Exasol AG

I’ve spoken with Stu Greenberg (VP, Business Development) from Exasol several times in the past month and wanted to share my impressions of the company and their primary product offerings, EXASolution and EXACluster OS.  Exasol is a privately held corporation based in Nuremberg, Germany.  They are currently opening up operations in the United States, an effort being spearheaded by Stu.

EXASolution is a parallel in-memory relational data base management system.  Key features include an in-memory clustered RDBMS, in memory query processing, massively parallel data processing, compression algorithms, and integration via industry standard interfaces (e.g., ODBC, ADO).  They also tout a self learning and self optimizing management system that reduces IT maintenance costs by “up to 50%“ by reducing or eliminating many of the time-consuming system administration tasks.  EXACluster OS is the underlying operating system, built from scratch starting with the Linux nano-kernel.  The combination of these two components, along with off-the-shelf Intel-based hardware, allows a customer to build a very scalable, low cost data warehouse appliance.

Their recently posted TPC-H marks are very impressive.  Exasol’s posted score of 580,729 queries per hour (QphH) in the 1,000 GB test was nearly twice that of the next fastest mark, by ParAccel.  Their Price/QphH was about a third of the ParAccel cost, and about 3% of Oracle’s cost.  Exasol posted even greater numbers in the 300 GB test, and was second only to newcomer Kickfire in the 100GB cost category (although with more than four times Kickfire’s QphH).  I realize that these benchmarks are not always indicative of real world processing, and that Exasol has yet to produce these numbers in the larger data size tests, but I’m sure there are a number of companies with 1TB systems out there that should at least add Exasol to their evaluation list.

Exasol offers implementation and technical consulting support, focused on the installation and configuration of the EXASolution.  The company is relying on either the client or systems integrators to handle the bulk of the implementation work, at least in the initial US installations.  They also offer a number of training courses on EXASolution.

Exasol has partnerships established with all of the major BI players such as Business Objects (SAP), Cognos (IBM), MicroStrategy, and SAS. They also have partnerships in place with other key players such as Oracle, Microsoft, and HP.

Exasol has identified two customer references, IMS Health and KarstadtQuelle Group, both European based operations.  They are looking to leverage these long standing relationships to build out their US business.

I think for someone evaluating data warehouse appliances, Exasol certainly deserves to be on the evaluation list.  As an evaluator of their technology, at a minimum, I’d want to understand the following:

• What is the size and complexity of their current production systems, and how does their real-world performance compare to their TPC-H marks?

• What happens to performance if my production data set grows larger than my available memory (forcing some queries to disk)?

• What are the steps associated with implementing EXASolution, and how much support could I expect to get from Exasol throughout this process?

• Has EXASolution been tested with my chosen toolset (e.g., ETL, BI, Modeling), and are there production systems currently running with those tools in use?

• I’d like to see a demonstration of their “self learning and self optimizing management” capability.  In my mind, this is as (or more) important than the processing speeds and Price/QPhH, since hardware is increasingly cheap but your System Administrator/DBA time is not.

I’d be interested in other’s input on this subject, particularly someone who has implemented this product in a customer environment.

This entry was posted on 23. April 2008 at 05:38 and is filed under Vendors. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response or trackback from your own site.