Computer Architecture Predictions 2015 and beyond

In the old days as horse coaches were replaced by steam engines the width between the wheels of the train coaches where kept the same simply because the tools being used to forge the chassis components and the people producing the coaches merely witnessed an incremental change even though transportation technology took a major leap moving from horses to steam power.

Perhaps you have heard the story that the dimensions of the Space Shuttle fuel tank, being transported from a safe place out to the shuttle launch pad, in fact was limited by the with of the tracks and hence in fact was defined by the size of two horses standing side by side. Architectural decisions live longer than anticipated.

We have seen such examples in many areas and computer architecture is no exception.

Over the past few years data storage have gone through great changes from a media technology standpoint by the advent of silicon based (FLASH) media flooding every aspect of our lives hence bringing the price per bit down dramatically.  In fact most everyday electronic appliances we come across these days are likely to have a flash storage element hidden inside.

The incremental change here is that we see companies offering products that is “simply” changing the media from spinning disks to silicon memory leaving everything else intact: the protocols, the physical connectors, the size of the physical box holding the media.

Another observation is a change from a software standpoint where software replaces traditional physical building blocks in computers, storage networks and storage devices, a.k.a. Hyper Converged Systems.

There are indeed a few examples of interesting technology showing up on the horizon, items that may appear to be more ground breaking than what we have seen so far.

One such change that holds the promise of big impact is the connectivity technology originally introduced by one of the storage media vendors called Kinetic drives where the protocol used to talk with the actual storage devices is promoted up a few layers from low level SCSI to HTTP rendering completely new and interesting ways to assemble computer systems. As a side effect the storage controller be it a monolith or internal to the server chassis is then politely omitted and replaced by software.

The three elements I am contemplating here caught my attention at about the same time. In concert they could completely change the way compute and storage systems will be architected and built over the coming 5 to 10 years.  The keyword: simplification.

Future of clusters pmard

The picture above is an attempt to create what turned out an almost organic looking visual representation of a highly resilient cluster using the triad.

– The Kinetic IP attached object media storage (Seagate)

– The Memory Channel Storage  (MCS) technology (Diablo and SanDisk)

– The Storage Virtualization Software piece eventually enabling the two items above play together (Atlantis Computing, USX)

Given some time this triad hold the promise to fundamentally change the architectural landscape of both compute and storage. The change will offer much higher speeds (IOPS) and Virtual Machine (VM) density, reduced power consumption through physical simplification and inline data reduction technology and offer web scale capacity.

What we do know from experience is that these things never happens over night so don’t hold your breath: there are strong forces across the business spectrum holding back such fundamental simplification attempts: billions of dollars invested in traditional ways of doing things, but it will come. It always does, eventually. From a customer vantage point however this is really good news.