AFCOM has taken flak for being vendor-centric and has recruited users to discuss the value of the peer network. John Parker --who manages the Internet Operations Group at ESRI, which makes geographic information systems software -- sat down to discuss his management style and projects with SearchDataCenter.com.
Parker manages ESRI's engineering staff, change management and works to ensure high availability for the company's online applications.
Parker also managed data centers in finance, insurance and healthcare industries, and did a stint as a consultant at Perot Systems.The large IT services consulting space has undergone consolidation. What has your experience been? Do you expect companies to contract large consulting services from companies such as Perot Systems (now Dell), EDS (now Hewlett-Packard) or IBM in the future?
John Parker: Consulting was a lot of fun, I love challenges and it was very challenging to go into a company and change the way they run a data center. You have to change their philosophy and thinking. I've done this wherever I worked.
I think large consulting engagements will be around in the future because more and more companies are focusing on their core initiatives and core competencies. That's why I'm working at ESRI. Their core competency is software development. With the Internet and SaaS [Software as a Service], we need to be strong in that market to continue to grow. Also, as more companies bring IT services back from offshore, companies are going to need to help from managed services.ESRI has been a Sun Microsystems customer in the past. What are your plans now that Oracle owns Sun?
J.P.: We're transitioning away from Unix. I won't say we're going off Unix altogether, but we're recommending Windows over Unix. That's where our expertise is growing. If you've got a hodgepodge of hardware and software, it's hard to create experts in all of those areas.
We have our tier one and tier two operations support teams offshore. We have people specifically trained in Unix and Windows, and we're trying to cross-train. Over the last five years, it's gotten harder to get a Unix admin to learn Windows and it's next to impossible to get a Windows admin to learn Unix.What are the big management challenges at your current data center?
J.P.: I always call them opportunities. Challenges give you headaches. I try to make it fun as opposed to being task-oriented. One of the opportunities we have here is high availability. With more companies offering services via the Internet, you need to have three nines or better of availability.
When you have human intervention and machinery, things are going to fail. I try to set documented, structured processes in place so when these events occur we can react quickly and resolve them as fast as we can. I don't see any challenges with high availability; it's just a matter of what you're allowed to buy. You need redundancy of N+1. Will your company allow you to purchase the equipment to [do] that?
Some of the best availability I had was working for the banks, and it was on the mainframe -- extremely reliable. Expensive but reliable.
Another opportunity is the new government regulations around the carbon footprint for data centers. I expect there will be a carbon regulation law in 2011, and in California we have Energy Efficiency Standard Title 24. Our local utility, Southern California Edison, has $1.3 million available for green initiatives. We found this out through our local AFCOM chapter. I just finished a data center audit with a consultant that Cal Ed paid for, telling us what to do better as a facility, how to be a more environmentally friendly company.What are you doing to make your data center more efficient?
J.P.: For data center power efficiency, you have to get rid of antiquated equipment. You have to use hot/cold-aisle design. The CRAC [computer air conditioning] units in our data center are old, we want to replace them, and put in variable-frequency drives on the CRACs.
I want to make sure that whatever we do around greening meets Title 24. The worst thing I could do is take these initiatives and have to redo them in a couple years.
This will also be the first time we've looked at power usage effectiveness (PUE) seriously. Depending who you talk to, PUE isn't the best thing since sliced bread, but it's the best way we have to measure energy use. It's going to be interesting. I know our PUE is going to great. We have a lot of work to do.You mentioned new servers as a way to save energy. What kind of projects do you have going on in that area?
J.P.: We're doing a server consolidation into a blade farm. We've got old servers; they don't have to be high availability, but they need to be faster and have more power. Blades have been around for a few years, they're safe and you know what you're getting into. When blades were new technology, [they] produced a lot of heat in a really small space. Now blade servers are faster, more efficient, save space. They use more energy, but when you consolidate, you save. We'll save 42% on our power bill from the 200 servers we are going to consolidate into this blade farm. It just makes sense.
What did you think of this feature? Write to SearchDataCenter.com's Matt Stansberry about your data center concerns at email@example.com.
Dig deeper on x86 commodity rackmount servers