Well, I’ve been pretty absent during these weeks, obviously I need to do some real work to put bread on my table, precisely I’ve been working on a big project for one of my favorite customers that marks a “copernican” step in his IT infrastructure, but let’s give some details (with names removed to protect the innocents πŸ™‚ ).

It’s a medium enterprise (At least by Italian standards), #1 in its market, with branch offices scattered in three different continents serving approximately 3000 users.

The Early Years.

The company itself was founded in the ’50s so they were a mainframe IBM customer during their first IT years (like almost every business in Italy at that time), the IT “architect” (even if he hates this monicker) at the time started to look at alternatives in the mainframe arena and they focused on Hitachi, they used to be Hitachi customers for both CPUs and External Disks for a long time (they’ve been Hitachi storage customer until recent years).

Exciting Times ahead.

In the late ’90s, with the UNIX powerhouse that was Sun, they decided to take the leap in the open system world and started a rehost of their CICS environment on the SPARC/Solaris platform.

Those were great years, Sun was the cream of the crop of the UNIX marketplace, they started using Sun for everything, from big servers to remote branches, from CAD workstations to FAX servers, definitely what I would call a great customer for Sun.

And then the oblivion.

Sun started to fall apart, the customer resisted, continued to purchase SPARC systems, in the meantime they also started a huge migration to Oracle E-Business Suite that needed more flexibility than their previous CICS (EBS instances proliferated due to development, QA and so on), this almost grinded to an halt the IT department: everything was simply too complex to manage, every task was time consuming, cloning an EBS instance literally took days. This was simply impossible to carry over.

The Revelation.

The customer was exasperated, he was so fond of Sun and their incredible engineering talent that they leveraged every single Solaris feature: from Containers to ZFS, from Jumpstart to Projects. The news of the Oracle takeover was really the straw that broke the camel’s back, the customer had enough.

He called me for a meeting to discuss the future, they already moved over to a next generation storage back in 2008 dissing HDS, so the next logical step was to migrate off SPARC to the land of the “x86 sh!t” (like the customer used to call it).

We settled for a two steps project:

  1. Add Solid State Drives to their current storage system in order to lower response times immediately.
  2. Acquire and create a new Oracle EBS production environment running on x86-64 and put it in production in a 3 months timeframe.

Right now the new system is humming away nicely, it’s a pretty decent system comprised of:

  • A dual-node RAC configuration with 32 Nehalem-EX cores and 256GB RAM each for the databases.
  • Five VMware ESXi nodes (12 cores/96GB RAM each) for the frontends and development systems.

As is common practice with this customer, we established a benchmark baseline using a specific job. This particular job ran during the weekends and used to clock at 46 hours when run on the old SPARC infrastructure (a top of the line machine with the latest SPARC processors), right now the same job, running on the x86 rig, clock in at 1 hour and 38 minutes. Impressive isn’t it?

Needless to say that the customer is enthusiastic, I used to design several transitions from SPARC to x86/VMware but this was definitely the project that left a different taste, this was a real revelation.