Last week we talked about the upcoming release of Intel’s Xeon E5 processor family. This week, we have some even more important announcements regarding Intel MIC and the strategic direction that Intel is headed regarding high performance computing.
Die shot of 'Aubrey Isle,' the silicon chip included in the Intel MIC 'Knights Ferry' development platform
We hope everyone has thus far had a great start to 2012! To help launch us further into the new year, we are running a special “social servers” campaign through the end of February to help bring everyone together for fantastic savings! The Stellar Deal on our 5017C-MTF 1U Single Xeon E3 “Sandy Bridge” server takes the form of social shopping for servers where a special discount of 20% will be applied if the campaign goal of 50 servers is met. If we reach our campaign goal, everyone gets the deal and benefits! If we don’t meet the goal, you won’t be charged but nobody gets the deal.
IBM’s new data center offers bold new ideas about how to cool servers. It has long been taken for granted that the best way to control the temperature in a server room was to keep it as cool as possible by blowing out the heat generated by processors inside the servers with chassis fans. IBM’s new data center in Zurich, Switzerland uses liquid cooling to do the same job. In a previous post, I talked about liquid cooling as an alternative to air cooling, but the new data center in Zurich is unique: it’s cooled by hot, not cold, water.
As described in an article on ServerWatch.com, a new idea is emerging in the server industry about how to moderate the heat produced by running servers. Instead of controlling the temperature of the server room itself, it’s much more important to focus on cooling the server components, regardless of how high the temperatures rise in the server room. Recent innovations in air cooling, such as allowing air from outside the building to be cycled through the server room even in locations as hot as New Mexico, show that a hot server room by itself does not dampen the performance of the servers.
But CPUs overheating does. The new Zurich data center, unlike previous liquid-cooled clusters, uses water at 140 degrees F to cool CPUs that run at 185 degrees F. It’s not necessary to chill the water, IBM reasoned, because even warm water is much colder than the temperature of the processors that it heats. So, to save money, the Zurich data center does not chill the water, and moreover pumps it (once it’s been boiled by the servers) to help heat nearby homes.
According to EnterpriseITPlanet.com, “The combined carbon reduction from using less electricity to cool the servers to the recaptured heat for heating purposes is a whopping 85 percent.” As equally important, IBM saved about 40% on their energy spending by switching to hot-water cooling of their servers.
The bottom line, it seems to me, is summed up by Jed Scaramella, senior research analyst for IDC’s Enterprise Platforms and Datacenter Trends as quoted in the ServerWatch article, when he said, “if you walk into the room and the room is cold, that doesn’t tell you much [about how well the servers are cooled]“.
Scientific Computing magazine’s latest Q&A article addresses a topic very important to most professionals who own or operate a data center: how to improve performance and save money on energy costs.
The magazine asked seven industry leaders in the HPC industry for their opinions about this subject. Their answers are very telling and reveal the state of data center technology today and the directions it will likely take in the near future.
Perhaps the greatest expenditure facing data center operators today is in the electric bill they pay to keep their servers running. While HPC (high-performance computing) servers have become relatively cheaper in recent years with such innovations as GPUs (graphics processing unit), the price of powering ever-expanding data centers has only grown.
There are several bottlenecks that are contributing to these higher costs. The most common problem mentioned by the industry experts in the Q&A article is how to reduce the cost of the cooling systems in server rooms. As Bob Masson of Convey Computer noted, “Every watt required to power a server nominally requires a watt to cool it.”
For this reason, VP of HPC Consulting at NAG Andrew Jones recommends that data center operators make a long-run analysis before they buy new equipment. The price of a server isn’t just what you pay on the day you buy it. The cost also includes what it will take to power that server and cool it over time (and, if you’re buying dozens or hundreds of servers, this begins to add up significantly). That is the true cost of a server, and from this perspective, power and cooling specs become very important.
This kind of an analysis is not only useful before buying a server but also for evaluating an existing data center. Old equipment, as Blake Gonzalez of Dell HPC describes, can be so expensive to run due to power inefficiencies that some companies can save money by buying entirely new servers altogether. But before any such decisions can be made, Gonzalez notes, one has to be able to measure current power consumption accurately and know how to estimate the long-run cost of replacement servers before buying them.
When it comes to server processors, bigger and faster isn’t always better. Linux Magazine has published an article about how using less powerful processors (but more of them) can actually be a smarter decision than using top-of-the-line CPUs.
The author of the article, Douglas Eadline, compares building a cluster with Intel Xeon processors to creating one with Intel Atom CPUs. The Xeon is more powerful and more expensive than the Atom, but requires substantially more power.
Eadline calculates the advantages of building with one processor over another. The Price of one Xeon/Price of one Atom ratio is 22 while the Performance of one Xeon/Performance of one Atom ratio is only 7.7. That means that, while you may pay 22 times more for a Xeon than an Atom processor, you are only getting 7 times the performance of the lesser expensive CPU. This comparison suggests that, based on these parameters alone, buying servers with Atoms may be a better choice than purchasing systems with Xeons.
But the matter is not so simple. While each Atom processor draws drastically lower quantities of power than a Xeon processor, these savings are negligible when one looks at the bigger picture. The Thermal Design Power (TDP)/Performance ratios of the processors, Eadline calculates, are about equal, meaning the two processors do not differ much in terms of power consumption.
Also, Eadline observes that while you may get better performance for every dollar with an Atom than a Xeon, this only holds true when comparing strictly processor prices. But, a cluster with Atoms will typically have more motherboards and memory DIMMs than an equally-powerful cluster with Xeon processors. To calculate whether Atoms are really a better deal than Xeons, one has also has to figure in the cost of those extra pieces of hardware.
In addition, some server functions are better performed by Atom processors than others. The Web hosting industry, which does not run large databases or complex modelling software, typically only needs to store and access web pages, a relatively simple function. Clusters with Intel Atom processors are well-suited for such functions. SeaMicro, a manufacturer of large server units, has built a 10U server system with 512 Atom processors that only consumes as much power as four desktop computers, reports ServerWatch.com. The individual Atom systems, which have smaller motherboards, were re-engineered to have a compact design which is not possible to achieve with clusters running more powerful processors.
So Atom clusters may be a viable solution for individuals or companies whose servers do simple functions. Even then, one must compare a variety of factors to determine whether “cheap in quantity” over “expensive with quality” is the wiser way to go. In any case, Atom clusters are definitely an option to be considered, especially for hosting companies.