About the Author

Douglas EadlineDouglas Eadline PhD, is both a practitioner and a chronicler of the Linux Cluster HPC revolution. He has worked with parallel computers since 1988 and is a co-author of the original Beowulf How To document.  Prior to starting and editing the popular http://clustermonkey.net web site in 2005, he served as Editor-in-chief for ClusterWorld Magazine. He is currently Senior HPC Editor for Linux Magazine and a consultant to the HPC industry. Doug holds a Ph.D. in Analytical Chemistry from Lehigh University and has been building, deploying, and using Linux HPC clusters since 1995.

User Rating: / 0
PoorBest 
A blog about making HPC things (kind of) work
In a previous post, I pointed out cases where a single socket server is better than a multiple socket server for HPC applications. This conclusion is based on real testing and some hands-on experience with single socket systems. Of course your mileage may vary, but in my tests, the most economical and efficient systems seem to be those that use a single socket. Of course there is some cost amortization with the multiple socket nodes in terms of power supplies, cases, fans etc., but single socket processors and motherboards always cost less than their big multi-socket brothers and sisters.

Based on the growth of multi-core and multi-socket systems and the growth of Cloud based systems, you might think that small single socket servers are on their way out. What possible market could there be for small servers? It turns out, plenty.

There are two separate (but related) markets where both Intel and AMD see growth. The first is the Cloud providers, or perhaps better called the co-location providers. These providers like to offer "exclusive" servers to their clients. That is, users want exclusive use of the real physical machine. They do not want a virtual machine, but rather a real server that they can control. In these situations a single socket multi-core server is adequate for customer needs. Thus, providers are looking for inexpensive "microservers."

In addition to the co-location providers, there are always those customers that want a local server. Often called the small home/office market, there are customers who still like to keep their server(s) safely withing physical reach. A single socket multi-core system is ideal for these situations.

As mentioned both Intel and AMD have recognized this market. Intel currently offers the E3 Xeon series (Sandy Bridge) for single socket systems. These processors offer 4 cores, 8MB of L3 cache, ECC memory support, and are supported by several motherboard vendors (Intel, Supermicro, Asus, Tyan). Similarly, AMD offers its 4200 line of processors for the same market also with 8MB of cache and ECC memory support. They have also announced the 3000 series, which is said to be a cheaper version of the 4200 series. AMD will offers 6 and 8 core versions of these processors. Motherboard are available from both Supermicro and Tyan.

In the recent past, there has been very little action in this area and choices were limited. All this activity is good news for the single processor node approach to HPC. Similar to the early days of cluster HPC, feeding off of the large markets is an acceptable approach. Now that the microserver seems to have a bright future, it is time for some HPC lunch.