This week we are looking at the basic cloud scenarios you might employ in your HPC environment. As you will see, each configuration offers benefits in the areas of cost, performance, scalability and capacity. The scenario that is right for you will ultimately depend upon your unique HPC requirements.
Public clouds are the most common cloud computing scenario. Computing resources such as hardware, software and other infrastructure, owned by the cloud provider, are delivered over the Internet to customers. In a public cloud, you share these resources with other customers, sometimes called ‘tenants’.
Advantages of public clouds:
A private cloud solution resides on a company’s network and often within their own facilities. This may be a good option for companies that already have expensive data centers and/or available shared infrastructure, as they can utilize their existing investments.
A private cloud most often consists of computing resources used exclusively by one company. It may be physically located on-site or hosted by a third-party. In either case, the services and infrastructure are usually maintained on a private network and the hardware and software are dedicated solely to you.
Advantages of private cloud:
Hybrid cloud is quickly becoming an essential tool for competitive HPC users. This scenario integrates on-premise infrastructure and public cloud, combining on-premises infrastructure with public cloud resources, allowing you to reap the advantages of both.
In a hybrid cloud, workloads can move between private and public clouds, creating great flexibility.
Advantages of hybrid clouds:
In the hybrid cloud configuration, cloud bursting occurs when a resource or workload runs inon premise until there is a spike in demand (or a new job is given priority), at which time the cluster can “burst” to the public cloud to harness the additional computing resources they need.
Many enterprises utilize the public cloud for bursting when their on-premises capacity is exceeded, perhaps because of high-volume workload spikes, or because a higher priority workload must take precedence. A cloud bursting strategy usually includes bursting based on backlog, on high-priority jobs, and/or on-demand.
Ready to learn more about hybrid cloud bursting? Check out this article by Univa President and CEO Gary Tyreman, “Capitalizing on Hybrid Cloud in HPC”.
We are always ready to talk cloud HPC, so send us your questions! firstname.lastname@example.org.