Willy Chiu, vice president of the high-volume Web sites organization within IBM Software, talked with Computerworld's Matt Hamblen about working with large customers to develop grid computing, or on-demand systems of servers for corporate needs. For example, IBM has announced that it's working with Federated Department Stores Inc. to test a new application called FedAd to make its advertising more efficient across the retailer's divisions, such as Macy's and Bloomingdale's.
Chiu went on to describe how grid computing initiatives are now moving from setting up grids for Web and application servers to general business uses such as database servers. Here are excerpts from that interview:
Q: What is your organization trying to do?
We're part of the IBM Software Group that reports to Steve Mills, with locations on the West Coast and New York, Japan and England. We're working with cutting-edge customers, solving problems in [the] area of scalability and have been doing that for a number of years. EBay is a good example, since they are running 700 million page views a day on thousands of servers, with WebSphere application server as the middleware, and they are incredibly successful. Over the last few months, we have been putting together solutions in the on-demand space. In December, we unveiled a Center of Competency at the San Jose Valley High Volume Web Site Lab for on-demand development. We're populating it with customer research projects and working with leading-edge customers to realize the ROI that you hear about for e-business initiatives.
Q: Recently, Charles Schwab & Co. said it has worked with IBM to create a Linux-based grid computing system to speed up its compute-intensive investment management applications (see story). So are there are other examples of this kind of work?
We have over 100 customers in on-demand projects, and some are not publicized. They are customers that we work with in the financial space in banking, finance and in the area of insurance and areas of e-commerce. At the center, we have critical engineering skills, and we work with customers to help them identify areas of opportunity, if they have bursty traffic and their own IT infrastructure and would like to virtualize their servers to run different apps across virtual resources. As they have bursts of traffic, they take advantage of spare capacity across the servers.
We automate the allocation of servers and automate management of service-level agreement [SLA] objectives and monitor what's going on in terms of different workloads and the high-priority workloads when traffic increases. For example, at health care provider Cigna, before on-demand, they would have to reallocate and reconfigure servers manually for high-volume periods. Now, with WebSphere and this server allocation for WebSphere, they are able to manage their WebSphere clusters in a virtualized way by identifying the workload that they want to have adhere to service levels, whether the levels are throughput or response times. As traffic increases, if more capacity is needed, it is provided automatically.
Q: Describe IBM's investment in this on-demand research.
IBM is spending $10 billion over four years on on-demand, and we are establishing the San Jose center as one of several, using our experience and best practices and exporting that to other centers around the world. We want to take lessons learned and apply on-demand solutions and capture business benefits.
Q: What is the next step for the Center for Competency?
We are expanding to 1,000 servers from several hundred and will be putting 1,000 servers in the other sites.
Q: How does IBM compare to the work being done by Hewlett-Packard Co. and others with regard to grid computing and the larger on-demand area?
What we're doing is more than an architecture. We are delivering real value today with hundreds of customers, with the three companies, Federated, Cigna and Schwab, highlighted to show what the benefits and are and what the value [is]. They are early IBM customers who have purchased IBM hardware or software or services and are partners.
Q: What is the next big thing with on-demand and grid computing?
One thing coming next is we are expanding solutions to the whole computing infrastructure, all the way from communications to the network through the Web server and app servers, then to database servers. We will manage the whole SLA cycle and, in addition to automating provisioning, we will be able to manage servers from the iron up and load up the whole software stack. Think about the enormous implications of that. EBay would have had to go out and procure another 500 servers without their new tools, and that takes time and planning.
Q: So adding the database servers and general business functions to grids sounds like one of the next steps?
The challenge is to go to the database servers and add dynamic capabilities and add caching ability. That way, you have a whole end-to-end view then. All these steps are coming simultaneously. It's all in the works, and we'll be able [to provide] that database grid capability with customers in near term, probably this year.