Analysis: Strong Reactions to Oracle's Database Machine
The Oracle/HP Database Machine is notable for being as ambiguous as it is ambitious. No one seems to know just how much it costs; which options it includes (or requires); or how much storage it encompasses.
- By Stephen Swoyer
- October 1, 2008
When Microsoft Corp. purchased data warehousing (DW) appliance specialist DATAllegro Corp. a couple of months ago, it threw down the gauntlet to rival Oracle Corp., claiming that DATAllegro's technology would enable its SQL Server database to "leapfrog" over Oracle in the high-end DW segment.
No one expected Oracle to ignore such a challenge, but few could have anticipated what Oracle announced last week. In tandem with hardware partner Hewlett-Packard Co. (HP), the company unveiled its Database Machine. The Oracle/HP Database Machine is notable for being as ambiguous as it is ambitious: no one seems to know just how much it actually costs, which options it includes (or requires), or how much storage -- high-performance or high-capacity -- it encompasses.
On the other hand, if the Oracle/HP Database Machine really does weigh in at or close to its $14,000-per-TB reference figure, and if it really can scale to support the DW configurations that Oracle chief Larry Ellison trumpeted during last week's unveiling, it will be a formidable competitor -- in combined Oracle/HP shops, at least. For mixed environments -- and in shops that run non-HP (or a heterogeneous mix of) hardware -- it might not get a second look.
"[Our] concern is that [the Database Machine] will obviously only ever work on Oracle systems, but we have several SQL Server [data warehouses] too, so we're also looking into [an appliance from] Dataupia -- which happens to be much cheaper … and [which] works on Sun as opposed to HP [hardware]," says a data warehouse architect with a prominent utility services company in the United Kingdom.
"I think [Oracle and HP] have a lot of work to do in positioning it, but for 100 [percent] Oracle shops, it looks like a dream come true! I think they'll struggle with trying to force people down a particular [hardware] route."
This user, like others in the industry, had heard some rumblings that Oracle was prepping something Big on the DW appliance front.
Just how big was the question -- until last week. Oracle's Database Machine announcement was probably the biggest news to come out of OracleWorld: CEO Larry Ellison even made it the centerpiece of his keynote (dubbed "Extreme. Performance."), touting Oracle's "first-ever hardware product."
Ellison cited a litany of familiar problems -- exploding data volumes and lagging storage performance foremost among them -- arguing that today's disk systems cannot cope with the volume of data that must be moved off [physical disk] drives.
Enter the Oracle/HP Database Machine, which comprises a combined hardware and software solution, sold by Oracle, serviced by HP. It's powered by a grid of database servers, which collectively act as a kind of query aggregator, distributing queries to each of the Oracle HP Exadata storage servers. The latter don't simply pass disk blocks back to the Database Machine -- they pass query results.
The default configuration specifies a grid of eight database servers (populated with 64 Intel processor cores and running Oracle Enterprise Linux) and a grid of 14 HP Oracle Exadata storage servers. The latter can support up to 168 TB of raw storage and deliver 14 GB/s throughput back to the database servers.
The performance boost, Ellison claimed, is dramatic: anywhere from 10 to 72 times the performance of a vanilla data warehouse.
A Mixed Reception
Industry watchers had mixed reactions to Oracle's Database Machine gambit.
Take Jill Dyché, a principal with Baseline Consulting (and a Teradata veteran). Dyché's reaction simultaneously highlights both the underwhelming and the overwhelming aspects of Oracle's announcement.
On the underwhelming front, Dyché explains, the Database Machine isn't exactly a novel entry. "My first thought was, hey, Ncube returns!" she comments, citing a Database Machine-like offering that Oracle announced in the 1990s. "In those days, [the] problem was that Oracle didn't perform well in the ad hoc query world."
On a more positive note, the Oracle Database Machine should help counter a lot of the fear, uncertainty, and doubt that Netezza, Teradata Inc., and other competitors have traditionally flung Oracle's way, Dyché suggests.
"There has always been a question about Oracle's commitment to data warehousing, since the bulk of its customers were running its DBMS in more transactional environments," she points out, "but in his keynote at last week's OpenWorld, Larry Ellison took on both Teradata and Netezza, comparing the latter to his new product in his slideware. His new product's name seems a blatant attempt at marginalizing the former. I'd wager that Oracle has surmounted whatever horsepower issues they've traditionally battled and are ready to go head-to-head with the other database hardware solutions."
Veteran data warehousing consultant Claudia Imhoff highlights another aspect of Oracle's announcement: it helps raise the profile of the DW segment in general -- and of DW appliance-like players in particular.
Indeed, Ellison attacked both Teradata and Netezza in his keynote -- a tactic that, regardless of his or Oracle's intentions, could redound to the benefit of both companies. Netezza claims that it has fielded a bevy of customer inquiries as a result of Ellison's attack. Ellison's keynote was as notable for the vendors he targeted (Teradata, Netezza) as those he ignored (Microsoft, IBM). "[I]t makes me wonder who Oracle sees as their real competition," Imhoff says.
On his DBMS2 blog, savvy industry-watcher Curt Monash, a principal with Monash Research (and an object of praise from no less of a luminary than Larry Ellison himself), assessed both the real-world cost and the comparative parallelization of the Oracle Database Machine.
Cost-wise, Monash notes, Oracle's default pricing scheme is very fuzzy: e.g., it doesn't take discounting -- which Monash says is rampant -- into account, nor does it provide any indication which features, such as Oracle's Real Application Clusters, or RAC, are actually necessary. It also doesn't distinguish between high-performance (300 GB, 10,000 RPM) and high-density (1 TB, 7,200 RPM) disk storage, which, according to Monash, means that "the price/terabyte can vary by a factor of 3.3."
With respect to parallelization, on the other hand, the Oracle Database Machine seems like a distinct departure from the massively parallel processing (MPP) status quo. "Oracle has moved into the node heterogeneity camp. By way of contrast, the usual-suspect MPP vendors -- Teradata, Netezza, Greenplum, Vertica, Aster Data, Paraccel, Exasol, DATAllegro -- do most or all of their subsequent processing on the same nodes that retrieve data," Monash writes. "Thus, Oracle is the first major vendor for whom it is important to remember that different parts of a query plan get parallelized across completely distinct sets of processors and processing units."
It remains to be seen what that will mean in practice, Monash concedes: "[H]ow good is all this parallel technology? On the one hand, we know Oracle has been shipping it for a long time, and has it widely deployed. On the other, we also know that Oracle performance has been very problematic for large parallel queries. Surely most of those problems were due to the shared-disk bottleneck, but were they all [or mostly all]? I don't yet know."
The UK-based data warehouse architect introduced above says he's now less likely to consider Oracle's Database Machine -- in spite of the fact that, by his own admission, he was waiting for Oracle to announce something about MPP. His reaction also highlights some of the difficulties that Oracle -- along with RDBMS competitors IBM and Microsoft -- might have in countering the attractiveness of strict DW appliances from Netezza and other competitors.
"[I]f it were purely down to me, I'd rather just scrap the current system [of mixed Oracle and SQL Server data warehouses] and start again with Netezza -- we'd have a brand new, dazzlingly fast solution within a few weeks," he concludes. "The best project I ever worked on was a greenfield telco [in] which I was lucky enough to select and implement a Netezza system back in 2003. It was a dream come true to someone who has worked with Oracle for many years."
The Database Machine raises a boatload of questions, starting with HP's commitment to its Neoview DW platform, but Oracle and HP seem to have settled at least one thing, Baseline Consulting's Dyché concludes: "Both Oracle and HP are taking data warehousing very seriously. Again."