August 30, 2012
The upcoming Supercomputing Conference (SC12) may not turn out to be the blow-out high performance computing hullabaloo it normally is. The recent GSA scandal involving overzealous spending at one of their conferences a couple of years ago has precipitated new federal policy that is forcing government labs to abandon their exhibits and cut back attendance at the world's largest supercomputing event.
The US General Service Administration (GSA) problems centered around the their 2010 Western Regional Conference in Las Vegas, which was being done for the purpose of employee training. Apparently, the agency spent over $800,000 on the event, which included things like excessive spending on scouting trips for the event ($130K) and an exorbitant food budget ($146K). The agency also hosted semi-private parties and meals for non-employees. A good time was had by all.
But what happened in Las Vegas didn't stay in Las Vegas. A report was generated by the GSA Inspector General that detailed the agency's misdeeds. Once Congress became aware of the lavish spending, the people in charge were hauled before the Senate and House and pummeled unmercifully. The revelations prompted the government to take a more systematic look at federal spending policies for conferences.
That rethinking prompted much of the wording in an Office of Management and Budget (OMB) memorandum that directs the heads of federal agencies to reduce travel expenses in FY 2013 by 30 percent. That memo begat this one from the US Department of Energy (DOE) that spells out how the agency is going to handle the new restrictions.
Essentially, the DOE is setting the nominal limit on conference spending to $100,000 -- not per DOE lab, but for all DOE spending at a conference. For any event that would exceed this limit, approval from the Deputy Secretary is required. If travel expenses in excess of $500,000 are desired, a waiver from the Secretary is needed. And in an election year, that kind of action is liable to attract unwanted attention.
This is a huge problem for the DOE laboratories that want to exhibit and otherwise participate at SC12. There are more than a dozen energy labs that exhibit and send numerous attendees to the conference. When you do the math, severe cutbacks are going to have to be made. Given that each of these booths probably cost between $100K to 400K, many of the labs may opt to apply their travel allocations to send attendees rather than host an exhibit.
In general, the DOE booths include some of the larger ones on the show floor, including super-sized ones from labs like Oak Ridge (ORNL), Berkeley (LBNL), Argonne (ANL), and Pacific Northwest (PNNL). As of yet, neither of those four labs have pulled their exhibits. But given the cost restrictions, it's hard to see how any of these will survive.
According to SC12 communication co-chair, Ian MacConnell, Brookhaven, Los Alamos, NNSA, and Sandia have already canceled their exhibits. He goes on to state that each of those labs will be in attendance -- presumably involved in paper presentations, tutorials, BoFs, and so on. But again, given the severe restrictions, it's likely that even those labs will not be able to send their full contingent of people.
MacConnell says that US government agencies represented less than 10 percent of the overall attendance last year at SC11. But that's still on the order of a 1,000 people that are in jeopardy. And they represent many of the key individuals in the US HPC community, inasmuch as a good number of them are involved in procuring, programming and running some of the most powerful supercomputers in the world.
NASA and the US Department of Defense (DOD) also have to follow the OMB restrictions, but since they have a relatively small presence compared to the DOE, they should be able to survive reasonably well with some belt-tightening and booth downsizing. The NSF-funded supercomputing centers like Pittsburgh, TACC, and NCSA, could theoretically be in trouble too, but none of the centers HPCwire has been in contact with have mentioned they were changing their plans. These centers are affiliated with local universities so are not entirely dependent on federal funds. The NSF has not yet responded to our inquiries about how the OMB directive will affect its SC12 participation.
In any case, a reduced US government footprint at SC12 is likely to change the look and feel of the event. From an HPC vendor's point of view, these agencies are among their largest customers. Certainly at the level of supercomputers, the US government is the largest buyer in the world, especially at the high end of the market. If these travel restrictions continue -- and the OMB directive states this reduced level of spending continue through FY 2016 -- HPC vendors might start to scale back their presence at SC events in the future.
It might create some interesting moments at the event, however. For example, it will be an odd spectacle if the US retains its TOP500 crown, but has nowhere to exhibit the supercomputing technology that put it there. Indeed, the government might save a few million dollars here, but given the international competition that this show highlights, it could end up getting some bad press anyway.
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?