HPC Systems Inc offers datacenter integration, annual maintenance, incident response services for HPC & Enterprise customers. We offer services that include installation of supercomputers, HPC clusters, storage clusters, user training, system administration training, parallel file system administration training, porting of applications to high-performance systems, benchmarking and Infiniband management.  Large  HPC systems are complex and expensive to implement, and we could provide expertise in designing, integrating, deploying systems while minimizing risk for end users.

 

System Design Services for HPC Clusters and Parallel File System Clusters:
Our HPC experts work with end users to identify their specific needs and requirements. Using the most relevant technology and products, our team of system design engineers develop a balanced system design to match your budget and performance requirements. Our engineers then explain how our design helps achieve customer goals.

Our engineers are experts in all components of high performance compute systems at various layers – including hardware, interconnects, protocols, file systems, operating systems, compilers, applications, schedulers, management software, etc. At each layer, your organization faces the daunting task of choosing the right components from a plethora of technologies & products like the latest multi-core Xeon, AMD CPUs, PCI-Express expansion, Infiniband network, Myrinet network, Lustre storage cluster, SAS & SATA storage, Linux, ROCKS cluster management solutions, Ganglia monitoring system, IPMI systems management and more. We help see through the complexity and deliver the best solution for your needs.

 

Cluster Management & Deployment:
To install, configure, monitor and manage all the nodes in the cluster is complex task. Depending on the size and features of the cluster, deployment can make or break a cluster. Our engineers are proficient in popular cluster deployment suites like ROCKS (from NPACI) and other commercial cluster stacks. Our HPC experts can also successfully integrate relevant technologies like IPMI, KVM, out of band management, headless operation etc. to provide an integrated and comprehensive cluster management solution.

 

Job Scheduling and Resource Management:
Achieving  high levels of cluster utilization is a daunting task for many customers and often result in under utilized resources. Integrating a job scheduler system or a resource manager can go a long way in realizing the full potential of your investment. A multitude of resource management software suites are available in the open source community (OpenPBS, torque, maui, sge) and from independent software vendors (LSF, PBS, Cluster resources moab). Our HPC experts can work with you to identify the right solution for your cluster and help integrate the scheduler with your cluster.

 

High Speed Interconnects:
The performance of a cluster or its applications is determined heavily by the type of cluster interconnect used.  Gigabit tends to be the most popular and cost effective interconnect. Infiniband is the defacto high speed interconnect on many clusters including the Top500 systems. End users often think of Gigabit as simple and cheap. However, the type of Gigabit switch, cables, networks and drivers used can make a huge difference in the performance of the system. The same holds true of Infiniband and Myrinet networks. Infiniband network design is a complex task, when not done properly can inflate costs in terms of lost performance, cable lengths, number of switches used and cluster stability issues.  Our HPC experts can design a well balanced single level or multi-level Infiniband network topologies keeping in view costs and performance.  We also offer multiple classes of Gigabit network designs to meet your requirements.

 

Application Porting and Performance:
Engineers at HPC Systems also offer assistance in porting , optimizing the performance, profiling and debugging your applications. We offer user training in compilers (GCC, PGI, Intel, SUN etc), tools like TotalView, jumpshot, gprof etc. Our engineers can also offer assistance in benchmarking with standard codes, top 500 HPL benchmark and in sanity testing of the clusters.

 

Contact us today for assistance in any of the above services.