Ten Cool Supercomputers, Illustrated

The Jaguar supercomputer combines processing power with a little bit of style. Photo credit: National Center for Computing Science.

Pingdom has assembled a review of Ten of the Coolest and Most Powerful Supercomputers, starting with the Cray 1 in the 1970s right up through Roadrunner 2. The story provides a timeline of the advancements in processing power over the years, but also provides photos of these mighty systems, and links to more. Highly recommended.

Using Containers for Disaster Recovery

We’re starting to see some interesting case studies for servers in shipping containers. In a profile of Revlon CTO David Giambruno, ComputerWorld has some details on Revlon’s use of distributed data center containers in its disaster recovery network. Here’s an excerpt:

“Rather than have parallel datacenters and SANs in various countries, Giambruno’s plan put high-capacity storage at five sites across the world, consolidating data and applications at its U.S. datacenter. Using the same shipping system as for its cosmetics manufacturing, Revlon sent out five pre-loaded “Mini Me” datacenter containers to its four other IT centers, creating a global disaster recovery network of identical systems that assured resources would work when moved. These Mini Me datacenters have the SAN, storage, and severs for both local operations and can support external fail-over from other locations if needed.

Revlon says this approach reduced its datacenter power consumption by 72 percent asnd cut disaster recovery costs in half, as well as dramatically reducing the time required to back up 6.5 terabytes of data each week. Read ComputerWorld for more.

Green HPC Firm SiCortex Shutting Down

SiCortex, which specializes in energy-efficient systems for high-performance computing, shut down its engineering operations Wednesday and is selling its assets and intellectual property. The company’s customers include the Department of Energy and many major universities.

HPC Wire reports that SiCortex ran out of funding when its venture capital firms balked at additional financing. There’s additional coverage from GigaOm, and perspective from SciCortex insiders on blog posts by Jeff Darcy and Matt Reilly.

Combining Colocation and Containers

Colocation and data center services provider i/o Data Centers has deployed an SGI ICE Cube data center at Phoenix ONE facility, the companies said this week. The 538,000 square foot Phoenix ONE project doesn’t officially open until next month, but i/o Data Centers is beginning some customer installations. 

i/o Data Centers joins a short list of providers of colocation providers who offer infrastructure support for data center containers, and is the first to spotlight SGI, the server vendor previously known as Rackable Systems. CRG West is partnering with HP and Verari Systems to offer hosting for their data center containers.

Combining containers and colocation can provide rapid expansion for companies that need to expand their data centers but are out of power or cooling capacity at their existing facilities. Containers can live outside but still require an external source of power and security.  i/o Data Centers supplies a secure infrastructure for the ICE Cube that includes conditioned power, temperature-controlled space and network connectivity.

WolframAlpha Struggles in Traffic Tests

Remember how the new “knowledge engine” WolframAlpha said it was ready for 175 million queries a day at launch? It turns out the service and its supercomputer-backed infrastructure hasn’t performed well in traffic load tests, and this evening’s scheduled soft launch (the official launch is Monday) could be delayed, according to the LA Times. UPDATE: The WolframAlpha team went ahead with a live “launch” broadcast, but was about three hours late with its scheduled kickoff as the WA team battled a logging database and latency between two colo sites. By 11 pm Eastern time the WolframAlpha site was live and responding to queries.

Here’s an excerpt of comments to the LAT from Stephen Wolfram, the creator of WolframAlpha: “We have several supercomputer-class compute clusters,” Wolfram said. “One of our tests was to use one cluster to simulate traffic and run it against the other cluster. And when we did that last night, we found that the through-put we got degraded horribly when we increased the amount of traffic that we were pushing from one cluster to the other. So we don’t quite understand that, and that would very much degrade the through-put that we could get.”

Wolfram Alpha’s Search Horsepower

The Wolfram Alpha “knowledge engine” is the latest new search tool to be compared to Google. As it prepares for launch, the team behind Wolfram is beefing up its infrastructure in hopes of avoiding comparisons to Cuil, another would-be Google killer that crashed upon launch and saw its buzz dissipate quickly.

Wolfram Alpha will run on infrastructure spread across five data centers, and be able to handle 175 million queries a day at launch, the company said in a blog post this week. Wolfram’s computing horsepower includes “two supercomputers, just about 10,000 processor cores, hundreds of terabytes of disks, a heck of a lot of bandwidth, and what seems like enough air conditioning for the Sahara to host a ski resort,” the company said.

Uptime to Form Tier Advisory Group

The Uptime Instititute is creating an advisory group that will enable members of its Site Uptime Network to suggest changes to the institute’s tier system for rating data center  reliability, the institute said today. 

Uptime’s tier system sets forth a four-tier rating system for data center reliability, with Tier IV being the most reliable. While many facilities have been certified using the tier system, a far larger number of data center operators use the format to describe their reliability, often in ways that are hard to verify and wander outside the specifications.

The advisory group provides a process for end users to suggest adjustments to the tier system to address the changing environment. ““We’ve listened carefully to what the marketplace was calling for and have delivered,” says Julian Kudritzki, Uptime Institute Professional Services Vice President. “Owners will now be offered a place at the table to incorporate their perspective into the standards-setting process.

Voltaire Outlines Converged Network Strategy

Cisco and Brocade aren’t the only networking equipment companies targeting the converged data center network. Infiniband specialist Voltaire Ltd. (VOLT) yesterday introduced new switches supporting 10 Gigabit Ethernet using Converged Enhanced Ethernet (CEE), the new standard for Ethernet with reduced packet loss. Infiniband is a high-speed networking technology used primarily in high-performance computing, supercomputers and Wall Street trading platforms. Voltaire sells InfiniBand switching platforms and software offering connectivity at 20 and 40 gigabits.

The introduction of CEE, which offers many of the features of Infiniband, will offer additional options for low-latency networking. Cisco has teamed with NYSE Technologies to offer customized networking solutions for financial services apps using 10 gigabit Ethernet.

Voltaire has now introduced a “scale-out” Ethernet architecture built around new Layer 2 core switches based on CEE standards, designed to deliver high performance and low latency over a virtualized data center fabric. Voltaire has outlined its new architecture in a presentation on its web site.