I am just finishing up the third day of my attendance at the 35th anniversary edition of the AFCOM conference, now affectionately known as Data Center World (which they combined with a dozen or so years ago). The conference has grown to 5 days in duration, over fifteen hundred attendees, and dozens of sessions were conducted. Along with the conference sessions agenda, more than 200 vendors were represented on the expo floor. Add to the mix all of the Las Vegas sunshine, swimming pools, great restaurants and a few dozen hospitality hours (not to mention the gambling) and you have an event that will be remembered by most. (And perhaps hopefully forgotten by a few, but that is another story.)
So, what were the session/conference highlights? What were people talking about on stage and in the hallways? Lots of things come to mind, presented below in no particular order.
1. Data Center In-house, Co-Lo, Modular, Cloud? Many discussions about what an end-user’s next computing platform will look like. Lots of discussions about the transitions. Co-Lo or in-house? Some Container or Modular? Retrofit or new construction? Where to build? What about the Cloud? Lots of opinions here. The folks at consulting company Plannet seemed to make a great case that if you have chosen to stand-up a new data center, you’ll have a second choice to make: retrofit available space versus new construction, and they used a figure of 5Mw as the threshold. Above that level they argued, NEW construction is usually better fiscally since much of the older available space is limited by density considerations. Modular seemed to be part of the discussions when smaller chunks of computing were needed in lots of distributed locations. And related, I also heard lots of discussions wondering about how quickly CLOUD will take over everything, and the general consensus was it would be in the far distant future, long after anyone at this show cares about Data Centers (a really long time in the future). Universally, everyone also agreed that Co-Location was a good thing, already part of their infrastructure today, and will be a growing part of their future… even if other strategies (mentioned above) were also increasing!
2. The DCIM Mantra. Everyone one was saying it. Every vendor said they had it, and every attendee wanted “whatever *IT* was”. So what was the “it” that they wanted? Responses varied widely as to what DCIM really was. I had no less than TWO DOZEN discussions with end-users at the show and each had a different view of what DCIM meant to them and to their peers and what they were actually looking for. The common theme however was, “I AM NOT SURE WHAT DCIM IS, BUT I KNOW IT IS MORE THAN POWER MONITORING!” Most agreed, senior management is now REQUIRING that the delivery of their IT business services MUST be more agile, more predictable, more defendable, more efficient, and more transparent…. and most attendees agreed that they are sorely ill-prepared to meet those challenges. (The word “embarrassed” would be a good fit in this sentence). Apparently the jig is up. The “common executives” in F1000 companies really DO want to know a great deal more about how IT does it’s magic. Where all of that money goes. Whereas for 30 years “Don’t ask, don’t tell” seemed to be fine for everything IT, now every executive wants to understand WHY and HOW. Vendors on the show floor were showing products that cooled better, software that tracked and managed resources and workflows better, and multiple sessions were presented about the various values of DCIM in production today. In one panel, a Fortune 500 end user with a larger instance DCIM already deployed spent a great deal of time talking about his results. He was very definitive that the BIG VALUE OF DCIM TO HIS COMPANY was being their ability to visualize and then streamline their operational processes and workflows. Consistency was the icing on his cake. Good to see that his experience shows what we’ve said all along, process optimization will yield more savings over time than that which will be seen by managing power and cooling alone.
3. The price of power has stabilized. Ten years ago the cost of power was the hottest topic in the Data Center industry. At this year’s conference I didn’t hear a single session or attendee voicing their concern about the price of power. Back in 2007, it seemed like the sky was falling. The Jonathan Koomey EPA report suggested that data centers were going to suck the life out of the power grid and power was going to just keep getting more expensive. We were braced for 7% compounded increases per year and could expect 20 to 25-cent kilowatts “soon”. Everyone began to look at renewal power sources and signed crazy commitment contracts to protect their future. That madness didn’t happen. While growth in IT still played out rapidly, efficient data center designed dulled the impact dramatically. Most of us still enjoy 10-cent kilowatts and two guys from North Carolina and one of the Dakotas told me at the reception yesterday that they are paying just 2.5-cents and 4.5-cents per Kilowatt (respectively)! This fundamental electricity cost stabilization changed the focus at this conference to that of the more interesting topics of growth and expansion, capacity planning and risk/reliability architectures. More than one session included discussions about choosing the right power redundancy and resiliency plan (including Tier selection) to match the specific business task at hand, rather than take a one-size-fits-all approach. They argued that putting all business services in the same Tier-3 data center for example was fiscally irresponsible, and showed a lack of business alignment.
4. Facilities, meet IT. It’s no surprise that at some point the facilities professionals who regularly attend AFCOM and similar shows were going to ask for a better understanding about what is actually happening on the white space (raised floor, production floor, etc) that they so diligently provide. Sure they see IT guys installing racks and rows, and servers and switches, but historically Facilities professionals don’t really understand how IT actually delivers computing services. In other words, they don’t understand the actual relationship of a circuit breaker to a business service. They know how much power and cooling the stuff on the floor sucks up and how much heat it generates, but not what it actually does with those consumed resources. There was a session presented by HP and Methode on Sunday morning entitled “IT and Facilities 101″ which explained applications, servers and virtualized operating systems and everything in between. Good starting point!
5. Finding Stranded Capacity. One of the most impressive ‘case study’ sessions was presented by a data center engineer at BOX, the Cloud based storage and services provider. Apparently BOX runs their namesake services on hundreds of racks spread across three co-location facilities. Through the recent deployment of new granular resource consumption monitoring and modeling, they were able to TRIPLE their service capacity WITHOUT adding a single rack! Most important, they can now easily place new services in the most appropriate locations, continuous conserving resources, optimizing their footprint. This was an eye opener for the audience attendees that knew stranded resources were an issue, but had no idea of the magnitude of the problem. (Feels a lot like some of the values seen with “DCIM”, right?)
So what about the 200 vendors? Weren’t there some new toys on the floor? Well, if I took last year’s show, and moved the calendar forward to today, each vendor had what I would expect to see, incremental progress. Rack PDUs were higher capacity and had smaller footprints. Cable trays were more adjustable and lighter in weight. More variants of racks were seen, each supporting larger weight/loads for higher density equipment. There were new blanking panels with active temperature indication. And a whole slew of air management offerings with higher capacity and more energy efficient fans. Etc., etc. Don’t get me wrong, this is all cool stuff, but it was expected stuff.
Lastly, congrats to Mike Cunningham, Director of Data Center, University of Texas @ Austin for winning AFCOM’s Data Center Manager of the Year. His background with DELL and IBM was all about process engineering, so it’s great that he is recognized for this key to all of our futures- significant efforts regarding process management.