Monday, July 8, 2013

Reflections on simplify, standardize, automate

Several years ago, I was the Senior Manager for Operations & Infrastructure in the Office of Information Technology, at the University of Minnesota Twin Cities campus. One tenet I always advocated was simplify, standardize, automate. The goal was to make our jobs easier, to free up our time from mundane or repetitive tasks, so we could give more attention and energy to working on more important issues. We used this to guide our future plans, to bring about transformational change in our IT practices.

One project that came from simplify, standardize, automate was the central web hosting system. Most of our web customers simply needed a place to put their web content, and didn't care (very much) about "web programming" such as Java, Perl, or PHP. Rather than continue instantiating a bunch of one-off web servers just to support this simple content, we set up a single robust web hosting service that could accommodate most (say, 90%) of our web hosting requests. This reduced the number of servers that we needed to support, which simplified the environment, and allowed us to standardize our web server administration. Over time, we further automated the process to create new websites for our customers. It was a large "win" that continues to benefit the university today. At Morris, we leverage the central web hosting service for several of our important websites, including some research websites.

Another challenge was database hosting. The university runs lots of databases (most of them are Oracle) and like most institutions, most databases require their own server. We had a tough time managing the server sprawl, as each new database usually meant a new server. Fortunately, the database administrators on my team, led by infrastructure manager Patton Fast, discovered a new way to support our growing Oracle database needs: Exadata. Although implementation happened after I'd left OIT to join the Morris campus, I was very glad to see this project kick off.

Prior to this, I'd long encouraged our Oracle sales engineers to develop an "Oracle/OS" distribution. This might come on a DVD (in several formats: say, one for Sun SPARC and another for Intel) that a database administrator installs onto an empty server. The server is automatically configured to run the database, and after installation, the database administrator doesn't need to worry about the operating system. The only difficulty would be managing enterprise storage, such as an EMC SAN.

After Sun purchased a storage company, and after Oracle purchased Sun, Oracle finally released such a product. To oversimplify, Exadata is essentially a "database in a can." The server and accompanying storage is self-contained in a frame. The underlying operating system is effectively "embedded," so database administrators don't have to manage the operating system. For all practical purposes, it is a database "appliance" and has saved the university in both time and budget. OIT's Andy Wattenhofer gave a presentation about U of M and Exadata at last year's EDUCAUSE Midwest regional conference.

I'd almost forgotten about the Exadata until I saw an update from the Enterprise Systems Upgrade Program a few weeks ago: ESUP Update, June 3-16.


ESUP Assistant Program Director Mark Powell was on hand last week to receive delivery of two new sets of servers: Exadata and Exalogic. The first set, Exadata, will house the Oracle databases. When OIT upgraded to the current version of Exadata servers, users experienced significant performance improvement.
The second set of servers, Exalogic, are new for ESUP. These servers will run PeopleSoft applications instead of the multiple servers currently used. The Exalogic and Exadata servers are designed to work together and Oracle will provide additional patching and upgrade services for these servers, which helps reduce system administration resources and costs to the University.
The servers will be split into two combined sets, pairing one Exalogic with one Exadata. The set in the WBOB data center will support ESUP’s production systems, while the second set (located in the 90 Church St. data center), will support the Program’s development, testing, and disaster recovery.

I'm glad to see this project continue. Way to go, OIT!

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.