8 Disruptive Technology Changes

From a technology perspective, we are in the most disruptive phase of Information Technology since the computer was born. It might not seem that way, but if you list the disruptive technologies currently in play, it paints an intriguing picture. So here’s a list:

1. Multicore Chips

Multicore chips have been around for a year or two. They were introduced as Moore’s Law ran out of steam. Each core is a distinct cpu and thus it is possible on PCs and laptops to simply split the workload in two and harness both cpus reasonably efficiently. However as you increase the number of cores on a chip, this strategy starts to fail and you need to specifically program for multiple cores using programming languages and models designed for the job. (A UK parallel programming consultancy, Concertant does an annual survey on multicore. If you’re interested in completing this, here’s the survey link.)

There has been talk of chips with hundreds of cores and, if general purpose uses can be found for them, they will likely be introduced in the coming 5 years or so. Multicore favors virtualization, so there is little doubt that multicore will persist both on the server and on the client at least up to 8 cores. However, the destiny of multicore may also be tied to the GPU.

2. The GPU Absorbs the CPU

This strongly relates to the previous multicore disruption. Nvidia believes that the GPU (Graphics Processing Unit) will absorb the CPU in the coming years. Clearly, that would rain on Intel’s parade in no uncertain terms, and of course, the sun would shine like a diamond on Nvidia’s parade. Nvidia has a point though. As interfaces increase in sophistication it’s the GPU that does most of the work, the CPU having topped out quite a while ago. As far as I can see there’s no reason not to put a few x86 core onto a GPU and make the CPU disappear, as in a magic trick. For further details read What is a GPU and Why is it Important?

3. Memory Replacing Disk

I wrote about this in detail quite a while ago here. The point is that flash memory, which is non-volatile and provides much faster access than disk, is gradually replacing disk. Samsung’s recently announced 256GB solid state drive (SSD) is 2.4 times faster than disk. That doesn’t sound too revolutionary, but the point is that nearly all the business applications we run assume that we are pulling data from a spinning disk and soon that will not be the case. Hence there will likely be an opportunity to rethink database at the physical level.

4. The Network Infrastructure Absorbs the Servers

Recently introduced all-singing-all-dancing network switch products can deliver astonishing network speeds – in the terabyte per second area. For full details see The Server Vendors v Cisco: Is This A New Technology War? and Brocade: The Heart and Lungs of the Network. Note that this bandwidth is greater than the bandwidth between memory and the cpu on a server board. (If you are going to get this fact in perspective, you need to understand that the distance between memory and CPU is very small, so the latency of a single instruction is lower memory-to-cpu than over the network, but the network bandwidth is higher because it embodies huge parallelism on fiber). The point is that with networks this capable it is feasible to have an architecture that sends instructions and data anywhere to execute. It diminishes the need to keep the processing close to the data. The ultimate consequence from this is that if you want to manage workloads on a network, you need to do it from the network infrastructure, which, naturally, has an operating system of its own. This will inevitably happen in time.

5. The Network Becomes Real-Time (or near-as-dammit-realtime)

Communications has been converging with computing forever, but only recently have they met and shook hands. The upshot is that IP Addresses becomes the physical communications addresses, Unified Communications rule (see What Is Unified Communications?) and Presence becomes important (see What is Presence and Why Should I Care?.) You should be able to get an immediate understanding of how computer networks become realtime simply by looking at the diagram on this posting and asking yourself whether it needs 99.999% uptime (yes it does).

6. The Network Gets A Management Circuit

A recent, almost silent, trend has seen the installing of Service Processors in servers to improve up-time and provide a circuit by which a network can be better managed. The story is told in greater detail in What is a Service Processor? And Why Should I Care?

7. Virtualization: Everything goes Virtual

I’ve written about virtualization extensively. You can find a page full of Virtualization links here. You could say that it all began with the advent of SANs, which eventually created a common storage pool which you could think of as a “virtualized storage resource.” With the advent of multicore, virtualization just made more and more sense, and it now looks to be the fundamental pattern of how processing resource is allocated and workloads are managed. Naturally the desktop gets pulled in to the virtualization whirlpool and as time passes, we end up with a “common resource space” with end-devices providing access.

8. Cloud Computing

Finally we have cloud computing where pretty much every resource and capability, hard or soft, is provided as a service. For a full description read Everything as a Service: The Growth of Cloud Computing

Anything else?

I could add that everything mentioned so far (hard or soft) could feasibly be addressed using a single 64 bit addressing scheme. That would be game changing, given that we live in an age of 64 bit computing that doesn’t yet exploit 64 bit addressing.

I could also mention that graphene is very likely to replace silicon in chips in the coming 5-10 years, which means that Moore’s Law would get a new lease of life and that the heating problems in data centers would evaporate. (See Graphene: The New Silicon

The point of summarizing all of this, by the way, is to provide a global perspective on the blistering pace of change. If you have anything else you think should be included, please post comments.

  1. June 26th, 2008 at 11:22 | #1

    This is a great post. I’m going to study these concepts and links, especially those having to do with “presence” and “realtime.” I’ve added your feed to my list.

  2. June 26th, 2008 at 11:33 | #2

    Robin — thanks for this great post. I have to admit a lot of this is over my head but I am going to study it. Even if I’m not in hi-tech I need to know what the issues/opportunites are.

  3. June 26th, 2008 at 22:31 | #3

    Great Post Robin and thanks for links for further exploring the topics.

    I’d like to add location based services on mobiles as another disruption. Although LBS has been around for a while, it’s only now with desktop-class web browsing experience and APIs that expose native phone services to applications that LBS are becoming viable. They are bound to change the way in which we interact with the world.

  4. Robin Bloor
    June 26th, 2008 at 23:36 | #4

    You are right about this, although I think of this as a different set of dimensions, and that particular disruption must also include RFID 2.0. I’ll probably discuss that as a separate topic.

  5. June 28th, 2008 at 10:49 | #5

    I felt a bit silly bringing up GPUs at a recent VMware awayday, it seemed to make sense but I wasn’t sure why. This adds another dimension, which makes 3 – graphics intensive front ends, embedded x86 and the GPU being capable of processing parallel workloads. Cool.

  1. June 27th, 2008 at 17:34 | #1
  2. June 30th, 2008 at 05:12 | #2
  3. October 1st, 2008 at 15:06 | #3