Article reposted form HPC in the Cloud Online Magazine. Article originally posted on Dec. 14th 2010:
2010 has been an incredible year for cloud computing in general and an even more exciting year for HPC and cloud. This is due, in part, to an increasing number of offerings designed to make high-performance computing applications perform better, flow with more steamlined management and make better use of the elastic resources that have become available.
As the end of the year approaches, it seemed like a great time to look back on some companies that shaped the HPC cloud ecosystem as a whole as well as to give a holiday “heads up” on some companies to keep an eye on in the coming year. There’s no way to put together a list that encompasses everything but here are a few honorable mentions.
This year Amazon took the world by storm with the announcement of services focused on HPC, HPC Clusters. Cluster Compute and Cluster GPU instances have been specifically engineered to provide high-performance network capability – allowing applications to get the low-latency network performance required for tightly coupled, node-to-node communication. Finally, it seems that affordable, flexible and elastic services have arrived for the HPC community.
Computing, and in particular cloud computing, is really all about the software and how to make the cloud work for you and not against you as a user. Adaptive has been around since mid 1990’s (formerly known as Cluster Resources) and provides intelligent automation software for data center, cloud, and high-performance computing environments. The company’s software solutions, powered by Moab, deliver policy-based governance that allows customers to consolidate and virtualize resources, allocate and manage applications, optimize service levels, and reduce operational costs. These services have allowed many users to get the most out of the cloud infrastructure.
Here’s a name that might be new to some of you. Nubifer’s mission revolves around making (and keeping) the cloud simple with a series of cloud program and services that enable users to easily configure and create cloud based services. One aspect of the company is its personalized and tailored architecture from any web-enable device–this means that part of their appeal is their technology-agnostic approach.
Clustercorp has an impressive sound byte – “Over 10,000 datacenters are power by Rocks Worldwide.”
Rocks+ is a complete cluster and cloud operating environment. Rocks+ can be used with Amazon’s EC2 to power large scale enterprise data and HPC workload. Rock’s creates single computing resource from multiple clustered systems. Remove the complexity drives down the costs.
First what a great name, not easy to forget. Whamcloud is basically picking up Lustre where Sun left off. The company provides vendor-neutral solutions for Lustre 1.6 and beyond. With years of experience developing Lustre features for high performance computing solutions – 50% of the TOP 500 fastest computers are powered by Lustre.
Yet another great name that’s certainly not easy to forget….
Cloud.com’s approach to cloud computing is to help organizations quickly and easily build, manage, and deploy private and public clouds. Extending beyond individual virtual machine images running on commodity hardware, the Cloud.com CloudStack provides an integrated software solution for delivering virtual data centers as a service.
The CloudStack’s secure cloud architecture, administrators can ensure that memory, CPU, network, and storage allocated to the individual virtual datacenter deployments are isolated from one end user to another. Certainly addressing one of cloud computing’s big challenges – security.
With many of the traditional big vendors reducing or even eliminating their spend in HPC markets Microsoft seems to be increasing their spend. Pushing the Azure and Azure services Microsoft’s cloud services vision starts to become a reality as the company continues to tout its proclaimed devotion to bringing high performance computing to the masses.
It is all about software management services here and many from traditional HPC have at least heard the name. After all, it’s the software makes the hardware work. The good news is that the world is recognizing that software and software management has been a missing link in the evolution of cloud computing. Platform has a rich set of cluster workload management software and have clearly targeted the HPC community and will likely continuting building its long legacy in HPC this year with more advancements for HPC cloud users.
With a broad array of system interconnects, Mellanox provide the fabric or glue that connects all the pieces together – Ethernet to Infinband, interconnect CPUs and Storage, adapter cards to switches. Mellanox has what can only be described as a “veritable smorgasbord” of interconnect products for high performance computing.
A pay-as-you go cloud computing model which is very attractive to small- and mid-size businesses as well as HPC users for the simple reason that it reduces capital expenditures and provides economies of scale not possible with the traditional datacenter model. Rightscale also provides a simple way to leverage Amazon’s EC2 platform, which is the top IaaS choice for many scientific and large-scale enterprise applications.
In 2009 the amount of digital content created and stored grew by 62 percent over the previous year, which had already been higher than any year on record. By the end of this decade the amount of data to be stored and created will be 44 times bigger than it was in 2009. This explosive growth in digital content, particularly unstructured content, has changed the rules of the game for businesses of all types. HPC is a huge creator and consumer of data, and it is more and more unstructured. Not only do you get both structured and unstructured but you also get high availability, manageability and high performance.
Is it conceivable that the HPC user community is ready for solid-state storage solutions? Answer is yes. Solid state has been around for 30 or so years in the HPC/supercomputing community from vendors such as Cray Research first half of 1980. Now SSD, based on NAND Flash memory, is back with a vengeance in several form factors as HDD replacements or more impressively as storage utilizing PCIe form factor. tachIOn from Virident provides a Tier 0 solution for high performance computing workloads, the goal is to eliminate the all to common IO bottleneck.