IBM’s New Power Servers Fuel Impressive Data Center Optimization

In 2008, IBM combined its “System i” and “System i” server architectures to form the Power Systems product line, also known as “IBM i.” Processors in this new line initially used Power6-based chips but were upgraded to Power7 architecture in 2010, thanks to a contract with the Defense Advanced Research Projects Agency.

Now, IBM is on the cusp of its next iteration, Power8, and has also made its existing Power technology available to outside manufacturers. As a result, the first-ever third-party Power server is poised to hit the market. Is this a game changer for enterprise IT?

Taking a Peek Under the Hood

According to a March 24 article from CIO, tech manufacturer Servergy will be the first to use Power processors in its new server, the Cleantech CTS-1000. This blade server is the size of a legal pad and weighs in at just over four kilograms (nine pounds). Servergy also has plans to join IBM’s OpenPower Consortium, which focuses on developing both hardware and software to support Power systems. Other members of the consortium include Samsung Electronics, graphics chipmaker Nvidia and search giant Google.

As for the CTS-1000, Servergy hasn’t released price details or an exact configuration, but the company says they’re targeting cloud and Big Data workloads with this release. The server comes with a 1.5 gigahertz (GHz), eight-core Power processor, which has led to speculation that Power8 is under the hood. But the server also uses PCI Express 2.0 rather than 3.0 ports, which are the new standard for Power8. Bottom line? Servergy says the Cleantech only uses around 100 watts of power, even running under full load, and is “16 times or more the I/O and compute density over traditional server technology.”

Wait for Power8?

There’s no official release date for Power8, but sites like IT Jungle predict a rollout at the end of April or the beginning of May. Intel’s Xeon E5 and E7 x86 processors are already dominant in this space and have recently undergone a “refresh,” so it makes sense for IBM to launch Power8 as soon as possible; combined with third-party Power servers, this should help boost its share of the market beyond internal IBM and Unix-based hardware.

There are also rumors that Big Blue will focus on companies looking to “scale out,” or add more servers, rather than on companies hoping to “scale up” by investing in more processors or RAM. Again, this is an Intel — and to some extent an ARM — market, but if Servergy’s focus is any indication, IBM plans to target enterprises looking to make best use of the cloud, Big Data and open-source technologies.

IBM is also working with Nvidia to move beyond the PCI Express interface by using a new interconnect called NVLink. This technology allows data transfer between Nvidia GPUs and IBM Power-based processors to be five times faster, effectively allowing graphics cards to make full use of CPU memory. Data flow is currently restricted to 16 gigabytes (GB) per second; when NVLink debuts in 2016, it will support transfer rates of up to 80 GB, accounting for the “even greater bottleneck between the GPU and IBM Power CPUs, which have more bandwidth than x86 CPUs.”

Power Up and Go Colo

One possible market niche for IBM’s Power Systems and Power8 architecture is coloaction hosting. For many enterprises, the cost of building and maintaining an entire data center is prohibitive, but the thought of moving critical data to provider-managed servers is daunting.

Colocation — in which a managed service provider offers the facilities, power and support necessary to host company data — is an attractive solution because companies supply whatever hardware they prefer. With its low weight and high power, the Cleantech CTS-1000 seems like an ideal colocation candidate, especially if it runs Power8.

There’s a new server in town, powered by IBM. Cloud-facing enterprise is Big Blue’s target market, and Power architecture has real potential, especially when it comes to colocation.

[image: Wavebreak Media/ThinkStockPhotos]