Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userAccurate runtime selection of optimal MPI collective algorithms using analytical performance modelling
66
0
0.0
(
0
)
Added by
Alexey Lastovetsky
Publication date
2020
fields
Informatics Engineering
and research's language is
English
Ask ChatGPT about the research
No Arabic abstract
The performance of collective operations has been a critical issue since the advent of MPI. Many algorithms have been proposed for each MPI collective operation but none of them proved optimal in all situations. Different algorithms demonstrate superior performance depending on the platform, the message size, the number of processes, etc. MPI implementations perform the selection of the collective algorithm empirically, executing a simple runtime decision function. While efficient, this approach does not guarantee the optimal selection. As a more accurate but equally efficient alternative, the use of analytical performance models of collective algorithms for the selection process was proposed and studied. Unfortunately, the previous attempts in this direction have not been successful. We revisit the analytical model-based approach and propose two innovations that significantly improve the selective accuracy of analytical models: (1) We derive analytical models from the code implementing the algorithms rather than from their high-level mathematical definitions. This results in more detailed models. (2) We estimate model parameters separately for each collective algorithm and include the execution of this algorithm in the corresponding communication experiment. We experimentally demonstrate the accuracy and efficiency of our approach using Open MPI broadcast and gather algorithms and a Grid5000 cluster.
rate research
Read More
Collective communication algorithms are an important component of distributed computation. Indeed, in the case of deep-learning, collective communication is the Amdahls bottleneck of data-parallel training. This paper introduces SCCL (for Synthesized Collective Communication Library), a systematic approach to synthesize collective communication algorithms that are explicitly tailored to a particular hardware topology. SCCL synthesizes algorithms along the Pareto-frontier spanning from latency-optimal to bandwidth-optimal implementations of a collective. The paper demonstrates how to encode SCCLs synthesis as a quantifier-free SMT formula which can be discharged to a theorem prover. We further demonstrate how to scale our synthesis by exploiting symmetries in topologies and collectives. We synthesize and introduce novel latency and bandwidth optimal algorithms not seen in the literature on two popular hardware topologies. We also show how SCCL efficiently lowers algorithms to implementations on two hardware architectures (NVIDIA and AMD) and demonstrate competitive performance with hand optimized collective communication libraries.
We present Task Bench, a parameterized benchmark designed to explore the performance of parallel and distributed programming systems under a variety of application scenarios. Task Bench lowers the barrier to benchmarking multiple programming systems by making the implementation for a given system orthogonal to the benchmarks themselves: every benchmark constructed with Task Bench runs on every Task Bench implementation. Furthermore, Task Benchs parameterization enables a wide variety of benchmark scenarios that distill the key characteristics of larger applications. We conduct a comprehensive study with implementations of Task Bench in 15 programming systems on up to 256 Haswell nodes of the Cori supercomputer. We introduce a novel metric, minimum effective task granularity to study the baseline runtime overhead of each system. We show that when running at scale, 100 {mu}s is the smallest granularity that even the most efficient systems can reliably support with current technologies. We also study each systems scalability, ability to hide communication and mitigate load imbalance.
With the increasing number of Quad-Core-based clusters and the introduction of compute nodes designed with large memory capacity shared by multiple cores, new problems related to scalability arise. In this paper, we analyze the overall performance of a cluster built with nodes having a dual Quad-Core Processor on each node. Some benchmark results are presented and some observations are mentioned when handling such processors on a benchmark test. A Quad-Core-based clusters complexity arises from the fact that both local communication and network communications between the running processes need to be addressed. The potentials of an MPI-OpenMP approach are pinpointed because of its reduced communication overhead. At the end, we come to a conclusion that an MPI-OpenMP solution should be considered in such clusters since optimizing network communications between nodes is as important as optimizing local communications between processors in a multi-core cluster.
Scaling supercomputers comes with an increase in failure rates due to the increasing number of hardware components. In standard practice, applications are made resilient through checkpointing data and restarting execution after a failure occurs to resume from the latest check-point. However, re-deploying an application incurs overhead by tearing down and re-instating execution, and possibly limiting checkpointing retrieval from slow permanent storage. In this paper we present Reinit++, a new design and implementation of the Reinit approach for global-restart recovery, which avoids application re-deployment. We extensively evaluate Reinit++ contrasted with the leading MPI fault-tolerance approach of ULFM, implementing global-restart recovery, and the typical practice of restarting an application to derive new insight on performance. Experimentation with three different HPC proxy applications made resilient to withstand process and node failures shows that Reinit++ recovers much faster than restarting, up to 6x, or ULFM, up to 3x, and that it scales excellently as the number of MPI processes grows.
Transparently checkpointing MPI for fault tolerance and load balancing is a long-standing problem in HPC. The problem has been complicated by the need to provide checkpoint-restart services for all combinations of an MPI implementation over all network interconnects. This work presents MANA (MPI-Agnostic Network-Agnostic transparent checkpointing), a single code base which supports all MPI implementation and interconnect combinations. The agnostic properties imply that one can checkpoint an MPI application under one MPI implementation and perhaps over TCP, and then restart under a second MPI implementation over InfiniBand on a cluster with a different number of CPU cores per node. This technique is based on a novel split-process approach, which enables two separate programs to co-exist within a single process with a single address space. This work overcomes the limitations of the two most widely adopted transparent checkpointing solutions, BLCR and DMTCP/InfiniBand, which require separate modifications to each MPI implementation and/or underlying network API. The runtime overhead is found to be insignificant both for checkpoint-restart within a single host, and when comparing a local MPI computation that was migrated to a remote cluster against an ordinary MPI computation running natively on that same remote cluster.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
