Wednesday, November 21, 2007

Green Computing: Environmental Concern or IT Crisis?

I've been buried in a number of projects lately, but as things slow down before the US Thanksgiving holiday, I've had a chance to peruse a number of bookmarks I've been tagging related to topics that I'm working on. And, I'm seeing some interesting trends linking many of these together which seems to be at the nexus for a number of new technologies and announcements. Looking at these patterns may help provide some focus understanding which of the emerging technologies has the greatest chances of having a significant impact on the IT market.

Of the trends, one of the most common right now is the constantly increasing price of oil, the global tension surrounding oil, its impacts on the global economy, and the ramifications on the overall cost of energy. Energy prices are obviously one factor leading to the nearly universal concern in data centers that energy prices are becoming a significant cost factor. Obviously, the rate of growth of data centers, the trend towards installing more cheap Intel based servers, the explosive growth of new workloads and better linkages between people and information are also rapidly consuming available power.

Related to this problem is of course the quantity of heat generated by all of these servers and the requirement that servers operate within a relatively narrow temperature range. This means that data center cooling is another source of power consumption. And, it means that air conditioning and cooling has to be fit into the same floor space as the servers. Since it raises the power consumption bill as well, it becomes a more frequently measured source of IT expenses. And, the dilemma of how to add more cooling and more compute power in the same, typically existing data centers, is causing many big IT organizations to build new IT data centers.
Of course, for those with some sort of conscience about the environmental factors at work here, this overall IT crisis is also becoming a Green concern. The amount of energy consumed, the amount of heat generated, the amount of often "prime" space being used for data centers is globally becoming enough to stagger the mind. And, the rate at which the new and innovative workloads is increasing is going to continue this trend in a superlinear fashion - I can't predict that it is an exponential growth because of containing parameters such as cost, floor space, the economy, power availability, human labor and related IT staffing, etc. But I can safely say from my personal experience that all signs point to something substantially greater than a linear increase in workload growth and data center sizes.

Another trend that seems to be prevalent in data centers is that, for many reasons, there has been a many decade trend towards heterogeneity, driven by many factors. Obviously, pricing/cost is a significant factor. Innovations happening at different rates in different companies also provides a value at some times beyond cost (those high value items that drive the margin behind any company with staying power in the industry). Personal preferences by the IT staff, relationships between company representatives and sales teams, interoperability of complementary products, etc. - the list goes on and on - all drive decisions which ultimately lead to the deployment of highly heterogeneous environments. In many ways, that heterogeneity of the data center drives a Darwinistic survival of the best/fittest environments, which can easily be viewed as a great thing.

However, this heterogeneity in the data center is also driving increasing complexity, increasing costs for interoperability, and, ultimately increasing IT management costs. In some ways, this last point might be one of the most expensive for most data centers. The cost of each person hired to manage data centers often exceeds the cost of the acquisition of those machines, and over time becomes one of the dominant expenses. While Systems Administrators might rebel at my line of thinking here - basically that paying people to administer systems is becoming a disproportionate cost of doing business - the reality is that this spending on administrating the infrastructure is taking money away from a company's ability to invest in its own unique business value.

In summary, I believe that we are approaching an IT crisis here, which happens to align with the increasingly popular concerns with the environment. Those points of crisis are centered around data center issues of power consumption and availability, the increasing need for data center cooling, the unconstrained growth of data centers, and the uncontrolled costs of IT management. While there are many solutions attempting to address components of these problems, such as the previously reported on, or new (new again?) chilled water cooling doors for racks, improved air flow or power efficiency in a compute cabinate, smarter data center layouts, improved systems management tooling, etc., I don't think any of these changes will be enough to resolve the superlinear problem of the growing IT crisis. However, I do think there are a few things on the horizon that have a chance of making a more significant difference, and hopefully over the holidays I'll have time to put up another entry that goes into some of the changes I believe are necessary to address this IT crisis.


Friday, November 16, 2007

Ottawa Linux Symposium 2008: Call For Papers

The Call for Papers for OLS, or, better to be known as "The Linux Symposium" since this will be the last year it is held in Ottawa, came out yesterday. This year the committee is going to go all out for the 10th anniversary of the conference and expects to have multiple keynote speakers in addition to the normal BOFs, presentations and tuturials. And, the facilitators will provide space for mini-summits before the actual conference. And, the program committee is working to expand the topics to include additional areas of innovation around Linux. These might include creative uses of virtualization, as opposed to strictly the implementation of virtualization, mechanisms for improved power management, overall improvements to desktop or handheld usability, or new technologies that are emerging in the Linux ecosystem.

The Call For Papers is online:

The 2008 Linux Symposium should be the best symposium yet! Make sure to submit your papers early and register early as prices rise closer to the actual conference.