Oracle Enters Hardware Business, Risks Angering Long-time Partners
Oracle has taken an interesting strategic turn by moving into the hardware business: a pair of servers, optimized to run Oracle's database and built and supported by HP. This seems like one of those decisions that will either prove to be a canny move or blow up in the face of Larry Ellison.
This isn't an impulse move, as the machine has been in development for three years and for the last year has been in tests with customers like Google. From a strictly technical perspective, if Google's demands can't break it, I think you'd be hard-pressed to find a company's that could.
Ellison, speaking at Oracle's OpenWorld conference, said large databases are creating a fundamental problem: Disk storage systems can't cope with data that has to be moved off of drives to database servers. He called it a "data bandwidth problem."As data gets larger the slowdowns become more unbearable. At one terabyte you will notice data bandwidth slippage. At 10 terabytes, storage systems crawl. "At one terabyte the problem rears its ugly head and it gets worse every year," said Ellison.There is some common sense to the assertion. If you have a PC, for example, a faster hard drive will give you better performance because the time it takes to write and retrieve data is always slower than the processing speeds. To make this work requires Oracle software to handle query processing and let the new platform search drives in parallel.
This isn't the first time Oracle wanted to get closer to hardware. Back in 1997, Ellison touted "server appliances" that bypassed operating systems by using a micro-kernel and "essentially running Oracle's database program on hardware." Even as Ellison bashed Bill Gates he said that this wasn't a marketing stunt, but rather an attempt to lower total cost of ownership by simplifying system management in the face of a shortage of trained database administrators. Micro-kernel or no, a database requires database administration, so it's hard to see, at least in retrospect, how the shift was supposed to reduce the amount of work. Perhaps even then Oracle was feeling squeezed between natural performance limitations and customer demands.
However, even if the hardware works as advertised, Oracle faces an enormous hurdle: its relationships with other vendors. Ellison specifically mentioned Teradata and Netezza in his speech, and he got a strong reaction:
"For years the old, incumbent vendors have tried to bolt together their products to try and tackle the data warehouse appliance market " and every one was a failure," said Jim Baum, president and COO of Netezza in a statement."You just can't slap together existing solutions in clever packaging and expect to deliver much faster performance," Baum added. "The power and simplicity of the data warehouse appliance model lies in integration and design from the ground up. Engineers in the same company, the same building, working to integrate a shared vision, "not patch it together with glue and spit."But there are two questions. One is about the market for these devices, and it's unclear whether the potential is that large:
"Pretty cool stuff, but it will likely have limited market reach, appeal," said Sageza Group analyst Clay Ryder. "The real question is just how many customers are there for a solution of this scale and performance?"[James Kobielus] of Forrester Research estimates that Oracle and its rivals sell some $9 billion in data warehousing software, hardware and appliances each year, a market that is growing at an annual clip of about 15 percent. Oracle has the largest share, followed by Teradata, then IBM, he said.The other big issue is that Ellison essentially threw a gauntlet down before every maker of high-end servers and every VAR and reseller that sells those machines and Oracle's software. The reaction from the sales channel and business partners has to be expectation that Oracle will try to steer as much business as possible to its own platform.
Even if logically the amount is not that much, a move like this will send a charge of fear through those who have traditionally sold Oracle products, and they might start hedging their bets by looking more strongly at where they could introduce competitive offerings. That is the potential "blow up" part of the strategy, because Oracle could conceivably create enough bad will to more than offset whatever financial gains it would make with the new product.