Storage performance has become a major issue in virtual environments, writes Wikibon Analyst Scott Lowe. The normal answer — flash server cards and arrays — are an expensive fix, particularly when they replace perfectly good disk systems. This year several vendors have appeared offering software-only solutions that can fix the problem for some users at a fraction of the cost of hardware solutions.

Lowe provides a high-level look at how three of the market leading products work. Each takes a different approach to increasing storage performance.

Infinio Accelerator creates a RAM cache between the hypervisor and physical storage layer on each host in a cluster. It runs in two vCPUs and uses 8 GB of RAM on each host, aggregating them into a read-only storage cache that does not require any flash or other new hardware. The disadvantage is that it is read only, but most organizations do many more reads than writes. And with a free trial period and a cost of $499 per socket it can be ordered on a charge card, downloaded, and installed with no impact on running VMs.

PernixData FVP sells a “Flash Hypervisor” that aggregates the solid-state storage across all the hosts in the cluster to create a scale-out data tier that sits between the VMs and storage arrays. It supports both reads and writes and can scale as more hosts or server flash is added to the cluster. It operates as a hypervisor module rather than a virtual appliance, which in some cases means it has less latency than appliance-based systems.

Proximal Data AutoCache also leverages existing server flash storage, caching hot data on host-based solid-state disks or PCIe cards. It provides a full read cache with write-through support in which writes are acknowledged only when they hit primary storage.

Lowe recommends that CIOs experiencing performance issues with legacy storage in virtualized environments try a software solution before investing in expensive flash hardware solutions. At worst if it doesn’t fix the problem it can be deinstalled. At best it may delay the CAPEX expense of new flash arrays for a year or more, during which time the price of those arrays will continue to fall.