Oracle has announced a batch of servers based on new Sparc processors and in the process has begun an expected shift toward converging its two families of Unix servers onto a single chip architecture.
Oracle sells two lines of Unix servers, the T-series, based on the Sparc processors it designs in-house, and the higher-end M-series, which have traditionally been built by Fujitsu and resold by Oracle, and which run on Fujitsu's Sparc64 chips.
[ Doing server virtualization right is not so simple. InfoWorld's expert contributors show you how to get it right in this 24-page "Server Virtualization Deep Dive" PDF guide. ]
On Tuesday, along with a clutch of new T-series boxes, Oracle announced the first M-series server designed by Oracle and based on one of its own Sparc processors, rather than Fujitsu's Sparc64 chip. "This is all Oracle IP," said Marshall Choy, Oracle's director of systems solutions and business planning, in an interview.
Observers have long expected Oracle to converge the two systems lines onto a single chip architecture, a move that could reduce its development costs for both hardware and software, at a time when sales from its hardware business are declining.
"This is really the first instance of them delivering on this promise," said industry analyst Nathan Brookwood of Insight64.
Oracle announced five new T-series servers on Tuesday. They're mid-range systems based on its new Sparc T5 processor, which doubles the core count over the Sparc T4, from eight cores to 16, and boosts the clock-speed, I/O bandwidth and memory bandwidth.
They include a single-socket blade system, and rackmount servers with two, four and eight processor sockets. Oracle says they'll give a big performance boost for databases, Java middleware and business applications.
The new M-series server, called the M5-32, is a high-end SMP (symmetric multiprocessing) machine that Oracle will position against IBM's Power 795 Unix server, its biggest. It supports up to 32TB of system memory -- a huge amount -- and runs on a new, six-core Oracle processor called the Sparc M5, which shares the same core as the Sparc T5 chip also announced Tuesday.
The difference between the T5 and M5 processors is that Oracle has removed some of the cores on the M5, which aren't as useful in SMP systems, and added a much bigger Level 2 cache memory -- six times bigger, Choy said -- and other features suited to such high-end machines.
It's a similar strategy that Intel pursues with its x86 cores, Brookwood noted, which is to develop one processor core and adjust the cores per chip, cache size and other features to suit larger and smaller servers.
Oracle says its new servers will be highly competitive with IBM's Power systems on a cost and performance basis, and it plans to release benchmark results Tuesday intended to show that.