Talentain’s HPC and Storage cluster experts are proficient with each of these layers, are backed by partnerships with industry leaders, and operate through a proven technology consulting methodology to deliver optimal HPC and storage cluster solutions. We strive to understand your business requirements and map these to relevant technologies to meet your goals. In short, we help you assess, identify and transform your business needs into working enterprise solutions.

Talentain can also work with your domain experts to ensure your organization is getting the most out of your enterprise solution infrastructure, using a proven solution assessment framework and performance tuning methodologies.

Here are a few of the challenges which HPC / storage cluster customers generally face in a new deployment or when tuning an existing configuration:

Cluster deployment and management: To install, configure, monitor and manage all the nodes in the cluster is a non-trivial and time-consuming task. A well designed tool will help achieve this with minimum use of human resources and provides a single point of control for the entire system. The tool should transform multiples of administrative domains to a single domain. Our engineers can help integrate relevant technologies like IPMI, vendor provided systems management, headless operation etc. to provide an integrated and comprehensive cluster management solution.

Resource Management: Cluster administrators often face the challenge of increasing the return on investment through higher resource utilization. Resource managers or job schedulers help manage the resource utilization of the cluster. With a variety of scheduling policies available, we can work with you to identify and implement the right policy for your configuration.

Server and Storage sizing: With the industry standardizing on the x86 platforms, users now can choose from an array of vendors and system architectures. Even though a platform maybe featuring the x86 technology and thus guaranteeing compatibility, the system design makes a good deal of difference in the performance you can expect from the machine. A system might feature a wide enough PCI bus for a high speed InfiniBand or a 10Gb Ethernet card, but does the system design provide a fast enough data path to the card? Your storage architecture might be feature the latest 4Gbps FC technology or the 3Gbps SAS technology, but is the storage subsystem configured to provide you with the most optimal performance? What is the right setting for the caching policies on your storage?

Interconnects: The application profile drives the choice of the interconnects – latency or bandwidth or both. However, majority of the clusters built are general purpose – i.e. cater to a number of applications in various verticals, not necessarily having similar profiles. Some tend to perform well on specialized networks and some on regular TCP/IP networks. The deciding factors for an interconnect are not just latency and bandwidth but also application portability, ease of integration in to existing infrastructure, compute node main board design and architecture, the load that the NIC places on the processor and more. With new protocol offloading techniques, one has to be prudent in making the right choice to meet the goals of the new investment.

Compilers and Performance Libraries: Vendor provided development tools were the only option, a while ago. Now with standards based computing products, users have the liberty to choose their development environment from a number of vendors or open source products. however the new challenge is integrating multiple components from different sources to provide a reliable and high performance development environment. This is especially critical when the user has to compile the code for his/her environment than buy a binary from an ISV.

Talentain delivers leading expertise in RISC and x86 clusters through partnerships with leading hardware vendors and enteprise technology providers.