One of the primary reasons that IT organizations have resisted deploying more virtual servers is concerns over storage I/O performance.

With multiple applications sharing the same virtual serve, storage I/O requirements are hard to predict, which invariable leads to contention between applications for storage resources. The end result is the performance of applications running on a virtual servers can vary widely when multiple applications are trying to share the same I/O path.

To tackle this issue, Marvell unveiled at the Storage Networking World conference this week a DragonFly Virtual Storage Accelerator (VSA) that plugs into existing servers as a PCIe adapter, Designed to optimize the management of I/O across a virtual server environment, Shawn Kung, Marvell director of product marketing says that DragonFly can improve the I/O performance for applications running on virtual servers by a factor of ten.

Based on Marvell’s HyperScale embedded processors, , Kung says DragonFly acts a co-processor that provides the caching needed to optimize writes and reads to storage systems.

According to Kung,  the DragonFly VSA provides two levels cache composed of a level-1 non-volatile DRAM (NVRAM) cache and a level-2 cache created using solid-state (SSD) drives. The two-level cache enables consistent, high-level performance to applications running on the host server.

According to Kung, DragonFly VSA is essentially an appliance on a card that makes it possible to scale storage more effectively across a virtual server environment. For solution providers in the channel this should create an opportunity to accelerate the adoption of virtual servers in production environments, which is an idea that many IT organizations have been hesitant to embrace, in part, because of I/O performance concerns.