1. CBHPC’08: Component-Based High Performance Computing (16/10/08) 1 A GCM-Based Runtime Support for Parallel Grid Applications Elton Mathias, Françoise Baude and Vincent Cave {Elton.Mathias, Francoise.Baude, Vincent.Cave}@inria.fr CBHPC’08: Component-Based High Performance Computing Karlsruhe - October 16th, 2008

2. CBHPC’08: Component-Based High Performance Computing (16/10/08) 2 Outline ● Introduction ● Related works and positioning ● Context – The ProActive Middleware – The Grid Component Model (GCM) ● Extensions to GCM ● The DiscoGrid Project and Runtime ● Evaluation ● Conclusion and perspectives

3. CBHPC’08: Component-Based High Performance Computing (16/10/08) 3 Introduction ● Strong trend to integrate parallel resources into grids ● Non-embarrassingly parallel applications must deal with heterogeneity, scalability and performance (+legacy applications) ● Message Passing (MPI!) is established as the main paradigm to develop scientific apps. – Asynchronism and group communications at application level – Applications must adapt to cope with a changing environment ● DiscoGrid project intends to solve this issues by offering a more high-level API, with treatment of these issues at runtime level

4. CBHPC’08: Component-Based High Performance Computing (16/10/08) 4 Related works and positioning ● Grid-oriented MPI – optimizations at comm. layer – unmodified apps. – strong code coupling – Ex: GridMPI, MPICH-G2, PACX-MPI, MagPIE ● Code Coupling - use of components - standalone apps. - weak code coupling - simplified communication – Ex.: DCA, Xchange mxn, Seine Approach: Code Coupling, but with advanced support to multipoint interactions (fine grained tightly component-based code coupling) DiscoGrid Project / Runtime: MPI boosted with advanced collective operations supported by a flexible component-based runtime that provides inter-cluster communication Approach: Code Coupling, but with advanced support to multipoint interactions (fine grained tightly component-based code coupling) DiscoGrid Project / Runtime: MPI boosted with advanced collective operations supported by a flexible component-based runtime that provides inter-cluster communication

5. CBHPC’08: Component-Based High Performance Computing (16/10/08) 5 ProActive Middleware ● Grid middleware for parallel, distributed and multi-threaded computing ● Featuring: – Deployment with support to several network protocols and cluster/grid tools – Reference implementation of the GCM – Legacy code wrapping ● C Java communication ● Deployment and control of legacy MPI application

6. CBHPC’08: Component-Based High Performance Computing (16/10/08) 6 Grid Component Model (GCM) ● Defined in the context of the Institute on Programming Models of CoreGRID Network of Excellence (EU project) ● Extension to the Fractal comp. model adressing key grid problematics: programmability, interoperability, code reuse and efficiency ● Main characteristics: – Hierarchical component model ● primitive and composite components – Collective interfaces

7. CBHPC’08: Component-Based High Performance Computing (16/10/08) 7 ProActive/GCM standard interfaces Collective interfaces are complementary: gathercast (many-to-one):synchronization parameter gatherind and result dispatch multicast (one-to-many): parallel invocation, param. dispatch and result gather Collective interfaces are complementary: gathercast (many-to-one):synchronization parameter gatherind and result dispatch multicast (one-to-many): parallel invocation, param. dispatch and result gather Standard collective interfaces are enough to support broadcast, scatter, gather and barriers But are not general enough to define many-to- many operation (MxN) Standard collective interfaces are enough to support broadcast, scatter, gather and barriers But are not general enough to define many-to- many operation (MxN)

8. CBHPC’08: Component-Based High Performance Computing (16/10/08) 8 Extending GCM collective interfaces: gather-multicast itfs ● server gather-mcast – exposes gcast itf – connected do internal components ● client gather-mcast – connects internal comp. – exposes mcast itf ● Communication semantic relies on 2 policies: gather policy dispatch policy Naïve gather-multicast MxN leads to bottlenecks in both communication policies Naïve gather-multicast MxN leads to bottlenecks in both communication policies

9. CBHPC’08: Component-Based High Performance Computing (16/10/08) 9 Gather-Multicast Optimization ● Efficient communication requires direct bindings ● Solution: controllers responsible for establishing direct MxN bindings + distribution of the comm. policies ● Configured along 3 operations: – (re) binding configuration (R) – dispatch policy (D) – gather policy (G) ● For now, these operations must be coded by developers

10. CBHPC’08: Component-Based High Performance Computing (16/10/08) 10 The DiscoGrid Project ● Promotes a new paradigm to develop non-embarrassingly parallel grid applications ● Target applications: – domain decomposition, requiring solution of PDEs – Ex.: electromagnetic wave propagation, fluid flow pbs ● DiscoGrid API – Resources seen as a hierarchical organization – Hierarchical identifiers – Neighborhood-based communication (update) – C/C++ and Fortran bindings ● DiscoGrid Runtime - Grid-aware partitioner – Modeling resources hierarchy – Support the DG API – converging to MPI when possible – Support to inter-cluster comm.

11. CBHPC’08: Component-Based High Performance Computing (16/10/08) 11 ProActive/GCM DiscoGrid Runtime

12. CBHPC’08: Component-Based High Performance Computing (16/10/08) 12 Optimizing the “update” operation DGOptimizationController.optimize (AggregationMode, DispatchMode, DGNeighborhood, … )

13. CBHPC’08: Component-Based High Performance Computing (16/10/08) 13 DiscoGrid Communication ● Point-to-point Communications ● Collective Communication – Bcast, gather, scatter ● Neighborhood Based – update DG Runtime convert calls to DG API into MPI or DG calls

14. CBHPC’08: Component-Based High Performance Computing (16/10/08) 14 Evaluation ● Conducted in the Grid5000 ● 8 sites (sophia, rennes, grenoble, lyon, bordeaux toulouse, lille) ● Machines with different processors (Intel Xeon EM64T 3GHz and IA32 2.4GHz, AMD Opterons 218, 246, 248 and 285) ● Memory of 2 or 4GB/node ● 2 clusters Myrinet-10G and 2000, backbone 2.5Gb/s ● ProActive 3.9, MPICH 1.2.7p1, Java SDK 1.6.0_02

15. CBHPC’08: Component-Based High Performance Computing (16/10/08) 15 Experiment: P3D ● Poisson3D equation discretized by finite differences and iterative resolution by Jacobi ● Bulk-synchronous behavior: – Concurrent computation ● jacobi over subdomain – Reduction of results ● reduce operation – Update of subdomain borders ● update operation ● Cubic mesh of 1024 3 elements (4GB of data) ● 100 iterations of the algorithm ● 2 versions: legacy version (pure MPI), DG version

16. CBHPC’08: Component-Based High Performance Computing (16/10/08) 16 Experiment: P3D (cont.) The entire DG communication is asynchronous DG update is faster ReduceAll in the DG version happens in parallel P3D Execution Time P3D update Time Time reduce as data/node reduces Simple gather-mcast interface is not scalable The update with the neighborhood happens in parallel

17. CBHPC’08: Component-Based High Performance Computing (16/10/08) 17 Conclusion ● Extensions to GCM provide many-to-many (MxN) communication – versatile mechanism ● any kind of communication ● even with limited connectivity – optimizations ensure efficiency and scalability ● The goal is not compete with MPI, but experimental results showed a good performance ● The DiscoGrid Runtime itself can be considered a successful component-based grid programming approach supporting an SPMD model ● The API and Runtime also permitted a more high-level approach to develop non-embarrassingly applications, where group communication/synchronization are handled as non-functional aspects

18. CBHPC’08: Component-Based High Performance Computing (16/10/08) 18 Perspectives ● Better evaluate the work through more complex simulations (BHE, CEM) and real-size data meshes ● Evaluate the work done in comparison to grid-oriented versions of MPI ● Explore deeper: – the separation of concerns in component architectures: consider SPMD parallel programming as a component configuration and (re)assembling activity instead of message passing. – adaptation of applications to contexts – the definition and usage of collective interfaces (GridComp)

19. CBHPC’08: Component-Based High Performance Computing (16/10/08) 19 Questions

20. CBHPC’08: Component-Based High Performance Computing (16/10/08) 20 Gather-Multicast μBenchmarks - factor = no of neighbors, spread throughout the sites (128 = All-to-All) - message size is constant independent on the factor - results are based on the average time to send the message and receive ack 0 MB (void) message 10 MB message for void messages, optimized version is slower for all-to-all void messages due to the nb of calls (8 2 +16*2 vs 128 2 ) optimized version is considerably faster when size increases because of reduction of bottlenecks Considering that all-to-all messages are avoided, we can expect a real benefit from the optimization