Where have all the DBAs Gone?
by Claus Mikkelsen on May 27, 2011
First, let’s get rid of some old business.
In some interesting timing, my last blog post questioned the future of Moore’s Law. The date of my post was April 26th. On May 5th, Intel announced the 22nm 3D tri-gate transistor, which will be available in their upcoming Ivy Bridge chip in the second half of this year.
Was I right? Yes.
Am I gloating? Perhaps a bit.
Was I lucky? Totally, but it’s a feeling I’ve had for a while. Thank you, Intel!! Read about this Intel technology and judge for yourself, but the days of the traditional planar transistors are over and it’s time to move on. It’s really quite a breakthrough, but it remains to be seen what the long term impact will be.
But I want to talk about another entirely different topic, and that’s the database administer (DBA). Full disclosure: one of my many previous “jobs” was a DBA. I learned a few things like how hard of a job it is, and I certainly learned to appreciate their work. I also learned that “getting it wrong” is what fuels DBAs’ nightmares.
I know. I got it wrong a couple of times, which is why I moved on to my next job: sweeping the offices of the remaining and competent DBAs. Incompetency can actually build a great and long resume!
Anyway, the obvious best approach was to “follow the rules”, something I was obviously not good at. The “rules” are what we now call best practices. That way, if you did get it wrong, at least you could claim it wasn’t your fault!
Many years later, when developing synchronous and asynchronous remote replication, I again got to work closely with some of the best database guys on the planet on the subject of data integrity with databases. I did enjoy it and learned a ton about storage performance (and data integrity). But I’m about to take issue with the profession. Let me explain.
When I was a DBA, we worked entirely on the application side and not with any storage folks. But this was still in the era of JBOD and “dumb” storage, so that made sense. But it got me recently thinking, and since my current job has me traveling more often than not, and spending a lot of time with customers, I began to wonder what has changed.
The answer, unfortunately, is not much. For about a year now, as I walk into a room, I quickly ask if anyone in the room is a DBA, married to a DBA, ever met a DBA, or even know where the DBA’s are hiding within their company. The answer generally is consistent with what I saw decades ago, that DBAs hide somewhere else, and are not part of the normal storage team.
Is this right? No, I think things need to change.
In the past month alone I have spent 2 full weeks on the road participating in what we call the “traveling EBC” or “EBC on the road”. That is, we fly some of our top techie folks and executives to where the customers live as opposed to asking them to fly to Santa Clara.
Three weeks ago, I spent a week in our offices in Hanover, MD outside of Baltimore and last week I was in Atlanta. In Baltimore I met with 19 customers; in Atlanta I met with 13. That means I asked my DBA question 32 times mostly with a resounding “no way”. But in Baltimore, one CIO said they recently reorganized and the DBAs and storage guys are now in the same group. Hmmm. In Atlanta, I met with a customer and there were 2 DBAs IN THE ROOM! I was speechless (for a bit), then got into a great discussion on storage and databases and performance. These two guys got me deciding to write this post (and I know you’re reading this, you know who you are, so thank you, and comment if you wish).
But here is my point. Storage is no longer JBOD and is no longer dumb. We’re living a world of storage computers that automate many tasks, including many done by DBAs. I’ve blogged on this in that past, and will challenge any DBA to provision a database that can outperform what we can do with Hitachi Dynamic Provisioning (HDP). You may get close (at great expense), but you will not exceed. Once I explain the magic behind HDP to a DBA, they immediately “get it” and agree.
Throw Hitachi Dynamic Tiering (HDT) into the mix and not only do you get the performance and throughput boost, you get the benefits of moving less accessed data to lower tiers of storage. Anyone disagree that (generally) large parts of databases haven’t been accessed months or longer? Demote those parts in 42MB pages to less expensive storage (or leave it all on Tier 1, if you wish).
When I speak to the pure storage guys, they get this well, which is why HDP and HDT adoption is very high. Now it’s time for the DBAs to get educated on this new (well, not that new!) technology and jump on board. Seriously, it’ll eliminate some of those nightmares.
And what better way to transfer the knowledge than to start working more closely together. All will benefit.
So here’s my plea: DBAs, wake up to the new storage technology available and Storage Admins, how about helping your fellows out. Like I said, all will benefit.
The pleasure was our Claus. I learned a great deal from you during our short time together. Let me know if you find a DBA willing to take on your challenge. I would like to see the results of their testing. How many man hours they invest and how poorly they compare with HDT.