Teraplex Center Waiting to 'Break' V4R4
When V4R4 of OS/400 is released to the market next month, the AS/400 user base won't be the only group looking to sink its teeth into the new technology. IBM's Teraplex Integration Center in Rochester, Minn. is poised to push certain aspects of V4R4 to the breaking point as a way to ensure the product does not break in actual production environments.
The AS/400 Teraplex Center was created -- along with Poughkeepsie, N.Y.-based centers for RS/6000 and S/390 technology -- by IBM to help businesses stretch the limits of current data warehousing hardware and software, and design future computing configurations.
V4R4 represents a unique challenge for the AS/400 Teraplex Center thanks to new technology designed to use encoded vector indexes (EVIs), according to Amy Anderson, Teraplex Integration Center technical specialist. Originally introduced with V4R3, EVIs are an IBM-patented technology loosely referred to in the industry as "bitmap index."
"Most relational database vendors are trying to come out with some kind of bitmap index technology," Anderson says. "On most relational database management systems, there's one kind of index that's either called a b[inary] tree or a radix [b-tree radix index], and it's a tree-like structure.
"In V4R4, what we're looking forward to working on is a way to use EVIs to improve the performance of joins when you run an SQL query," Anderson continues. "[Traditionally] if you have two tables and you want some information from each table, the end user -- or the database system on the user's behalf -- has to tell the system how you want those two tables joined together. At a very high level, [EVIs] are just a different way of storing information about what's in a row."
All three Teraplex Centers -- created at a combined cost of $47 million -- are a focal point of IBM's company-wide business intelligence initiative unveiled February 26 of this year. "The Center's fundamental mission is to test databases and try to break them," says Daniel Graham, global solutions executive for IBM Global Business Intelligence Solutions. "In the data warehousing space, you want to know if a system can handle it."
IBM executives initially wanted proof that the AS/400 would do what IBM said it would, according to Graham. "When the Center opened in Rochester in the spring of 1997, it was the first time anyone in IBM has seen a clustered AS/400 solution," he says. "DB2/400 had originally been built to do parallel processing. With clustering, four independent AS/400s were linked together with OptiConnect -- a fiber-optic connection between machines -- to function as a single machine."
Graham points out the Teraplex Center also welcomes IBM customers and business partners looking for proof of concept. "Things happen at different stress points that you may never see unless you test for them," he says. "Proof of concept looks a little like a benchmark, because of course everyone wants to see how fast it will go, how big it can get, etc. It's very much focused on ensuring that, first of all, we can scale up -- meaning we do a test at one size, then double the amount of data and test again."
The AS/400 Teraplex Center is quite different from the AS/400 Benchmark Center, also located at IBM's Rochester facility. While the Benchmark Center reacts to sales requests and will take on workloads to achieve performance goals in a variety of areas, the Teraplex Center is more concerned with database scalability and is more proactive in its testing, according to Graham. "We know what the AS/400 has to prove, so we go and find circumstances where this is appropriate," he adds.
"We bring the customers and the business partners in, and we help them to scale test the AS/400 in these large configurations," Graham says. "That way the customer and the business partner both have a lot of confidence when they walk out that, collectively, IBM and its partners have demonstrated that it can be done, and whatever bugs that do need to be fixed have been exposed before the customer tries to get into a production environment."
According to Anderson, the Teraplex Center has had the greatest impact on two distinct groups within IBM. The first is developers of DB2 for AS/400 -- particularly their optimizer team -- which now has line items referred to as "Teraplex" line items.
"The other group we have managed to influence is the I/O subsystem of the AS/400, those who make the disk run faster," Anderson says. "We've done a couple of different tests that they were very keenly interested in, and they have gone off and made product plans based on what they learned from us."
The Teraplex Center has had a very cooperative relationship with the AS/400 development team, according to Anderson. "We've never had to go to a development team and push them to develop something," she says. "What I think is interesting is that, when the AS/400 was initially developed as the System/38, the guys who were developing it were reading the current research at the time, which said, 'someday we won't have disk at all, we'll just have memory.' The architects in Rochester at the time said, 'if there isn't going to be any disk, then let's design a system that won't need disk.' And that's what they did. With 40 GB of memory, you can literally run an AS/400 data warehouse in memory."