, migration or anything else required.
4. Improved insight. Ultimately, when storage becomes deeply integrated and VM-specific it usually enhances visibility, which can fundamentally alter the day-to-day operation of storage. If integrated enough, a virtualization-specific offering may be able to collect highly granular data -- from VM latency to host latency to storage latency, for example -- and, in turn, provide better visibility into troubleshooting and tuning application performance.
By any other name ...
Call it what you want -- software-defined storage, virtualization-specific storage or VM-centric storage -- but the real story is that these capabilities are obtainable today and make up the capabilities and benefits of some seriously innovative virtualization storage products.
Still, there are some caveats with virtualization-specific storage. The biggest is that early products look poised to change responsibilities in the data center, with fewer administrators involved in managing storage. Historically, IT hasn't exhibited a great deal of good governance in managing storage. As more tasks become consolidated in the hands of fewer administrators, it's worth thinking about how we maintain responsible use and control of storage. The upside is that a reduction in operational costs might allow us to rethink the operation of storage and, for the first time, invest in much better storage operations. But getting there will take some foresight and planning. Now is the time to be thinking about how you might "do" storage better in your company if you could radically reduce the time and effort spent on its upkeep.
About the author:
Jeff Boles is a senior analyst at Taneja Group.