Two years ago, Anthony Hill found an IT Tower of Babel in the data center of San Francisco-based Golden Gate University....
The technology architecture -- which served 1,000-plus desktops, 6,000-plus staff and student users and 50,000-plus directory identities -- contained almost every software platform and hardware system extant. "Everything except a mainframe," Hill said. As the institution's new chief technology officer, he was charged with creating order in this chaos and restructuring the data center to facilitate GGU's goal of operating as efficiently an e-business.
Linux is playing a key role in GGU's move from hetero to homogeneity. In this two-part interview, Hill describes the process of making platform and application decisions and the pleasant and unpleasant surprises he's encountered during the implementation. In this installment, he explains how GGU created an IT jumble, why Linux won him over and how he persuaded GGU to use Linux. In part two, he describes his experiences with Oracle 9i and Oracle 11i and explains why he's interested in Linux on the desktop.
How did GGU end up with so many different systems?
Hill: GGU had experienced the typical ad hoc database and hardware and operating system proliferation over the last 10 years. Most universities had a decentralized and somewhat democratic IT process then, and that results in a tremendous amount of redundancy and waste.
What I faced was the result of a decade of best-of-breed purchasing without any overarching technology strategy. We needed to transform the entire technology architecture in order to be able to function as an e-business.
Could you describe the system mishmash you were faced with two years ago?
Hill: For every core platform we had, we had a different operating system, a different database and a different hardware platform. Each application was running on a completely different environment.
We were running Windows, Solaris, MPE on the HP3000, and the Macintosh operating system in production in the data center. We had Microsoft Access and SQL Server as the primary database and a UniData database (now an IBM product, but a fairly archaic one). We also supported Novell as the LAN operating system. Not much was left out. Oh yeah, there was no mainframe. We had HP3000, but no big mainframes.
How well did these IT fiefdoms work together?
Hill: Unfortunately, there was no integration. So, HR, financials and our student administration system were on different platforms at zero interoperability and data sharing. Now, in a university, the student system is analogous to a commercial business' customer database and the processing inventory system.
So, how did you choose what to use and what to scrap?
Hill: Our enterprise was so heterogeneous two years ago that we could have gone in just about any direction. I narrowed it down to two directions: Microsoft and .NET, when .NET was just emerging; or J2EE (Java 2 Enterprise Edition) and Oracle 9i.
Looking at the technology needs of a higher education institution, I felt that the J2EE and Oracle platform provided the broadest destination for all of our systems. If we had chosen the Microsoft Windows, SQL Server and .NET path, for example, we would have had about 50% fewer ERP vendors to select from. All the major vendors supported Oracle, and J2EE was the preferred application architecture for all of the Web applications that we wanted to build.
Also, choosing Microsoft would have meant the application architecture would revolve around Visual Basic, SQL Server and the Microsoft technology stack. Two years ago, when we were making decisions, the Microsoft technology stack direction was very unclear. .NET was the heavy buzzword, but it wasn't being delivered. We could have done J2EE and Oracle on Windows, but it makes a lot more sense to use J2EE and Oracle on Unix. So we made the Unix decision.
What flavor of Unix did you favor?
Hill: Initially, we were leaning toward Solaris. We had some experience with Solaris in-house. All of the prior large e-business development efforts that I'd been involved in and witnessed were based around Solaris.
With Solaris, however, our big challenge was cost. Implementing Oracle, new ERP and a whole new data center with Sun servers and Solaris is a very expensive proposition.
Hill: Yes. So we began looking at Linux. But, despite its cost advantages, it had its downside. Two years ago, Linux hadn't really been adopted as the standard it is today. As a CTO, I didn't feel comfortable recommending Linux at that time.
What changed to make you feel more comfortable with Linux?
Hill: Starting two years ago, every month bigger and bigger announcements were coming out relative to Linux and vendor support of Linux. More and more, Linux seemed like a viable operating system in the data center. Then, Oracle announced a partnership with Red Hat and Dell to deliver Oracle products on Linux and to support those products with an integrated help desk; I began to see Linux as the way to go.
How did you persuade your executive team to go with Linux?
Hill: I had to prove the business case for Linux to the business manager, our COO and our president.
I described the cost model for a Linux deployment at that time. I did a very detailed five-year, total-cost-of-ownership analysis on Linux running on generic hardware. I compared that to Solaris on Sun hardware. There was a very significant cost of acquisition difference at that time. That gap is narrowing now, but there was a very significant difference then.
Then I made the case that there is industry support for Linux. When Dell and Oracle supported Linux very visibly, it took a lot of the risk out for the business managers.
After that, I discussed hardware vendor lock-in scenarios that you can't avoid with Unix. I explained that Linux gives us some leeway to change our decision at any point in time. Let's say we're six months into a two-year server migration to Linux, and it turned out to be a bad decision. We can change without a huge hardware cost or [software] effort. With Unix, the hardware cost would be high, and many applications would run only on Unix. If we were in a Microsoft world, a switch threatens your database, application architecture, and packaged application decisions, because all run on Windows and probably not Unix.
Once I made my case, winning approval was easy.
Today, is it your plan to use Linux across your whole organization?
Hill: It is. My goal is to achieve the greatest economies of scale that we can relative to data center consolidation. So we'll put as much on Linux as possible, not because of the magic of Linux, but for the business goal of consolidation. For consolidation, Linux happens to be the platform of choice today, in my opinion.
FOR MORE INFORMATION:
FEEDBACK: What has to happen for Linux to become the dominant data center platform?
Send your feedback to the SearchEnterpriseLinux.com news team.