LINUX DOES TANGO WITH DECISION SUPPORT
 

Metapa's the name. Selling the value of Lintel decision-support software is their game. Not that they've come to bury DB2; they want to dance with it.

    
 
by Nancy Cohen
November 6, 2003		      
     
  Metapa Inc. is an LA-based company that for the last two years has been operating in the business intelligence/data warehousing marketplace, selling analytic solutions to very large service providers and telcos, and building specific data models and reports for these customers. Nonetheless, there is a growing demand for affordable solutions for business intelligence and data warehousing.

According to a white paper on business intelligence, industry analysts such as META Group peg the growth in the volume of data at approximately 125% per year. The high growth has heavy cost implications and headaches for CFOs as well as CIOs. How to continue to choose the right designs to store and access all this data in step with competitive goals that can't tolerate business at the speed of paper? How to scale proprietary databases to accommodate information growth, and at what killer price? Metapa's strategy is to pose and answer questions of their own, in perfect resonance for those enterprise users who need solutions  like yesterday: 'How do we bring software solutions that can change the very economics of business intelligence?' 

Last month, Metapa seized the database clustering advantage in acquiring Didera, a Linux database clustering company, expanding its own product development forays into business intelligence and data warehousing. In the works for Q1 2004: Something called 'Metapa's Cluster Database (CDB), which is to be a Linux database clustering platform.

The benefits of a database cluster is worthy of a white-paper in itself, but general benefits include:

1. BI applications in a clustered arrangement can aggregate greater amounts of data and deliver reports in much smaller batch windows, for real-time analytics.
2. Database clusters can provide new data warehousing functionality for a company that may not have implemented a mass data storage and retrieval system.
3. Database clusters can help relieve overloaded monolithic databases by reducing the extra load from the current database.
4. Distributing data into clusters enables 'parallelizing' the query process, shortening time needed to retrieve information from business systems. 

  
         
  With the acquisition of Didera, Metapa, which already had  technology capabilities to extract, transform and load (ETL) data, also wanted to do the data store portion. When they came across Didera,  they thought it made a lot of sense to put the two companies together. Metapa, which has been working with Linux from the start, will now be advocating the Linux clustered database approach as the ideal approach  for organizations coping with large amounts of data. 

"What we are doing now is building on top of PostgreSQL," says Powell. Why PostgreSQL? In going one-up on the big vendors in changing the 'economics' of business intelligence while matching the big vendors in performance, an Open Source database makes perfect sense, but why PostgreSQL? When Powell and crew looked at databases at the end of last year, he says, PostgreSQL at the time had the the most functionalities to match  for what Metapa wanted to do.

  
Metapa's Q1-Targeted CDB:
What is Metapa CDB:
Cluster database software application running on Linux and initially supporting PostgreSQL

Suitable for what:
Decision-support systems in a share-nothing environment where each instance acts on its own partition of data.

Sample Features:
Advanced SQL parallelization engine
Standards-based client interface (JDBC/ODBC/DBI)
 
     
  "Not that we aren't going to do it with MySQL somewhere down the road. Certainly we were aware of MySQL and its work with SAP, with all the functionality around SAP, but my view is that the SAP work will push them even further into transactional mode. And that relates to a divergence in database  functionality, where one becomes really focused on high-performance transaction systems or on high-performance business intelligence systems." 

Transactional mode is not where Metapa wants to go. "We are not doing anything  where you deal with very high number of small transactions," says Powell. "Our direction is decision-support systems, where you deal with a much smaller transaction load, but each query is much greater in its volume of what it asks for."

But hang on. Aren't there some other players in Linux database clustering game, names like Oracle and IBM,  that can easily outpace newcomers like Metapa? Last time we coughed, there were piles of press releases, case studies, and customer testimonials, for example, about the merits of IBM's DB2 Integrated Cluster Environment for Linux. How does Metapa intend to go up against the formidable marketing machinery behind IBM DB2?

Powell notes that IBM's DB2 Universal Database (UDB) offering  attracts business users for both data warehousing and transaction modes and that it has adopted a  shared-nothing clustered DBMS architecture. In turn, Powell says Metapa is hardly capable of unseating an IBM--to borrow from the Dana Carvey impersonations of George Senior--couldn't, wouldn't be prudent. 

Metapa will work with, not fight, the strengths of DB2 technologies. "One strategy that we are looking at  is how do we engage with the Linux side of IBM? With IBM's global services piece? Are there solutions we could deploy in concert with DB2?" Powell says that for those customers who still harbor reservations about putting a database of record on Open Source, "We will say go ahead and put data on a commercial database like Oracle or DB2,  but will recommend they pull their material for reporting off into a data mart and use Metapa's software built on top of Open Source PostgreSQL for that--i.e., extract for reporting purposes." 

In turn, he adds, "We won't seek to displace IBM. It's not that we are taking dollars out of IBM's pockets. Rather, we're dancing to the side, to allow customers to have additional reporting systems to meet their growing reporting needs, spending less dollars using commodity hardware and our software for a better price."

General availability of the Linux clustered database product is targeted for early 2004.