Adaptive Computing spins up Moab 6.1 control freak
Company 'fired up' about 'explosive growth'
Adaptive Computing – formerly known as Cluster Resources – thinks it's now in the IT sweet spot. As virtualized cloudy infrastructure becomes the norm in the data center, the company says, outfits encounter the same kind of complex job scheduling issues that supercomputer clusters have had for decades on physical infrastructure, and it's there to help solve these problems.
With Moab Adaptive Computing Suite 6.1, which comes in cloud and HPC flavors, Adaptive has tweaked the Moab job scheduler to be more useful for companies puffing up clouds as well as running jobs on HPC clusters.
The cloudy version of the tool is called Moab Adaptive Computing Suite 6.1, and the big change with this release is a new self-service portal that allows end users to request resources on a compute cloud and set up VMs, operating systems, and applications on virtualized servers as well as to activate physical servers if the cloud doesn't have enough iron turned on.
The cloud control freak has also had some of the code at the heart of the scheduler altered to make it more amenable to clouds. "The jobs running on HPC clusters are a little more static, but private clouds are more dynamic," Robert Clyde, the new CEO at Adaptive Computing, tells El Reg. "Moab had to be changed to reflect this difference."
And unlike a lot of other job schedulers grafted onto hypervisors and cloud fabrics, given its history in the HPC racket, the cloudy version of the Moab scheduler is future aware and is able to schedule workloads as much as a month in advance. "This lets you deliver better service level agreements," says Clyde. The system also has what Adaptive Computing calls "amber alerts", which means hooking into the predictive failure features of servers and storage to gracefully move workloads off machines before they fail.
On the HPC side, the underlying Moab 6.1 job scheduler is at the heart of three different products, called Moab Cluster Suite 6.1, Moab Grid Suite 6.1, and Moab Adaptive HPC Suite 6.1. The big change with the job scheduler in the 6.1 release is that it sports a factor of ten increase in the number of jobs that can be scheduled over the previous Moab 6.0 release for HPC clusters that came out in November 2010. The Moab 6.0 scheduler could manage a queue with as many as 50,000 jobs in it, but Clyde says the new Moab 6.1 scheduler has been tested juggling up to 500,000 simultaneously scheduled jobs with an average job submission time of under one second.
Clyde would not divulge pricing on the updated cloud and HPC releases using the Moab 6.1 scheduler. But last fall, the company said that Moab Cluster Suite 6.0 cost under $100 per server socket it runs on, while Moab Adaptive Computing Suite cost under $300 per socket.
Adaptive Computing's co-founders, David Jackson and Michael Johnson, are still involved with the company and have hired Clyde to take revenues up to the next level. And to duck.
On July 19, the same day that Clyde joined the company, Adaptive Computing made the headlines in the Deseret News when the company blew up an ancient Mitsubishi sedan owned by the VP of sales. This was not a retaliatory action by the Provo, Utah company for missing sales quotas, but rather a celebration for surpassing sales targets in 2010 of $10m and pushing it on up to $11m.
The car, packed with black powered and gasoline and its only crime being that it leaked a little oil, created a 60-foot towering fireball and gave the Spanish Fork Fire Department a training exercise. (Don't try this at your campus. The VP of human resources at Adaptive Computing is a Hollywood special effects guru).
Adaptive Computing has doubled its employee count to over 130 in the past year and boosted revenues by more than 50 per cent, says Clyde. The company secured $14m in its first round of venture capital funding from Intel Capital, Tudor Ventures, and EPIC Ventures back in September 2010 and is using that cash to fund hired and further development.
Adaptive Computing was founded in 2001 as Cluster Resources and has grown organically and has been profitable every year since then. Co-founder and current CTO David Jackson has done stints working at Lawrence Livermore National Laboratory, San Diego Supercomputer Center, the National Center for Supercomputing Applications, Pacific Northwest National Laboratory, and the Maui High Performance Computing Center and was involved in the creation of the open source Maui Scheduler. Michael Johnson is the other co-founder of the company and is currently its president he was previously the manager of Internet and security products at Novell. ®
Sponsored: IBM FlashSystem V9000 product guide