Exadata Update Dramatically Boosts Speed of Data Warehouses, On Line Transaction Processing
Oracle claims that the latest version of Exadata can beat all competitors, including specialty solutions.
- By Stephen Swoyer
- September 23, 2009
When Oracle Corp. unveiled its first-ever Database Machine last September, it touted the combination of Oracle database, warehousing, and clustering software running on top of hardware from Hewlett-Packard Co. (HP). Few outsiders could have imagined that -- one year later -- Oracle would itself be a hardware vendor.
Oracle and Sun Microsystems Inc. last week showcased the latest version of the Oracle Database Machine, Exadata Version 2, touting a combination of Oracle software and Sun hardware -- with a special emphasis on Sun's FlashFire NAND technology -- which they claim can best all comers.
That includes its predecessor.
"Version 1 wasn't a bad effort. It was the fastest database machine in the world," said Oracle CEO Larry Ellison, during the product's launch event.
Exadata Version 2, Ellison claimed, ups the ante considerably. "It's twice as fast as Exadata Version 1 for data warehousing, but that's not the most interesting thing. The most interesting thing is it runs online transaction processing. Something Netezza just can't do at all. Something Teradata just can't do at all."
Ellison noted that " there are a lot of interesting new specialized data warehousing … start-ups that just can't do online transaction processing, but we can do both," he said. "We are still the fastest machine in the world for data warehousing, but now we're also by far the fastest machine in the world for online transaction processing."
Exadata Version 2 taps server and storage hardware from Sun. It can support raw disk capacities of up to 100 TB (in Serial-Attached SCSI configurations) or 336 TB (in Serial-ATA configurations) per rack. "The interesting thing about the architecture is that if you need more capacity, just plug in another storage server, just plug in another compute server, or just plug in another InfiniBand networking switch and you expand the system, and it expands from one rack to 8 racks to 16 racks to 32 racks … to petabytes and petabytes of data."
It uses a "smart flash cache" -- a memory hierarchy that comprises conventional DRAM and flash storage, powered by "very sophisticated algorithms" -- to accelerate random or non-sequential access. "This is a very smart memory hierarchy where the Oracle software manages that memory extremely efficiently," Ellison said. "The storage servers in aggregate have … 5 TB of flash storage," he continued, noting that the Exadata compute servers are populated with 400 GB of DRAM. "The aggregate throughput is 880 GB/s. "We can move data much more rapidly than any other computer in the world," he maintained.
Ellison also took aim at specialty players that tout the performance benefits of an in-memory architecture for either data warehousing and OLTP applications.
This is a grouping that includes solidDB (now an IBM Corp. property), Groovy Corp., QlikTech, and Exasol, along with VoltDB (nee H-Store) -- a new in-memory, OLTP start-up that features data management legend Michael Stonebraker.
"We're even faster than this new generation of database that's designed specifically to optimize queries for databases that are stored entirely in main memory," he said. "There's some new start-ups … that all they can do is [you have to] be able to load the entire database into memory. If you load the entire database into memory, they run pretty fast."
Ellison claimed that Exadata Version 2 is faster than these specialty efforts -- and said Oracle has the benchmarks to prove it. Performance claims aside, in-memory has been getting a lot of recent attention. In a white paper published this summer, for example, former SAP chief Hasso Plattner championed the use of an in-memory column store (with massive parallel underpinnings) as a means to accelerate both OLTP and OLAP workloads.
Industry veteran Mark Madsen, a principal with BI and DW consultancy ThirdNature, says Oracle's combined DW/OLTP pitch should chiefly appeal to thekinds of customers that bought Exadata the first time around -- namely,Oracle shops.
"You use Exadata for DW workloads, otherwise why would you buy it? [It] seems to be selling well to the Oracle client base, and it does speed up Oracle BI workloads, so it's a reasonable (if costly) solution relative to the appliance databases," he observes.
"I suspect they're saying this so people can move to Exadata from a regular Oracle install -- for example, not [having to] have two Oracle systems, or maybe to say that you can run BI and OLTP workloads on the same database. That's something Teradata and [HP, with] Neoview were both claiming, e.g. the 'active DW' model."
What About HP?
During his introductory remarks, Ellison positioned Exadata Version 2 as "a computer that's come out of a long-standing partnership with Sun Computer. For a long time, we've been working with Sun and this is the latest output of that partnership." Long-standing partnership or no, Oracle's relationship with Sun was considerably cozier back in May, when it announced its $7.4 bid to acquire the Unix giant.
Oracle's acquisition of Sun isn't yet finalized; regulatory agencies in both Russia and the European Union have said they require additional time to study the proposed acquisition.
Oracle's acquisition of Sun, and its Sun-centric Exadata Version 2 release, raise questions about its relationship with HP. Oracle officials, for their part, say they'll continue to sell Exadata Version 1 until they've exhausted existing inventory.
"Oracle will continue to sell the original HP-based systems [viz., HP Oracle Database Machine and HP Oracle Exadata Storage Server] while existing inventory is available, for those customers who request it," said an Oracle spokesperson in a statement.