PCIe Rides the SSD into the Enterprise

Arthur Cole

SSDs hold as their claim to fame faster throughput in high-traffic environments, such as Web servers and database platforms. So it's no surprise, then, that many of the newest enterprise designs are turning to PCI-Express (PCIe) as their primary networking format.

Earlier this month, we spotted a number of new PCIe-based flash technologies from the likes of Dolphin ICS and Smart Modular Technologies. Since then, a whole new crop of systems has hit the demonstration floor, each promising to boost the already impressive I/O of SSDs even further.

NextIO has teamed up with Marvell Semiconductor to match a PCIe-based SSD with NextIO's virtual networking technology to deliver more than 200,000 IOPS and 400 GB across a single PCIe slot. The entire system is expected to hold 12 PCIe slots, bringing overall throughput to more than 1 million IOPS and total capacity to 4 TB -- all within a 3 U chassis that draws less than 500 watts when active.

Fusion-io is bringing PCIe to its latest SSD, the ioXtreme, aimed at gamers and other consumer uses, but equally well-suited to enterprise applications. The device has an 80 GB capacity and provides 520 MBps average bandwidth and 617 MBps burst bandwidth, which the company says are more accurate gauges of real-world performance than maximum read/write measurements. The drive is on the pricey side, $895, and is only compatible with 64-bit Windows or Linux, but it does offer a lengthy lifespan of 2 million hours MTBF and six years write endurance, according to the comany.

PCIe is also making its way onto the wider enterprise network. Netronome Systems has turned to Denali Software to add the protocol to its NFP-3240 Network Flow Processor, the company's 40-core packet processing system aimed at high-volume networking environments. The Denali Databahn PCIe core supports the PCIe 2.0 implementation that provides single-root I/O virtualization (SR-IOV) and verification IP technology suitable for unified computing architectures. The NFP-3240 provides 256 queues of I/O co-processing and is capable of delivering more than 56 billion IOPS over 320 hardware threads.

Most state-of-the-art PCIe systems feature the 2.0 version of the protocol, the so-called Gen-2. But it won't be long before Gen-3, with its 8 Gbps functionality, becomes the norm. Although that standard is still under development, it's far enough along for some vendors to bring initial products to the channel. Snowbush, the IP division of Gennum Corp., recently introduced a four-lane Gen-3 PHY (physical layer) and endpoint/root/dual-mode controller block, a combination that has already been picked up by PLX Technology for its newest switch chip. Gennum was quick to point out that the design is highly configurable, meaning it should be able to accommodate any changes to the standard outside of something major, like a reconfiguration of the main encoding scheme.

PCIe is not likely to become a central enterprise fabric any time soon. But the fact that it has pushed its way beyond the server rack and is now emerging as a short-haul networking solution at a time when SSD technology is hitting the enterprise is evidence that you can never tell what will happen to a technology once it gets into the hands of creative engineers.

It also means that any effort toward unified networking fabrics would be wise to incorporate PCIe along with Ethernet, Fibre Channel and InfiniBand.

Add Comment      Leave a comment on this blog post

Post a comment





(Maximum characters: 1200). You have 1200 characters left.



Subscribe to our Newsletters

Sign up now and get the best business technology insights direct to your inbox.