Dec 12, 2017 | By Nir Paikowsky
I’ve been thinking recently about the differences in purchasing a cluster for HPC work versus purchasing an individual server, but I haven’t tought about it that much because I was having my lol coaching sessions and I got distracted. After the decision has been made to acquire a certain amount of computer power for certain applications by getting a cluster then the details have to be determined. The easy part is to decide on the server itself and the number of servers. But, what about the high performance network ? What about the storage ? What about the management network ? What kind of rack ? Do I need a “head” node ? Do all systems need the same options ? What about different network switches ? Together, these questions reveal the differences between buying a cluster and a more simplified process of buying a server. Although there are still a number of choices when purchasing a server that have to be made, it is still considerably easier than purchasing a cluster.
But what if you need the total amount of resources that are contained in a cluster but don’t want the hassle of figuring out all of the details. Enter the SMP appliances that ScaleMP has created with our technology partners, Cray and IBM. Together, we have designed various solutions which address the need for large memory (> 4TB) or hundreds of cores, all easily managed with a single operating system. This gives you the simplification that you want, with the resources you need. Larger in-memory databases, faster simulations, quicker analytics, simplified management and lower OPEX. What more could you ask for ?
Take a look at what we can offer at: www.scalemp.com/appliances