Half of All Data Centers Understaffed

alphadogg writes "Fifty percent of IT executives say their data centers are understaffed, and companies are still looking for more ways to cut costs, according to Symantec's latest 'State of the Data Center' report. Sixteen percent of survey respondents said their data centers are extremely understaffed, and another 34% called their data centers somewhat understaffed. At the same time, data centers are becoming more complex and harder to manage, with more applications, data and increasingly demanding service-level agreements. 'Data center complexity has led to a lot of staffing challenges,' says Sean Derrington, director of storage management and high availability at Symantec."

Comments

Re:Should this be surprising?

[vlm]

Ahh I was somehow under the false impression that they were able to make cheaper cars due to lower wages, less environmental regulations, and the lack of labor unions.

Actually it only takes about $2K of labor to build all cars and trucks. Some robot factories cost less, some cost more.

Most of the revenue goes to executive bonuses.

I'll buy American made, Japanese managed, cars. But I won't buy Mexico made, American managed, cars.

Don't forget Western Europe

[brucmack]

I work for a Swedish company that understands the value of IT and invests resources in it accordingly. Based on my experiences with other Western European countries, this isn't abnormal.

The difference in work culture between here and the US is astounding. While it seems most American companies see IT as the place to save costs, the companies I've dealt with here recognize that our IT systems contribute directly to our competitiveness in the global market, and invest accordingly.

Re:Th e other half

[Penguinisto]

Laugh all you want, but there's a kernel of truth in that. All the *nix servers in my care mostly run on autopilot, and I pop in only once in awhile to check up on them, change/enter something in BIND, occasionally patch the ESX machinery, or put in the occasional patch that yum or ports can't get out of a repo (e.g. our custom help desk site software).

OTOH, a huge chunk of time is spent in Exchange and SharePoint - mostly chasing down errant mails, or fixing bugs and glitches. To be fair, those two bits are customer-facing, thus more open to calls - but even still, so is our help desk site (which runs on Linux), and I rarely have to bother with that on the back-end. Also, I've run pure *nix email setups before, and it never ate as much time percentage-wise as Exchange does now - even when chasing bounces.

On average, the 'doze servers eat about 95% of my time, but they comprise only 60% of the population.

Nota Bene: One thing I've found to be awesome - get up a script that sends a copy of your Exchange logs to another box... that way you're not fighting store.exe for RAM when you want to parse through them, and you can use a real text editor (vi or EMACS - you pick) to read them.

Re:Th e other half

[ka8zrt]

-1 or more about not thinking this through though. (and not funny at all)

As someone who has until recently done research in data centers and their operations, and personally dealt with the *NIX side of *NIX vs. NT years ago, I know the reality as opposed to the half-thought-out dreams some have. Yes, *NIX makes it much simpler to manage a machine, and increase the (servers/admin) ratio, among others, but it is not a solution which scales to where one person can administer 10K servers. As you add servers and applications, that ratio will reach a limit where you have to add yet another admin (operational, network, hardware, etc.). And should that site not be willing to do so, you end up with one of those "understaffed" data centers. Where that point is reached depends on a multitude of factors, including the behaviour of those using the data center (stupid developers, hands on users or workload characteristics cause that point to be reached sooner), the applications (a bunch of database servers will likely reach it before an equivalent amount of web servers), the amount of storage on those servers, and even the individual admins and how they are organized themselves. Throw in things like buying the cheapest hardware, or buying bleeding edge hardware (say 1.5TB drives when they first come out, or 10Gb ethernet cards), and it gets even worse as you try to deal with first generation drives failing or buggy drivers.

Can two people administer 500+ servers with 1.5PB of storage? I know personally that it is possible. But to do it and keep everyone 100% happy? No. And that precludes things like having people who are hard to satisfy, having to backup all that data, running it in a non-university production environment, etc. When I left CompuServe in 1997, the numbers were far different, with IIRC 25-30 operators of varying skill levels, about 10 of us in admin positions (who were called upon by the operators when they could not handle something), and around half a dozen or so network and hardware folks. Total number of servers? Around 1200 running BSD/OS, and around another 1000 running either our proprietary OS on systems which came out of the DecSystem 20 designs, or systems running a specialized NT 3.51 load, and perhaps a total data storage of around 1.5TB. And things were simplified by things such as having dozens of machines which were identical handling application X. Of course, we also had 3 data centers, and did backups of at least one of each machine in a given group. And then there is the fact that some applications required the developers to administer the application itself.

And looking forward... There were no regular 12 hour shifts at either of these. Yes, I was on call darn near 24*365 (I got vacation time off at my latest employer, but at CSI, I was on call even during vacation, and averaged 80hrs/week at the end). But when the fecal material hit the fan, and we had unusual problems like a computer room flooding or a critical server failing... it was possible to have to put in a 24 hour shift. Such is the life of a senior systems engineer in an operations group, which is one reason I try to avoid positions like these.