Multicore Architectures for Cloud Computing. A Race not as open as it may appear
August 14, 2011
Grazed from GLG Research. Author: Editorial Staff
Only a few years back Azul Systems was building multicore server chips based on their own proprietary architecture, to efficiently support Java workloads. They had excellent hardware technology and even today they offer a 54-core chip in their Vega line. Lately they changed direction, and now they sell a JavaVirtual Machine that runs on top of multicore x86 server platforms. There is a steep effort to introducing new hardware infrastructure in the computation server market.
Multicore server chip development using ARM, MIPS, RISC, or x86 architectures are now all competing for a share of the Cloud computing market. With no clear winner, venture capital money has recently flown to some of them. If past history is to be used as a reference, de-facto standards and strong existing ecosystems will, once again, have the upper hand in the longer term…
Only a few years back Azul Systems was building multicore server chips based on their own proprietary architecture, to efficiently support Java workloads. They had excellent hardware technology and even today they offer a 54-core chip in their Vega line. Lately they changed direction, and now they sell a Java
Another company, Tilera, founded only two years after Azul Systems, develops RISC-based multicore chips for faster computation in data centers. Now, with a new infusion of capital earlier this year, they are targeting the Cloud, and they are not the only ones. Calxeda, a start-up building ARM-based servers, captured the attention of investors a few months ago. Several companies like Loongson Technology in China are betting in the MIPS architecture for their multicore offerings. Intel has been strongly defending their Atom platform for use in power-efficient server deployments and only last year they were showcasing 48-core chips in their technology labs in a clear move to massive on-chip multicore. Other public and private companies have as well posted press releases on new multicore product offerings, and many more are to be expected.
As the race to capture market share unfolds, customers will continue to look at support, compatibility, and low cost of adoption and maintenance, as key drivers when deciding their investments in Cloud infrastructure. Platforms other than x86-based, despite of their efficiency claims, still need to bring their entire offering to server levels to really meet customer’s needs. This implies not only performance and reliability from hardware components, but also a mature, dependable, and extensive software base. Not a small undertaking, which will require strong players in the industry with deep financial pockets, strategic vision, and long-term continuous commitment, to drive adoption.