Oracle’s timing for “unveiling Exalytics In-Memory Machine” was unfortunate because it was in a shadow of Steve Jobs. In addition It was a lot of distraction between Larry Ellison’s and Mark Benioff’s egos.

Oracle is late to Analytics appliance game and have to fight already released products like Netezza/IBM (proven performer), SAP HANA (has large sales pipeline already), family of Teradata Appliances (Teradata Columnar coming in 2 months and sounds very good to me plus it packaged with Information Builders BI) , EMC/Greenplum Data Computing Appliance (doubled the sales during last year!), Microsoft Parallel Data Warehouse Appliance (Based on CTP3 I expect the great things from SQL Server 2011/2012/Denali) etc. They all are in-memory Machine, capable to store and process big data (exabytes? I guess depends on price…), almost all of them already have or will have soon columnar database.

Larry Ellison claimed during Oracle Openworld this week that “Exalytics is 10x faster than…just about everything.”

Yes, It runs a software stack that includes parallelized versions of Oracle’s TimesTen in-memory database and memory-optimized Essbase OLAP Server (“BI Foundation”), but it is not a columnar database, so I wonder how Oracle is going to prove Larry’s bold claims. However, Oracle TimesTen In-Memory Database for Exalytics supports columnar compression that reduces the memory footprint for in-memory data. Compression ratios of 5X are practical and help expand in-memory capacity (Qlikview, PowerPivot and Spotfire can do much better “columnar compression” then 5 times, claimed by Oracle)

Hardware itself looks impressive with four Intel Xeon© E7-4800 series processors (40 cores total) and 1TB of RAM but pricing is unclear. It has total 8 high speed ports:

  • 2 quad-data rate (QDR) 40 GB/s InfiniBand ports. When connected to Oracle Exadata, Oracle Exalytics becomes an integral part of the Oracle Exadata private InfiniBand network and has high-speed, low latency access to the database servers. When multiple Oracle Exalytics machines are clustered together, the InfiniBand fabric also serves as the high-speed cluster interconnect.
  • Exalytics has Two 10 GB/s Ethernet ports for connecting to enterprise data sources
  • Exalytics has Four 1 GB/s Ethernet ports are available for client access

Exalytics includes 3.6TBs of raw disk capacity. Optionally, clusters of Oracle Exalytics machines can leverage network attached storage.

Hardware portion of it probably below $100000 (I saw a guesstimate of $87000) but most expensive probably will be the Essbase (Business Intelligence Foundation Suite with in-memory Cubes now and ability to replicate entire data warehouse into TimesTen in-memory database) with list price about $450000, so we are talking  here about millions of dollars, which is (let’s wait and see the final pricing) will definitely reduce the number of potential buyers, especially considering weak Data Visualization and average BI functionality of Oracle’s software stack. According to Larry Ellison, Exalytics has 1TB of RAM but can hold five to 10TB of data in memory thanks to COLUMNAR compression.

Oracle Exalytics promotes self service analytics and makes it easier to develop analytics content by introducing a Presentation Suggestion Engine (PSE) which provides recommendations on type of visualizations to use to best represent a data set.

I do not expect anything spectacular from this “PSE”. For example Oracle proudly introduced “new micro charts and multi-panel trellis charts to visualize dense multi-dimensional, multi-page data on a single screen. The multi-panel trellis charts are particularly effective at displaying multiple visualizations across a common axis scale for easy comparison, to see a trend and quickly gain insights”:

but this micro charts available in much better shape and form for many years from Spotfire, Qlikview, Tableau etc. and relatively recently even from Excel.

In any case, Exalytics suppose to be well integrated with Oracle’s Exadata database machine and Exalogic application server. Mr. Ellison did some other bold claims like:

  • “For a given task, it will cost you less on an Exadata than it would on a plain old commodity server.”
  • “we move data around a hundred times faster than anyone else in this business”
  • “1,000 Exadata machines have been installed and 3,000 more will be sold this year”
  • “Java applications’ response times are 10 times as fast on Exalogic, and companies can serve many more users at once”

Special Note about Java.

I am not sure why Java is advantage for Oracle. Java is not welcome at Apple (can you say Objective C?), at Microsoft (can you cay C# ?) and recently even at Google (after Oracle sued Google for “misuse” of Java, which reminded me the Sun, disappearing after it sued Microsoft for … “misuse” of  … Java). Together those 3 companies have almost all cash (almost $200B if you exclude Oracle as a Java Owner) software companies have worldwide (Apple has $76B+ in a bank, Microsoft has $60B+ and Google has about $40B – may be less after buying Motorola Mobility) and I am simply following the money here. If Oracle wishes to have the Java-based advanced Data Visualization, they are better buy Visokio and integrate their  Omniscope with Exalytics and Exalogic instead of the inventing the wheel with PSE.