Program Behavior at Extreme Scale
- Performance tool notes
- ISP
- Applications
- Communication Pattern Detection
- Precise Dynamic Location of Bugs
- ScalaTrace
- Developing Ideas
- Benchmark Suites
- General Analysis Notes
- References
Task Mapping
[17,18,19]
- [19] uses simulated annealing (SA)
- Analysis runtimes on medium scales are several hours for plain SA and 30 min with graph partitioning.
- Uses a communication matrix to do off-line analysis.
- [17] discusses various consistent mapping orders and gray-codes.
- [18] uses graph fitting.
- [17,18] assume a regular, known topology.
Communication Pattern Detection
- Pattern detection according to [1] correlates similar event traces, with some freedom (describe...).
- Well suited from SPMD programs.
- Does not handle communication patterns between heterogeneous processes.
Precise Dynamic Location of Bugs
- A fault caused by a bug may be separated from the bug in time.
- How to determine causality and at what abstraction?
ScalaTrace
[3,5,6,7,10,12]
- Original paper is [12].
- Yields the semantic information of MPI events to compress traces.
- Attains near constant size traces.
- [7] attempts to explain sub-linear and super-linear trace sizes given the run-time and scale.
- [6] extends the compression used in ScalaTrace to include time stamps.
- [6] suggests the use of MRNet for trace collection to offset memory overhead.
- Must evaluate whether compression is good for irregular communication patterns.
Developing Ideas
- On-line analysis.
- Make inferences at run-time.
- Can a meaningful analysis be done locally?
- Impact on compression.
- May reduce the data needed to be stored. May allow new semantic compression.
- ScalaTrace methods will not be directly applicable. Some of the code may be reused.
- Make inferences at run-time.
- Data profiling.
- How to summarize the data exchanged?
- Predictive Performance Evaluation
- Predict performance with a different configuration (different topology, etc.)
- Predictions need not be complete. Reports may include only effects in the near future.
- Grammar inference based model extraction [16]
- Focuses on finite state machine extraction from traces.
- Methods may be applied locally, not to a complete trace.
- Event-based model of behavior [15]
Benchmark Suites
[11]
References
[7] MPI Trace Compression Tuning Project.
[9] An Open Framework for Scalable, Reconfigurable Performance Analysis. (html)
[11] The ASCI Purple Benchmark Codes.
[14] Huband, S. and McDonald, C. 2001. A Preliminary Topological Debugger for MPI Programs.
[15] Peter C. Bates. Debugging heterogeneous distributed systems using event-based models of behavior.
[16] Cook, J. E. and Wolf, A. L. Discovering models of software processes from event-based data.
[17] Yu, H., Chung, I., and Moreira, Topology mapping for Blue Gene/L supercomputer.
[18] Smith, and Bode. Performance Effects of Node Mappings on the IBM BlueGene/L Machine.
[22] Springer and Upchurch. Using SPEEDES to Simulate the Blue Gene Interconnect Network.