Fresh off its acquisition of cache hypervisor company IO Turbine earlier this month, NAND flash memory maker Fusion-io announced a new product that allows virtual machines (VMs) to share physical NAND flash cache, accelerating performance.
Fusion-io has bundled IO Turbine’s hypervisor software, which allows cache storage on servers to be shared across multiple virtual machines through its ioCache adapter card. The combination allows VMs to share NAND flash to improve I/O throughput.
Salt Lake City-based Fusion-io, which filed for an IPO in March , makes specialized NAND flash cards designed to boost server throughput up to 278,000 I/Os per second while offering up to 1.28TB of data storage capacity.
Most NAND flash-based solid state drives (SSDs) shipping today mimic hard drives in form factor and serial ATA connectivity. On the other hand, Fusion-io’s ioDrive consists of NAND flash and a controller on a PCI Express (PCIe) expansion card. The PCIe technology offers throughput of up to 128Gbit/sec, while SATA tops out at 6Gbit/sec today.
When deployed in a VMware environment, the ioTurbine software dynamically provisions memory capacity and I/O performance across multiple virtual machines (VMs) by using Fusion-io’s flash-optimized OS subsystem called the ioMemory Virtual Storage Layer OS, which stores and serves data for each VM.
“The main value proposition behind it is it allows virtual machine servers to get three to six times the effective workload,” said Fusion-io CEO David Flynn. “This lets you get more workload per server by increasing the workload per guest machine, or by packing more VMs on the same machine.”
Available exclusively in bundles with ioTurbine software, Fusion’s ioCache delivers 600GB of capacity. The ioCache and ioTurbine virtualization bundle will be available this fall with a suggested retail price of $6,900 per physical server. Stand-alone licenses for ioTurbine software sell for $3,900 per physical server.
Flynn said the card can deliver a good return on investment because it will increase utilization rates of VMs by placing large amounts of flash-based cache right next to server processors thereby increasing the number of VMs that can be deployed per physical server.
He said the ioCache card also reduces the expenses related to powering, cooling and managing expensive traditional storage systems that struggle to meet the demands of virtualization.
“Microprocessors are doubling in capacity every 18 months, but the output servers are able to produce is not,” Flynn said. “You can’t get the data into the processors fast enough. If you double the processing, you need to double the data I/O rate.”