Note: You are experiencing only the raw content of this site, without the intended layout and design. Either your browser has ignored the Cascading Style Sheet (CSS) files for this site, or you are using an outdated browser which does not support Web Standards. Learn more.

EDS' Next Big Thing Blog: Read and Respond to What the EDS Fellows Say About Technology

Read and respond to what the EDS Fellows have to say about the future of technology on EDS' Next Big Thing Blog on eds.com.

The data centre of the future - part 2 (forces of change)

by Charlie Bess

In the last entry, I briefly touched upon the concepts of commoditization and virtualization. There are other major factors like integration and complexity/simplicity that will affect the data centre of the future.

The forces of standarization/commoditization and virtualization will drive down the cost of the data center and reduce the time to get a configuration online, providing greater capability through the assembly of lower cost processing capability into massive networks of computing capability. Even if the speed of processors starts to flatten (as speed is limited by the ability to remove the heat they produce), the battle of numbers through multi-core techniques is just starting. As the industry learns how to create higher cores per die in volume, we should expect a rapid increase in capability, likely higher than what Moore's law would have predicted for a single processor solution. Since we'll be able to use all those transistors more effectively.

Until now, data centre complexity was managed using the most flexible tool available - people. With the increase in processor speed, size of storage and speed of network mentioned in the previous entry, the data centre environment is much more complex. Unfortunately, having a huge, virtual network of capability needs extensive integration functionality and introduces management complexity that is too much for anyone to understand.

Data centres of the future will need to perform real-time, automated assembly of modeled configurations based upon performance parameters. These same parameters will feed into automated monitoring capabilities in order to simplify the management process into something that people can understand. They'll also have improved predictive and response capabilities, bringing processing capability online and spinning down capabilities before they break or cause disruption. Once these capabilities become available, the concept of a true utility computing approach that can be drawn upon to meet general business needs will be a reality.

Published Wednesday, November 02, 2005 1:49 PM

Subscribe to this post's comments using RSS

Comments

Posted by Colonel Nikolai Wednesday, November 02, 2005 6:29 PM

While I like the ideas you are presenting for future data centers; real-time automated assembly of modeled configuration based upon performance parameters, etc, I think for this sort of thing to really happen, a major culture change will have to ocurr first. The data centers we have today are, in practice, little changed from what they were when the first data centers were created, with the only major technical advances made when the switch from batch to interactive came into existence, whereas even that isn't as deep-rooted as we may think. I'm finding in my experience that, for instance, MQ Series plays a huge role navigating the impedence mismatch between batch and interactive / realtime technologies in today's modern data centers, allowing the mainframe people to maintain their batch mentality in the face of an increasingly interactive / realtime world around them.

The answer, I believe, lies in looking at what Marshal McLuhan said more than 30 years ago. "Real technological innovation is almost always predicated upon social change, often radical in nature."

We can talk about technological enhancements all day long. But they will never come to pass without the precursing catalyst of social change. History is filled with examples. From the invention of the wheel, the development of written languages, horse-mounted warfare to moveable type, social change, radical in nature, was required to sustain the innovation and change the world.

Today, for instance, I forsee the two major triangulating issues of our time coming into full force. These issues will be the true disrupting catalysts of the datacenter, indeed of innovation at large. Those two issues are Climate Change and Peak Oil. A premature prediction that datacenters will need to become much less power consumptive and much more mobile might be worth trying on for size, even if the prediction is wrong.

This change is not only a Good Thing; it is necessary. It is not only necessary: It is ... coming.

Posted by Charlie Bess Thursday, November 03, 2005 2:15 PM

I agree with your comments wholeheartedly. We are already seeing with announcements from Sun and others about a strategy with much lower power consumption. When data center consumption is measured in KW/sq. foot, its clear that this is a huge cost issue.
Fortunately, with the higher densities described by Moore's law, lower gate voltage requirements come along as part of advancement. With the movement to greater parallel processing (multi-core) and less reliance on higher clock rate to push the work through, power consumption goes down accordingly.
The cultural change issues required to incorporate the advancements are significant. When the other compute service providers have a significantly lower cost base, traditional cost cutting techniques will be insufficient. You can either plan for it or react to it. Arguing about it is a waste of time.

Posted by Jay Prestipino Thursday, November 10, 2005 5:33 PM

I buy into the value of virtualization and utility computing and that it will happen at some point in the future. I believe many other factors will effect real utility computing near term. First companies got in the mode of owning their own networks (primarily the gear) many years ago and staffing has been built around it. The origin of that was cost reduction and control. While staffing has been reduced by outsourcing plans companies still seem to want to own their own gear. Very strong vendors have sold the concept of ownership well. It seems that the market is potentially willing to shift from ownership if someone can show a model that works for all the constituents. Vendors don't want to own equipment, just sell it. Each has an agenda to provide the "full" solution. Customers are frustrated by the pace of technology change, security problems, integration and network costs.

While computing is reaching the the point of utility there has to be a change in the underlying basic thinking around ownership cost and the perceived benefit.

Posted by Charlie Bess Thursday, November 10, 2005 9:35 PM

Jay - when you say there has to be a change in the basic thinking, I think your on the right track. The organizational shift needs to be on the value that can be derived and the parameters necessary to ensure that it is done reliably, with sufficient flexibilty while remaining cost effective.
If someone has other aspects that I've left out, I'd like to hear more.

Posted by Jay Prestipino Friday, November 11, 2005 3:10 AM

Charlie,

There is another part that has to be worked out as well. Virtualization created by higher processing speeds (due to multiple processors) and greater storage capacity will allow for centralization of more corporate data. This will potentially reduce the need for servers in remote locations. The final part that can make this possible is better WAN capacity either through additional bandwidth (a potentially prohibitive cost), handling latency by better caching (still can cause capacity issues) or eliminating protocol chatter. We need more options at the WAN to have it perform like the LAN.