As Linux establishes itself as a mainstream operating system, and open source tools and applications prove their enterprise readiness, a growing number of organizations are talking publicly about their open source deployments and direction. Recently, Mok Choe, CTO at Union Bank in Monterrey Park, Calif., spoke with Network World Senior Editor Jennifer Mears about the financial institution’s decision to scrap proprietary Unix systems for commodity servers running Red Hat Linux. Here is an edited transcript of their conversation:
|
Give me some background on your move to Linux and open source.
We were faced with having to do an upgrade of our Web infrastructure last year, which gave us an opportunity to really look at our architectural direction. We decided to head in the direction of horizontal scaling using commodity hardware and open source tools. There are multiple drivers behind this thing. Cost is clearly one of them. First and foremost always is the reliability and performance. The second thing is we wanted to be able to scale much easier than we do today. Since we run on big boxes, at some point to scale you have to buy another big box. Instead of that, we wanted to be able to rapidly manage capacity by adding or subtracting commodity hardware.
Had you been using any open source software previously?
On a smaller scale, a Linux deployment did exist, but in very small, departmental solutions.
How did you decide that it was time to look at a broader deployment of Linux and open source?
That was part of the overall architectural direction that we decided upon for our Web applications. From the enterprise perspective, it was a big decision. Going with the commodity hardware was a big decision. Going with open source was a big decision.
So you determined that open source had reached a point that you felt comfortable bringing it into larger production environments?
Yes, we have a number of people who have had experience with Linux, including myself in my previous job, so we were rather comfortable going with that decision. We did make sure that we are meeting all the compliance requirements, so both internal and external regulatory bodies were consulted.
How did you go about that?
Since we are going the open source route, everything must be open, so to speak. All the documentation that’s required has been satisfied, and we have an internal arm auditing our use of open source tools.
You were running AIX servers, WebSphere and Oracle databases. Was there reluctance among the bank’s management on the business side to move away from proprietary best-of-breed?
I wouldn’t call it reluctance. But we did do a lot of education upfront. We started really small and proved the maturity of the technology, as well as covering all the processes from the enterprise project management perspective, as well as from the risk management perspective.
A year or two ago I would think the move to open source would have been a harder sell.
Yes, but now, even in a highly regulated space like banking, there are a number of solutions that take advantage of open source tools. As a matter of fact, let’s say you go to a contract on IBM’s WebSphere, you will actually find open source components in there. That’s a fact.
Do you know how much you’ll save once you make the transition?
Yes, we did a complete multiyear ROI analysis, but I don’t want to give specific numbers. Any savings has to do with our overall approach, so we didn’t separate Linux over AIX, for example, [and say that] is giving us this much savings. It was our existing architectural approach vs. what we’re building.
Previously, your architecture was centered on big SMP [symmetric multi-processing] systems?
Yes, we use systems running on RISC-based processors. But we decided to put together a plan and we did an ROI analysis across all of the infrastructure, as well as the applications, in our Web space — from the server to the network to security to load balancing and to the applications themselves. What we’re doing now is beginning to move in the direction of our new architectural road map. We are currently tackling the server infrastructure, the network infrastructure and the security infrastructure. Linux is part of that initiative.
How did you launch into the migration?
As I said, we were faced with the upgrade of our Web infrastructure and we decided that instead of buying the next version of the same thing to look at our architecture. We did a quick pilot. We did have an internal system that was built on a completely LAMP [Linux, Apache, MySQL, PHP/Perl/Python] stack of solutions. So we took two smaller systems and migrated them to our new architecture successfully in September. And then we built the case for the overall architectural migration.
At the same time, we took a brand-new project and built it on the commodity hardware, Linux, JBoss, MySQL set of tools. That project, which is about to go into production, is one of the bigger systems any bank that deals with check images has to have. It has major performance and throughput requirements.
What kind of load does that system handle?
We process millions and millions of checks every night. We capture data, as well as images of those items. So we’re talking a large system here.
And you saw good results?
There are always minor issues, but if you look at the pilot we did, it was a tremendous success. It went off without any problems at all. It generated real excitement in the technology community here, as well as the business side. So there is a lot of momentum behind it now. Had the pilot failed, we would have been in a different situation. Its success was in its ability to scale, and we saw the performance increase.
In September you did the smaller pilot projects, which were a success, and then what?
We put together the overall proposal and then got approved. We’re now hot and heavy into the first phase of this project, which is actually building the network. That involves building a separate network from what exists today with more enhanced load-balancing capability and security, and that’s where we will build the servers and then migrate the applications on top of that.
Is this for your entire Web site?
Yes, mostly customer facing, as well as some internal applications. If you can imagine all of the Web-based solutions that a bank provides today.
So today you’re still running most of the Web site on the AIX systems but you’re slowly migrating them over?
Right. Like any other project you have a set of requirements to build on an existing platform. It’s not stopping and waiting for the migration, so we need to coordinate between that and this project.
As far as what’s being migrated from AIX, can you give me an idea of how many AIX machines you have and how many commodity servers you’re migrating to?
We’re increasing in terms of sheer number of servers by a lot. But that’s an excellent question because as we go down this approach there are some disadvantages: No. 1 is the power requirement; No. 2 is the number of network ports you have to have; No. 3 is the management perspective. So we’re doing a lot in terms of automating provisioning.
At the same time, though, you’re expecting significant savings?
I think we’re going to generate some savings, which we’re going to reinvest to build our next set of solutions. But that’s not necessarily the sole driver of this thing. We do feel we’re going to be much more reliable and quicker in terms of showing responsiveness, as well as being able to launch the next set of products more quickly with our new architecture.
It’s not really cost savings, then, but flexibility you were looking for?
Right — reliability, performance, flexibility, scalability, as well as doing all of that in a cost-effective manner. Those are all drivers of this project.
Why did you settle on Red Hat vs. the other Linux distributions, such as SUSE?
There were several things: We do have folks who have some experience with [Red Hat] and we liked the solution stack that Red Hat provides from Linux to JBoss to MySQL. We plan to have a number of solutions eventually built on that stack. The third thing is these guys have a strong relationship with the hardware vendor we picked.
Have x86 advances in performance and capabilities enabled you to look at this new approach?
Absolutely, if you look at the price/performance of an Intel or AMD chip today and compare that to a RISC processor, there is an enormous difference. We want to do things in a cost-effective manner to reduce the cost per transaction.
Getting back to the open source angle, where are you in the stage of using JBoss and MySQL?
The overall initiative for JBoss will be following this particular initiative, but we have done the pilot phase.
So you know you’re heading in that direction?
For any J2EE application moving forward, yes.
What about the database?
We are a huge Oracle shop. We don’t have any plans to do a wholesale migration, but whenever the opportunities arise we will look at the MySQL-based solution.
Any advice for your peers who still may be a bit reluctant to move to open source?
I don’t know if I would ever look at it as just a Red Hat initiative. I think organizations should have an overall game plan that includes open source components and try to build that way. Linux is just a piece of the overall puzzle. After all, it’s the operating system — we do a whole lot more than operating systems.
As far as your plans go, what other open source tools are you looking at?
We talked about some big ones already. We, at the same time, are looking at smaller-scale tools, maybe engines, whether it’s a workflow engine or a documentation engine. We are always looking at the viability of using some of those things. That’s a little more out on a limb than going with Red Hat Linux, so that will take a little longer time. We do have a working prototype of it, though, but it’s not something we’re going to put into production.
Do you find open source coming up more often when you’re making software decisions?
There is clearly a buzzing in our technology shop. The folks that have been pushing the idea all along now have a following.