Eaton Corp. is introducing its Power Xpert Architecture today at the AFCOM Data Center World Conference in Las Vegas. It includes a combination of power meters and other hardware that Eaton says will help data centers determine when their power supplies are acting funny.

The power meters – already in Eaton’s portfolio – come in three different versions, the 4000, 6000 and 8000. The higher the number, the more the capability; for example, the 4000 meter records 1,024 samples of the power per cycle, while the 8000 records 100,000.

Those readings can then be viewed on the Web so IT folks can determine irregularities. Jim Thompson, a product line manager at Eaton, said the Power Xpert Architecture works with Eaton and non-Eaton power equipment. It also includes a 1U rack server that synchronizes time between multiple power devices, so when data centers want to backtrack to investigate a power hiccup, all the hardware is in tune.

“It detects small events so you can be proactive,” Thompson said. “One of the biggest issues is you end up with manufacturers pointing fingers at each other. A server manufacturer will walk in and say the server is locking up because of the power. By putting a power meter on the line, you can turn around and say, ‘No, it’s not.’ It gives you visibility into an item that historically has been dark for IT professionals.”

The power meters are $4,000 to $8,000, and the time synchronization server is about $4,000, Thompson said.

A post on Slashdot from last week addresses the disconnect between the increasingly multi-threaded, multi-core processors that are being developed by server vendors and the traditional serial programming aspect of most applications. The content referenced is Unix-focused.

Gartner analysts brought this issue up as one of the key problems facing IT at the Data Center World conference last November. Gartner predicts the number of cores on a chip will double approximately every 18 months through 2015. But those gains won’t mean much unless users address software. “We’re entering a software crisis,” Gartner analsyt Carl Claunch said. “If your software runs one big workload, you need to rearchitect it to be parallel. There is a challenge to make our apps multithreaded.”

It takes a while to get to the really interesting comments on this Slashdot thread, but if you wade deep enough you’ll find some real insight.

IBM has announced a breakthrough in chip design that will help cool processors more efficiently. Ars Technica reports that IBM announced the breakthrough at the IEEE Semi-Therm conference. Processors are designed with a copper cap that sits above the CPU core and transfers heat. This cap is coated in a thermal paste — which now thanks to the new method — can be more evenly distributed, resulting in a thinner layer of paste and nearly double the heat transfer according to IBM. No word yet on when to expect to see this in production, or whether AMD or Intel will buy in.

Rich Miller at Data Center Knowledge is reporting on the Data Center Dynamics conference in New York this week. He has a post on HP’s Dynamic Smart Cooling — a system of fans and computational fluid dynamics software that helps data centers avoid overlapping too much with their CRAC unit coverage. According to HP’s Ken Baker, HP is looking input into what third party cooling equipment vendors HP should begin writing drivers for to tie into the Dynamic Smart Cooling system. Baker also said HP’s own data center consolidation project is going well.

IT Jungle is reporting that a widening rift between IBM and Fundamental Software is starting to cause systems running Fundamental’s Flex-ES mainframe emulators to shut down.

The battle between the two began last fall when IBM and Flex-ES decided to get nasty with one another. IBM would not renew its agreement with Fundamental to allow them to sell Flex-ES emulators that included IBM technology, and so Fundamental refused to give IBM any more firmware keys to keep Flex-ES systems running.

The result? When the firmware dongles expire, say bye-bye to most of the Flex-ES systems.

The majority of Flex-ES systems are in the hands of PartnerWorld for Developers (PWD) members, who help developers create mainframe software. IBM employees also own a chunk of them to use on the road for mainframe software demonstrations (lugging an actual mainframe around isn’t feasible). The few commercial end users that have Flex-ES systems have firmware that won’t expire, so they can keep on existing as long as they don’t mind operating in 31-bit code and without IBM support.

If IBM and Fundamental can’t come to an agreement soon, PWD members may have to pony up the money for a real mainframe, which is much costlier, or come up with other options such as Hercules or Platform Solutions Inc. And we all know well that IBM is suing PSI, and PSI is fighting back.

IDC released a forecast today estimating that multicore processors and virtualization will cost the x86 market more than 4.5 million shipments and $2.4 billion in customer spending between 2006-2010.

“The server market is at a crossroads and customer buying behavior is increasingly driven by the strategic business benefit of the IT investment rather than a singular focus on cost containment,” said Matt Eastwood, program vice president for IDC’s Enterprise Platforms Group in a statement. “In today’s business environment, it is clear that technologies such as virtualization and multicore are particularly important enablers for the consolidated IT infrastructure IT organizations are increasingly seeking to deploy.”

“Each of these technologies is impactful to the market in their own right. However, the use of multi-core technology in conjunction with server virtualization tools has a compounding impact on server configurations, and accelerates the ability of IT organizations to exploit the benefits of multi-core technology,” said Michelle Bailey, research vice president for IDC’s Enterprise Platforms and Datacenter Trends in a statement. “Unlike other previous multicore introductions that took time to become mainstream as customers changed their application code, virtualization allows customers to fully exploit the improvements in x86 processors immediately, accelerating business benefits and thereby increasing adoption rates.” Looking forward, IDC believes the server and component vendors will optimize around quad-core technology before moving ahead to octi-core technology.

Other highlights from this study include;

• Server revenue growth rates will be lower in comparison, but are reduced to a lesser extent than shipment growth rates as customers deploy more richly configured systems in terms of memory, disk, and I/O to balance the increase in processing and server utilization.

• Despite the decline in the number of physical shipments, over the forecast period, growth in the number of effective processors continues to climb at a 25% annual rate due to multi-core technology advances.

• The number of virtual servers rises dramatically at a CAGR of 40.6% during 2005-2010 so that by the end of the forecast period, more than 1.7 million physical servers will be shipped for virtualization activities resulting in 7.9 million logical servers. This represents 14.6% of all physical servers in 2010 compared to just 4.5% of server shipments in 2005.

John W. Backus, leader of the team that created the Fortran computing language, died Saturday at the age of 82. Fortran, an abbreviation for Formula Translator, was released in 1957. It helped to “open the door to modern computing,” according to a recent New York Times obituary. After stints at the University of Virginia and the United States Army, Mr. Backus enrolled at the Columbia University where he studied math, a degree that helped him land a job at IBM attending to their punch card-operated SSEC “computer.” It was there where he assembled a team to tackle the problem of human-computer interaction. The result, Fortran, is considered the first successful high-level programming language by many computer historians.

Mr. Backus went on to develop other language-point interfaces that were all focused, in one way or another, on improving computational methods (as opposed to developing more powerful languages), including the Backus-Naur Form. It was John W. Backus’ rogue-style leadership and belief in the possibility that computing can be better that has helped us get to where we are today.

In an effort to push their products and cut back on energy consumption, Sun Microsystems and Pacific Gas and Electric (PG&E) have partnered to offer ways to cut costs in the data center with an exclusive rebate on Sun Fire T1000 and T2000 servers with CoolThreads technology. The program rewards PG&E customers who replace power-hungry servers with Sun’s  CoolThreads servers.

These are the first and only systems to qualify for a power rebate on energy-efficient data center products. Sun’s new CoolThreads servers were selected as a result of achieving the highest energy efficiency rating among servers, according to Sun. As part of PG&E’s Non-Residential Retrofit Program, customers who replace old, inefficient servers with Sun Fire T1000 and T2000 servers can now apply to receive a cash savings between $700 - $1,000 per server, PG&E reports.

You can check out the program at Sun’s website: http://www.sun.com/emrkt/energy-rebate/index.jsp

Server vendors have set very strict parameters for the temperatures and humidity levels at which their machines will function properly. Thus, datas centers have to support these fragile conditions with energy hogging cooling infrastructure. The question came up at the Uptime Institute Symposium: Why not just spend the money hardening the servers, expanding the thermal requirements and lowering overall cooling (power) consumption?

The idea is interesting and may have promise in some applications. But most high density data centers run the hot aisle, cold aisle method of cooling that faces server intakes toward air conditioned raised floor vents, directing all hot air to aisles on the backside.

If you increase the allowable thermal requirement, nobody will be able to work in the hot aisle and your Category 6 cabling is going to degrade in performance down to Cat 5.

Coding Horror writes about how Google was custom-building its servers way back in the beginning (images included). Thanks to Data Center Knowledge for the link.

Believe it or not, the Google history is only a decade old. The author, Jeff Atwood, applauds Google’s approach as an extension of any computer enthusiast building their own, better PC for less money.

Some people might look at these early Google servers and see an amateurish fire hazard. Not me. I see a prescient understanding of how inexpensive commodity hardware would shape today’sinternet. I felt right at home when I saw this server; it’s exactly what I would have done in the same circumstances.

The Google patents are also a fascinating diversion, with entries for cable management and cooling baffles.