Network Infrastructure Blog

In a recent article from The UK register, new estimates are out that show IT now consumes 10% of the world’s electricity.  The increases can be tracked to many things including BYOD initiatives, data center growth, and the increase in internet sites to name a few.  With increasing pressure to lower power consumption whether it be from a green standpoint or just lowering the bottom line costs, IT professionals are tasked with finding more energy efficient ways to compute and support business needs in parallel.

SDN and newer fabric based technologies can help.  These technologies can be delivered with centralized switches nearly eliminating unused ports.  Likewise with SDN, assets can be placed on the data center floor where they make the most sense for power and cooling and assigned to the network needed without having to physically move them.

Stranded power (power allocated and not used) becomes much easier to manage in a more stagnant data center.  Decommissioning of unused equipmetn is critical in maintaining the balance.

Where do you think power will be in 10 years?  Hopefully we continue to learn to compute smarter.  Active electronics manufacturers continue to improve their products.  Design and layout matter more now than ever.  But most importantly, and like we have been saying all along, it’s time to view the data center as an ecosystem.  Every department that uses a data center has a responsibility to that ecosystem.  It’s time for decisions to stop being made in department silos, but rather by the ecosystem support team.

Google, IBM, Nvidia, Mellanox and Tyan work together on a new faster and more highly integrated Data Center Environment. Could this be a game changer?

Google alongside IBM, Mellanox Technologies, Nvidia and Tyan Computer are set to form a development group to improve datacentre technology. The alliance called the “OpenPOWER Consortium,” an open development alliance based on IBM’s POWER microprocessor architecture.

The Consortium intends to build advanced server, networking, storage and GPU-acceleration technology aimed at delivering more choice, control and flexibility to developers of next-generation, hyperscale and cloud datacentres.

The move makes POWER hardware and software available to open development for the first time as well as making POWER IP licensable to others, greatly expanding the ecosystem of innovators on the platform. The consortium will offer open-source POWER firmware, the software that controls basic chip functions. By doing this, IBM and the consortium can offer unprecedented customization in creating new styles of server hardware for a variety of computing workloads.

“The founding members of the OpenPOWER Consortium represent the next generation in datacentre innovation,” said Steve Mills, senior vice president, and group executive, IBM Software & Systems. “Combining our talents and assets around the POWER architecture can greatly increase the rate of innovation throughout the industry. Developers now have access to an expanded and open set of server technologies for the first time. This type of ‘collaborative development’ model will change the way datacentre hardware is designed and deployed.”

As part of their initial collaboration within the consortium, NVIDIA and IBM will work together to integrate the CUDA GPU and POWER ecosystems.

The newly coined term describes a staggering amount of data. The Pentagon is attempting to expand its worldwide communications network, known as the Global Information Grid, to handle yottabytes (1024 bytes) of data. (A yottabyte is a septillion bytes—so large that no one has yet coined a term for the next higher magnitude.) In the following link you can see Wired magazines look into the NSA data center.

See: Wired Looks at the NSA’s $2 Billion Datacenter in Utah

The storage war is on in the data center.  There are some early adopters of Fibre Channel over Ethernet, but Fibre Channel in native form still holds a commanding lead in the data center.  This is due in large part to existing fibre channel investments and the fact that this equipment is not fully depreciated.  Secondly, Fibre channel is higher speed than FCoE in currently shipping equipment with 16Gb/s versus 10Gb/s respectively.

There is also talk of flattening networks for optimal performance in a FCoE environment.  Flat fabrics will certainly operate faster (wire speed faster to be exact), but may not be feasible for data centers that have recent investments in a more hierarchical layer 3 configuration.

Another hurdle to FCoE adoption is that the current CNAs (Converged Network Adapters) are higher power than the Ethernet and SAN adapters together.  Granted it isn’t by much, but it is still measurable.

When implementing the FCoE protocol, the line between the storage team and network teams dissolve and equipment must be shared between teams.  Administratively, this can create issues with responsibility.

So will FCoE become viable in the near future?  Time will tell.

According to PHY manufacturer Aquantia, 95% of all the major switch vendors now have a 10GBASE-T offering. These switches increase bandwidth and throughput in data centers and are significantly less power than the equivalent 10ea 1GbE ports. Power is now in line with fiber, providing a lower cost option that supports 1 an 10 GbE out of the same hardware. This provides additional savings for data centers. Intel (R) began shipping 10GBASE-T native on motherboards in 2011.

For other benefits of 10GBASE-T including port savings and fewer switch purchases, please see Siemon whitepapers.

Siemon DC Design services will be happy to provide assistance on data center layout options to help you utilize this new technology. It is important to remember that with current shipping hardware, approximately 1-1.5 Watts per port can be saved using 6A or 7A cabling under 30m. The right design can further your power savings while upgrading network links to support virtualization, cloud computing, and other high bandwidth applications.

When you put 10GBASE-T in the switch to server links, the backplane also needs to improve.  While some manufacturers suggest using  multiple 10GbE uplinks, some also use QSFP assemblies for 40GbE, and some are already supporting 100GbE.   Visit Siemon SIS for supporting direct connect products.