Server vendors and the dead hand of commoditisation

Comment The leading server vendors have known about the inhibiting effect of slow disk drive performance on their users for years, yet have done nothing about it. It's been left to companies like Fusion-io and Virident to solve that problem by inventing and popularising PCIe flash.

The problem is well known. When users run applications those apps need to be loaded into memory and access data. Memory access happens in microseconds whereas disk drive access takes milliseconds. The answer is in the "bleeding' obvious" category; insert an intervening layer of memory between DRAM and disk.

Fusion-io with its ioDrive and Virident with its tachION card use this concept and have their NAND flash memory inserted as an intervening tier of storage - slower than DRAM but faster than disk - between DRAM and disk. It can be viewed as a cache and other hairs can be split, but the effect is the same; apps run faster.

Michael Dell was an early investor in Fusion-io. But apart from that there was no interest in PCIe flash cards from the server vendors until Fusion-io started running various million IOPS demos with, for example, IBM. Now it and Virident have landed several server OEM deals and the concept is well understood.

Sun made clever efforts to use flash to speed up its ZFS filesystem and Solaris but these efforts did not spread to the big three - HP, IBM and Dell - which preferred to wait for a third-party, commodity solution to the problem rather than engage in their own engineering development.

Minor server manufacturers such as Acer, Hitachi and even Intel with its white box efforts, were also conspicuously absent from the PCIe flash party. We are seeing the dead hand of commoditisation. By all means make a better server mousetrap but the main architectural ingredients inside the server box, the X86 CPU, DRAM, PCIe bus, third-party O/S, and disk drive storage, stay inviolate.

Commoditisation crippled innovation in an area which mattered, mattered deeply because virtualising servers made the disk I/O problem worse.

In fact virtualisation, the work of VMware, was itself directly related to server disk I/O performance drag because earlier efforts to keep the server or desktop PC CPU busy when disk I/O delays caused it to be idle, meaning multi-tasking operating systems, were so bad at sorting out the problem.

VMware is, after all, just a glorified way of multi-tasking apps in servers and, originally, PCs, that was necessary because Windows and Unix were so crap at the job.

In the storage world there was one vendor that did see the importance of a flash memory layer in the server. That was NetApp with its PAM (Performance Acceleration Module) now called onomatopeically the Flash Cache. The server in question is the X86-based controller in NetApp's FAS arrays. Kudos is due to NetApp's engineers for looking further than their nose when fixing the problem. No other storage array vendor did that at the time, except Sun.

Therefore we can argue that NetApp and Sun (now Oracle) are the two most innovative server suppliers, while Acer, Dell, HP, and IBM are the least innovative. What do you think? ®