History of the company - lots of firsts:
1996: Predictive & backfill scheduling
2001: Integrated accounting
2006: Dynamic provisioning
2008: QoS Management
2008: Petaflop scale
2011: 1M+ jobs
The company is a leader in both HPC and cloud computing and both share a similar code base while the fact that they are involved in cloud does not mean they are deemphasizing HPC in any way.
The company is very technical - he is a past CTO and programmer - even the marketing team is filled with tech people. They have 16 patents and 30 pending - he believes this is greater than other companies their size. There is a huge focus on R&D.
Backed by top tier investors such as Intel Capital, Epic Ventures & Tudor Ventures and partners consists of HP, Penguin, Cray, NICE, Bright Computing, IBM, SGI.
We have 200+ customers and focus on large scale projects.
We power 40% of the top ten systems and 33% of the top 100 by cores. Sample customers below:
- Oak Ridge National Laboratory
- Los Alamos National Laboratory
- Lawrence Livermore National Lab
- Sandia National Laboratories
Last year at Moabcon you told us to get more involved with TORQUE. Although it is open source, you said we are the main contributors. We have dedicated our resources to making it more reliable and scalable.
We have also determined that our major product releases shouldn't have a "0.1" designation - [we will use whole numbers to signify a major change]. We will likely have one major and one minor release each year.
We will begin calling our releases by code names internally until they are ready for release. We will use names of explorers (Evans, Gatling, etc). Expect better usability and increased hardware support going forward until 2013.
We believe HPC experience better enables us in the cloud - our HPC knowledge in scalability, usability and migrations are useful while our experience in the cloud - dealing with elasticity, services-based UIs and cloudbursting helps HPC.
We see Exascale as the next frontier. Many of our customers have topped out their power – so you can’t multiply power by 100x.
Utilization issues are crucial – we don’t want to go from high nineties in utilization to 40s and 50s. Scalability will be job #1 for us.At Adaptive Computing Labs we are working on Exascale and other issues.
HP Keynote panel
Rob Hamilton (center) and Jerome Labat (right) of HP are interviewed by Adaptive Computing CEO, Rob Clyde. Brief overview of cloud benefits – we are putting together an architecture which works across models. You would think HPC customers would be early cloud adopters – an impediment could be having to move petabytes across the cloud. Customers are buying hundreds and thousands of servers at a time and you need to design servers, networking, storage, power, cooling and management as a unit. This is driving change – not only for software but hardware.
Yesterday we announced converged infrastructure cloud – from private to manage to hybrid and more. We are merging HPC and big data. Why not take crash data and merge it with HPC and big data?Cloud interop is a problem - can't move to Amazon Web Services from Google App Engine. We are using OpenStack in our public cloud and have made it partner friendly. We want to be open and leverage industry standards and innovate on top of them.