October 14, 2013
This week we can expect news from the Ethernet Alliance as it outlines its roadmap during the “Future of Ethernet” Technology Exploration Forum, which gets underway Tuesday and Wednesday.
In addition to featuring a talk from the inventor of Ethernet, Bob Metcalfe, members will strike a new path toward 400 gigabit Ethernet (the working group was recently formed)—a goal the group hopes to reach by the first half of 2017.
In advance of upcoming announcements from the event, and to recognize the fortieth birthday of Ethernet, we spoke with John D’Ambrosia, Dell’s Ethernet “evangelist” and chair of the Ethernet Alliance, an industry organization committed to advancing awareness of Ethernet technologies, expanding the ecosystem and adoption, as well as pushing interoperability, reach and new developments and Ethernet specifications.
We discussed Ethernet’s current role in HPC and what it might be going forward in some detail, with recognition of gigabit Ethernet to 10 GbE ratio of the Top500 as its own set of issues—not to mention how Ethernet in general will continue to match with InfiniBand clusters on upcoming lists.
The last Top500 list from the International Supercomputing Conference (ISC ’13) in Germany had a total of 205 InfiniBand systems, giving it a 41% share of the Top500. Gigabit Ethernet stole 140 systems (28%) and 10 GbE laid out 75 systems for 15% of the total list. To say that it’s a close split is an understatement, but as 10 GbE finds its way onto the board, this share is expected to alter in future lists—all matters we talk with D’Ambrosia about in the audio interview below.
D’Ambrosia notes that InfiniBand was given an opening in the market when it took longer to roll 10 GbE out, but as it emerges on the motherboard, it will be granted new life in HPC and beyond. Decisions about Ethernet are often simply a matter of cost, says D’Ambrosia, and the recognition for many datacenters, even those in HPC that aren’t as latency and bandwidth conscious as others, that it is “good enough” is still a strong argument.
He points to that matter of latency, however, noting that “there are things being presented in the Ethernet Alliance around how to bring those latency levels down closer to InfiniBand levels” including developments in the RoCE space. Although a look at the agenda for the upcoming event doesn’t reveal much in the way of conversations around that technology, D’Ambrosia insists it’s a topic that’s garnering a great deal of internal attention.
HPC is just a small slice of what’s happening in the wider world of Ethernet, of course. One look at the event’s lineup reveals that other topics, including standardizing SDN, the meshes between IEEE 802 wired and wireless technologies, the future of photonic integration, and new uses for Ethernet, including in vehicles and aircraft are on the menu for this week’s event. More info on this week’s Silicon Valley TEF event here.
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?