Gabriel Consulting Group (GCG) is a research, analysis, and consulting firm dedicated to helping our clients achieve maximum return on their Information Technology investment.

Find Us on Facebook Find Us on LinkedIn Find Us on Twitter

Read More...

Main Menu
GCG in the News
GCG Press Room
Recent Research
GCG Products & Services
GCG News and Views
About GCG
Search
Contact Us
Making Big Ones out of Small Ones, Part 1: 3Leaf PDF Print E-mail
Friday, 20 November 2009 00:00

The wheel is turning full circle. In the 1980s and 1990s, the HPC crowd pioneered scale-out processing and advanced parallelism in response to cost and scalability limitations imposed by the shared-everything systems of the day, ushering in an era where huge numbers of small computers were harnessed to solve large-scale problems.

'Read More' to see our video interview with 3Leaf...

Now we’re seeing new technology that allows users to join these small servers together to present a single system image and large memory space. Fast Infiniband connections are taking the place of custom-designed system buses and crossbars to provide system coherency at a much lower cost. This new model for large system images is more flexible than most of the old-school SMP servers, allowing users to deploy different configurations quickly and easily. Users can boot up a single o/s image spanning, for instance, 12 out of 16 servers. This would provide a large, single shared memory system to handle a workload that lacks the parallelism to run efficiently on scale-out. Later, the system can be deconstructed back into the 16 individual servers.

While this technology will certainly be useful, it isn’t a panacea. It probably won’t work very well on OLTP workloads or other commercial transactional applications. To us, it seems like the best fit is for apps that are memory-starved, or those that are difficult to parallelize. We also increasingly foresee a need for ‘unbalanced’ systems that can throw radically different amounts of processor, I/O, or memory to match unique workload needs. Different computing problems have quite different needs – some will become processor-bound, while others hit bottlenecks in memory or I/O. The standard ‘balanced’ x86 system isn’t offered in enough flavors to match each and every workload need, and non-x86 systems that have the capacity to provide massive amounts of resource also require massive expenditures. The easiest way to get around the limitations inherent in x86 systems is to allow them to be used as building blocks for larger boxes.

To learn more about this new technology, we talked to B.V. Jagadeesh, CEO and President of 3Leaf Systems. 3Leaf is a new entrant into the market, offering an innovative hardware ASIC which, according to BV, ties distributed systems together with a fast, low-latency Infiniband-based circulatory system.

 

Share this post

Submit Making Big Ones out of Small Ones, Part 1: 3Leaf in Delicious Submit Making Big Ones out of Small Ones, Part 1: 3Leaf in Digg Submit Making Big Ones out of Small Ones, Part 1: 3Leaf in FaceBook Submit Making Big Ones out of Small Ones, Part 1: 3Leaf in Google Bookmarks Submit Making Big Ones out of Small Ones, Part 1: 3Leaf in Stumbleupon Submit Making Big Ones out of Small Ones, Part 1: 3Leaf in Technorati Submit Making Big Ones out of Small Ones, Part 1: 3Leaf in Twitter