Virtualization pushes traditional storage architectures to their performance limits, prompting companies to look at server-dedicated solutions, typically PCIe and SSD flash, to solve their performance problems. The challenge with server dedicated solutions, since they’re limited to the storage devices that can connect to a single server, is that they often can’t provide the raw capacity that’s required. They’re also not cluster aware, critical for a virtual infrastructure. PernixData’s FVP software provides a host-level acceleration solution that can pool flash and RAM across servers while complementing existing shared storage.
FVP runs in the hypervisor kernel to accelerate read and write workloads on a per-VM basis. It creates a server-side performance tier of storage using resources like SAS SSDs and PCIe flash cards. FVP can also use available RAM for acceleration, addressing its volatility with “Distributed Fault Tolerant Memory” (DTFM), a technology that prevents data loss even if the host, the device or the network fails.
Pooling Storage Performance and Storage Capacity
PernixData’s FVP Cluster technology pools local resources to create a virtual high-performance tier across servers. In this way any VM can access the flash or RAM that’s available on any other host, as well as on its own host. PernixData’s clustering technology allows all servers to share that high-performance tier, making more flash and RAM capacity available than would be possible within a single server. The result is improved efficiency and greatly increased flexibility in a scale-out architecture.
FVP 2.5 Upgrades
In the latest release, FVP 2.5 makes using RAM for performance acceleration an even more viable alternative to flash with DTFM-Z, an implementation of adaptive compression to increase the effective RAM capacity by up to 4x. With more data stored on the available RAM, more applications can be accelerated by this highest performing media, providing the lowest latency storage for critical use cases.
With Intelligent I/O Profiling, FVP 2.5 can also detect, in real time, which workloads are most appropriate for the high performance tier and which can bypass server-side acceleration and be routed directly to shared storage. This saves flash and RAM space for the most important workloads and can reduce wear on flash as well. FVP can apply this optimization function at the VM level.
Also new with FVP 2.5 is Role-Based Access Control (RBAC). This feature enables administrators to control which users have access to the PernixData UI and what information they can see. Service Providers can use RBAC to support multi-tenant environments and enterprise customers can use it to tailor access to specific departments’ VM acceleration data.
SAN’s Best Friend
There are several solutions, like VMware’s VSAN, that virtualize server-side capacity and replace traditional shared storage as well as the existing server footprint. But PernixData is different. It’s designed to complement the existing infrastructure, not replace it. It’s also non-disruptive to the existing storage services on those shared storage systems that IT trusts for data protection and application availability.
One of the themes that PernixData uses when they talk about this product is ‘intelligence’, leveraging information about VM workloads to optimize server-side resources and accelerate data. This intelligence also moves high-stress I/O traffic off of the SAN, saving the network resource for its most appropriate data sets as well. But there’s another aspect of this product that speaks to intelligence.
FVP works with existing resources, both server-side and on the storage tier, which is a more intelligent implementation strategy than ‘rip and replace’. Software that installs on the host and maximizes performance with minimal disruption while leveraging existing resources is one that’s much more appealing for most IT managers.