Conference Paper

Instrumentation and Analysis of MPI Queue Times on the SeaStar High-Performance Network

Scable Syst. Software Dept., Sandia Nat. Labs., Albuquerque, NM
DOI: 10.1109/ICCCN.2008.ECP.116 Conference: Computer Communications and Networks, 2008. ICCCN '08. Proceedings of 17th International Conference on
Source: DBLP


Understanding the communication behavior and network resource usage of parallel applications is critical to achieving high performance and scalability on systems with tens of thousands of network endpoints. The need for better understanding is not only driven by the desire to identify potential performance optimization opportunities for current networks, but is also a necessity for designing next-generation networking hardware. In this paper, we describe our approach to instrumenting the SeaStar interconnect on the Cray XT series of massively parallel processing machines to gather low-level network timing data. This data provides a new perspective on performance evaluation, both in terms of evaluating the resource usage patterns of applications as well as evaluating different implementation strategies in the network protocol stack.

14 Reads
  • Source
  • [Show abstract] [Hide abstract]
    ABSTRACT: Inthisworkwepresentapredictiveanalyticalmodelthat encompassestheperformanceandscaling characteristicsofan importantASCIapplication.SAGE(SAIC'sAdaptiveGrid Eulerianhydrocode)isamultidimensionalhydrodynamicscode withadaptivemeshrefinement.Themodelisvalidatedagainst measurementsonseveralsystemsincludingASCIBlueMountain, ASCIWhite,andaCompaqAlphaserverES45systemshowing highaccuracy.Itisparametric - basicmachineperformance numbers(latency,MFLOPSrate,bandwidth)andapplication characteristics(problemsize,decompositionmethod,etc.)serve asinput.Themodelisappliedtoaddinsightintotheperformance ofcurrentsystems,torevealbottlenecks,andtoillustratewhere tuningeffortscanbeeffective.Wealsousethemodeltopredict performanceonfuturesystems. Keywords Performanceanalysis,ful lapplicationcodes,parallel systemarchitecture,Teraflopscalecomputing. 1.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The Seastar, a new ASIC from Cray, is a full system-on-chip design that integrates high-speed serial links, a 3D router, and traditional network interface functionality, including an embedded processor in a single chip. Cray Inc. designed the SeaStar specifically to support Sandia National Laboratories' ASC Red Storm, a distributed-memory parallel computing platform containing more than 11,000 network end-points. SeaStar presented designers with several challenging goals that were commensurate with a high-performance network for a system of that scale. The primary challenge was to provide a well-balanced, highly scalable, highly reliable network. From the Red Storm perspective, a balanced network is one that maximizes network performance relative to the computational power of the network end-points. A main challenge for SeaStar was to maximize the bytes-to-flops ratio of network bandwidth - that is, to maximize the amount of network bandwidth relative to each nodes floating-point capability
    IEEE Micro 06/2006; 26(3-26):41 - 57. DOI:10.1109/MM.2006.65 · 1.52 Impact Factor
Show more

Preview (3 Sources)

14 Reads
Available from