InfraSpace specialises in products and services to enable high performance workloads – which are an important and fast growing niche. These unique capabilities are enabling the company to win a significant market share in the Big Data and AI and machine learning spaces.
InfraSpace previously hosted all of their AI and Big Data workloads in AWS and Azure, but the cost profile was far too high for the performance profile they required. They looked at ways to reduce costs and concluded that building their own datacenters was the best option. 2 years later, all of their workloads have been migrated to their own infrastructure. Initially they used VMware as their virtualisation technology of choice, but found that not only were they getting a fraction of the performance they expected when deployed on modern hardware - such as NVMe drives and processors using NUMA memory. Also, management of the VMware estate was getting more and more onerous. The focus on demanding workloads together with VMware's performance constraints meant a constant need to buy more servers, memory and storage. Tasks like upgrading from vSphere 5.5 to 6.5 were a major project that detracted from their ability to service customers.
In addition - InfraSpace had started to offer rural 'microdatacenters' in the Indian market that could be deployed near to smaller centres of population - but by their nature required extremely low power consumption without compromising performance.
InfraSpace evaluated the Hyperconverged Computing Infrastructure (HCI) solutions on the market to find a virtualisation product that was able to access the performance of new technologies such as NVMe storage, could scale out easily and could operate on resource-constrained processors. Sunlight was the natural choice.
This paper takes a look at how InfraSpace used Sunlight to underpin their MSP services.