Given the amount of attention it received at the recent Informix Solutions Portal ’99 in San Diego, Calif., it appeared the Internet Foundation.2000 platform is the only important new offering from the data warehouse vendor.
That, however, is far from the truth, since two of Informix’ key technologies are scheduled for major updates.
Both Informix’ own Informix Dynamic Server with the Extended Parallel Option (XPS) and the company’s acquired Red Brick star schema data mart are scheduled for major relaunches in the very near future — relaunches that are going to bring the products closer together.
The Spider-man release of Red Brick, due out by the end of the year, will add variable-length character strings, Level 4 JDBC capabilities and high speed data exports, but more importantly, from the Informix point-of-view, more seamless integration with the rest of the company’s product lines.
“What we’ve done so far,” explained Bob Walters, vice-president business development, data warehouse division, “is ensure that Red Brick talks to the other tools within the Decision Frontier Suite. We’ve done the first level of integration. All the parts work together and work together well.
“The second level is what you call user transparency among the products. There already exists a lot of end-user transparency, but I’m talking about the ability to administer the different database products from a common Java-based console. We have that capability in everything but Red Brick.
“The other aspect is connectivity — which is the great burden of all databases: how do I connect to this thing? We’ve got to use the same ODBC drivers from the same vendors, and we’re not there yet because of the Red Brick acquisition. So that’s also what I call the second level of integration.”
Since commonality is key to furthering integration, Walters said Informix has asked its developers to work on some small but key modifications to the foundation of the data mart and to its counterpart warehouse XPS.
“We have to get Red Brick and XPS to speak the same SQL. Each of them independently developed data warehousing extensions to SQL. We have to make those extensions common.”
Informix’ approach to this problem is to tinker a little bit with both engines, and not just adapt one to fit the other.
“It really has to be the greatest common denominator. You can’t subtract features from one, so it has to be a total cross pollination. It sounds like a lot of work but on the scale of database development it is really trivial.”
In addition to the reworked SQL, the new version of XPS, which is scheduled for a Sept. 30 release and carries the code-named Yellowstone, will focus on greater and more manageable power.
The 64-bit version, with ports to HP, Digital, IBM, Siemans and Sun Alpha servers, has a number of new features.
“The first is what we call dynamic workload control,” Walters said. “XPS was built as a power-user’s tool. When a query comes in, XPS will very aggressively give that query large hardware resources.
“With Yellowstone we’ve given the DBA the ability to dynamically alter the personality of the server anywhere from the previous power user mode to what we call information consumer mode. Information consumer mode is very stingy with resources because in this mode we’re expecting several thousand concurrent users. And of course you can run anywhere in between in the spectrum,” Walters said.
“Now that’s cool, but we’ve gone even further. What you can do with Yellowstone is fractionally change the machine. You can say ‘I want one-third of the machine to run at power user mode and two-thirds at information consumer mode,’ and then I can send certain user groups to each mode, and I can dynamically alter this.”
A dynamic join co-server feature has also been added to the Yellowstone release to increase the warehouse’s high availability configuration.
“Consider a cluster where I might have three nodes running and I might designate a fourth as a joined co-server. In normal operation, the joined co-server performs the high availability function. It just sits there and if one of the other nodes breaks, it fairly transparently kicks in and replaces that node. However, I also have the flexibility in a high workload scenario to give up the HA temporarily and having it join in queries — having it add its CPUs to the query load. And it is all dynamic too.”
Another key component to the new XPS is improved star schema capabilities.
“In Yellowstone we’ve made some fundamental breakthroughs in bitmap star-join technology. Now I’m sure my engineers would hate me for saying this, but this is sort of akin to what Oracle does.
“What we are able to do — and this has the unfortunate name of a push-down, semi-hash join — with relatively easy-to-administer bitmap indexes is accelerate a star join by a factor of three to five over joins without this acceleration. That’s a great improvement, and it is with bitmap indexes that are relatively easy to maintain. They don’t hurt your load times that badly.
“Contrast that with Red Brick, and its true star join. In the same scenarios we see a 10 to 30 times acceleration over a normal join. However, the Red Brick indexes are much more onerous, much more complex, so there is some degradation of load times. Most of our customers don’t find it bad, because the Red Brick loader is so strong, but that’s the trade-off.”
As for the joint future of Red Brick and XPS, Walters would only drop a few hints.
“What you’ll see in the future is that as the two products already overlap a bit, there will be an inevitable increase in overlap as the two products get more mature and powerful, however we will keep that in check largely by keeping the enterprise data warehouse positioning for Yellowstone, and by keeping the star schema/snowflake schema and its variants – blizzards, constellations – the domain of Red Brick.”
What is public knowledge, however, is that Yellowstone is already scheduled for an update next year. That release, dubbed Independence, will focus on improving both the warehouse’s high availability and OLTP functions.
Although Informix is willing to talk about the future of Red Brick and XPS, at least one customer isn’t sure if he is hearing the right messages.
“Having three codes [XPS, Red Brick and IIF.2000] is still a bit confusing,” said Mark Christmas, senior database consultant with CrossKeys Systems Corp. in Kanata, Ont. “I think they need to clean it up a bit more, or make it very clear where 8.x [XPS] fits in with 9.x [Informix Dynamic Server with the Universal Data Option — the Illustra Code]. For a while it looked like XPS was going to fold into 9.x. And now it is still a very large data warehouse without the universal data option [available in the 9.x]. They say the products are very complementary but the waters are still muddy.”
Christmas added that the lack of the universal data option will cause some confusion among developers.
“Which stream do I go down if I want to build a very large database now with a universal data functionality? The image was starting to get clear but now it is not so clear.”
Still he does like the proposed improvements to XPS.
“We already have a server that rocks, so let’s make it even better. Now it will be able to run on a single node, which is both good and bad. It just gives me another option, and adding failover and additional nodes are amazing concepts.”