Technical Computing Suite Job Management Software

Size: px

Start display at page:

Download "Technical Computing Suite Job Management Software"

Reynard Jackson
8 years ago
Views:

1 Technical Computing Suite Job Management Software Toshiaki Mikamo Fujitsu Limited Supercomputer PRIMEHPC FX10 PRIMERGY x86 cluster

2 Outline System Configuration and Software Stack Features The major functions of job scheduler Efficient Resource Usage Fair Share Scheduling System-optimal Resource Assignment Summary and Future 1

3 Hybrid System Configuration Supercomputer PRIMEHPC FX10 PRIMERGY x86 cluster 6D mesh/torus Interconnect (Tofu) Fat-Tree Interconnect (Infiniband) Local file system (Temporary area occupied by jobs) IO network (IB), management network (GbE) Management nodes User Login nodes Login Compilation Job submission Global file system (Data storage area) Job management nodes File management nodes System operations management Job operations management Control nodes Administrator 2

4 System Software Stack System operations management System configuration management System control System monitoring System installation & operation Job operations management Job manager Job scheduler Resource management Parallel execution environment User/ISV Applications HPC Portal / System Management Portal Technical Computing Suite High-performance file system Lustre-based distributed file system High scalability IO bandwidth guarantee High reliability & availability VISIMPACT TM Shared L2 cache on a chip Hardware intra-processor synchronization Compilers Hybrid parallel programming Sector cache support SIMD / Register file extensions Support Tools IDE Profiler & Tuning tools Interactive debugger MPI Library Scalability of High-Func. Barrier Comm. Linux-based enhanced Operating System Supercomputer PRIMEHPC FX10 3 Red Hat Enterprise Linux PRIMERGY x86 cluster

5 Features Same job operations in FX10 and PRIMERGY Efficient, fair and system-optimal job scheduling See slide below for details Resource / Access control Elapsed time limit / CPU time limit / Physical memory limit Enable / Disable execute permission of job operation commands Reduce OS jitter / Power saving control Job statistical information The amount of CPU time / Memory / IO SIMD rate / MIPS / MFLOPS 4

6 Job Scheduler Renew our job scheduler for large-scale system Our job scheduler features: Multi-process enable to coexist multiple scheduler in a cluster. Multi-thread enable to balance the load of scheduling. 5

7 Efficient Resource Usage Backfill scheduling for keeping the resources busy Our scheduler manages space(compute nodes) and time. It will backfill the low priority jobs so as not to prevent high priority jobs. Time Now t1 t2 t3 Not backfilled Running job Job C Job B Job D Job C Backfilled Running job Job B Job D Job D Job C 6

8 Fair Share Scheduling Fairly share resources between users/groups based on past usage. 1 Fair share value is issued in advance for each user/group. 2 The value is changed by the result of resource usage. 3 The job execution priority is determined dynamically according to the value. Fair share value (money) Payment time Fair share value is like money. Return of overpaid Deposit Payment[P] = (#Node allocated) x (Elapsed time limit of job) Deposit[D] = (Elapsed time) x (Recovery rate) Return of overpaid[r] = P - ((#Node allocated) x (Actual elapsed time of job)) 7

9 Optimal Job Scheduling for FX10 Interconnect topology-aware resource assignment One interconnect unit : 12 nodes (2 x 3 x 2) Job assignment rule: rectangular solid shape Guaranteeing neighbor communication Avoiding interfering with other jobs Rotates rectangular solid of interconnect unit to reduce fragmentation In-use unoccupied 6 z y x

10 Optimal Job Scheduling for FX10 Asynchronous file staging Compute nodes PRIMEHPC FX10 Interconnect IO nodes Stage IN/OUT Local file system Stage IN Asynchronously transfer files from Global to Local FS before the job starts. Stage OUT Asynchronously transfer files from Local to Global FS after the job ends. Compute nodes Time Now t1 t2 t3 Running job Async. Async. Job B Job C IO network (IB), management network (GbE) IO nodes Stage IN Stage IN Stage OUT Stage IN Stage OUT Login nodes Global file system (Data storage area) Co-scheduling of computation and file transfer. 9

11 Optimal Job Scheduling for PRIMERGY Fine-grained node assignment Node selection method : balancing / concentration Rank placement policy : pack / unpack Priority control of allocated nodes Execution mode : node is occupied or not by a job. Strict core assignment Node#0 Node#1 Node#2 Job C Job B Job D Node concentration Node#0 Node#1 Node#2 R0 R0 Job B R1 R1 Rank pack Node Rank unpack Processes are bound to cores in the job territory No process can move to cores in other job territory. 1 3 P 5 Job B 7 core 10

12 Summary and Future We developed the job management software. Unified operability on PRIMEHPC FX10 and PRIMERGY New job scheduler : Efficiency, Fairness and System-optimization Practical resource control and job statistical information Future Work Operation simulator Administrator will be able to simulate the operation situation subsequent to operation parameter changes. 11

13 12

Operating System for the K computer

Operating System for the K computer Jun Moroo Masahiko Yamada Takeharu Kato For the K computer to achieve the world s highest performance, Fujitsu has worked on the following three performance improvements