Astra, deployed in 2018, was the first petascale supercomputer to utilize processors based on the ARM instruction set. The system was also the first under Sandia's Vanguard program which seeks to provide an evaluation vehicle for novel technologies that with refinement could be utilized in demanding, large-scale HPC environments. In addition to ARM, several other important first-of-a-kind developments were used in the machine, including new approaches to cooling the datacenter and machine. This article documents our experiences building a power measurement and control infrastructure for Astra. While this is often beyond the control of users today, the accurate measurement, cataloging, and evaluation of power, as our experiences show, is critical to the successful deployment of a large-scale platform. While such systems exist in part for other architectures, Astra required new development to support the novel Marvell ThunderX2 processor used in compute nodes. In addition to documenting the measurement of power during system bring up and for subsequent on-going routine use, we present results associated with controlling the power usage of the processor, an area which is becoming of progressively greater interest as data centers and supercomputing sites look to improve compute/energy efficiency and find additional sources for full system optimization.
The Message Passing Interface (MPI) has been the dominant message passing solution for scientific computing for decades. MPI point-to-point communications are highly efficient mechanisms for process-to-process communication. However, MPI performance is slowed by concurrency protections in the MPI library when processes utilize multiple threads. MPI's current thread-level interface imposes these overheads throughout the library when thread safety is needed. While much work has been done to reduce multithreading overheads in MPI, a solution is needed that reduces the number of messages exchanged in a threaded environment. Partitioned communication is included in the MPI 4.0 standard as an alternative that addresses the challenges of multithreaded communication in MPI today. Partitioned communication reduces overall message volume by creating a buffer-sharing mechanism between threads such that they can indicate when portions of a communication buffer are available to be sent. Separation of the control and data planes in MPI is enabled by allowing persistent initialization and single occurrence message buffer matching from the indication that the data is ready to be sent. This enables the usage of underlying hardware primitives like triggered operations, where commands (destination, size, etc.) can be set up prior to data buffer readiness with readiness triggered by a simple doorbell/counter later. This approach is useful for future development of MPI operations in environments where traditional networking commands can have performance challenges, like accelerators (GPUs, FPGAs). In this paper, we detail the design and implementation of a layered library (built on top of MPI-3.1) and an integrated Open MPI solution that supports the new, MPI-4.0 partitioned communication feature set. The library will enable applications to use currently released MPI implementations and older legacy libraries to provide partitioned communication support while also enabling further exploration of this new communication model in new applications and use cases. We will compare the designs of the library and native Open MPI support, provide performance results and comparisons between the two approaches, and lessons learned from the implementation of partitioned communication in both library and native forms. We find that the native implementation and library have similar performance with a percentage difference under 0.94% in microbenchmarks and performance within 5% for a partitioned communication enabled proxy application.
Remote Direct Memory Access (RDMA) capabilities have been provided by high-end networks for many years, but the network environments surrounding RDMA are evolving. RDMA performance has historically relied on using strict ordering guarantees to determine when data transfers complete, but modern adaptively-routed networks no longer provide those guarantees. RDMA also exposes low-level details about memory buffers: either all clients are required to coordinate access using a single shared buffer, or exclusive resources must be allocatable per-client for an unbounded amount of time. This makes RDMA unattractive for use in many-to-one communication models such as those found in public internet client-server situations.Remote Virtual Memory Access (RVMA) is a novel approach to data transfer which adapts and builds upon RDMA to provide better usability, resource management, and fault tolerance. RVMA provides a lightweight completion notification mechanism which addresses RDMA performance penalties imposed by adaptively-routed networks, enabling high-performance data transfer regardless of message ordering. RVMA also provides receiver-side resource management, abstracting away previously-exposed details from the sender-side and removing the RDMA requirement for exclusive/coordinated resources. RVMA requires only small hardware modifications from current designs, provides performance comparable or superior to traditional RDMA networks, and offers many new features.In this paper, we describe RVMA's receiver-managed resource approach and how it enables a variety of new data-transfer approaches on high-end networks. In particular, we demonstrate how an RVMA NIC could implement the first hardware-based fault tolerant RDMA-like solution. We present the design and validation of an RVMA simulation model in a popular simulation suite and use it to evaluate the advantages of RVMA at large scale. In addition to support for adaptive routing and easy programmability, RVMA can outperform RDMA on a 3D sweep application by 4.4X.
Proceedings - 2020 IEEE 22nd International Conference on High Performance Computing and Communications, IEEE 18th International Conference on Smart City and IEEE 6th International Conference on Data Science and Systems, HPCC-SmartCity-DSS 2020
The Message Passing Interface (MPI) standard allows user-level threads to concurrently call into an MPI library. While this feature is currently rarely used, there is considerable interest from developers in adopting it in the near future. There is reason to believe that multithreaded communication may incur additional message processing overheads in terms of number of items searched during demultiplexing and amount of time spent searching because it has the potential to increase the number of messages exchanged and to introduce non-deterministic message ordering. Therefore, understanding the implications of adding multithreading to MPI applications is important for future application development.One strategy for advancing this understanding is through 'low-cost' benchmarks that emulate full communication patterns using fewer resources. For example, while a complete, 'real-world' multithreaded halo exchange requires 9 or 27 nodes, the low-cost alternative needs only two, making it deployable on systems where acquiring resources is difficult because of high utilization (e.g., busy capacity-computing systems), or impossible because the necessary resources do not exist (e.g., testbeds with too few nodes). While such benchmarks have been proposed, the reported results have been limited to a single architecture or derived indirectly through simulation, and no attempt has been made to confirm that a low-cost benchmark accurately captures features of full (non-emulated) exchanges. Moreover, benchmark code has not been made publicly available.The purpose of the study presented in this paper is to quantify how accurately the low-cost benchmark captures the matching behavior of the full, real-world benchmark. In the process, we also advocate for the feasibility and utility of the low-cost benchmark. We present a 'real-world' benchmark implementing a full multithreaded halo exchange on 9 and 27 nodes, as defined by 5-point and 9-point 2D stencils, and 7-point and 27-point 3D stencils. Likewise, we present a 'low-cost' benchmark that emulates these communication patterns using only two nodes. We then confirm, across multiple architectures, that the low-cost benchmark gives accurate estimates of both number of items searched during message processing, and time spent processing those messages. Finally, we demonstrate the utility of the low-cost benchmark by using it to profile the performance impact of state-of-The-Art Mellanox ConnectX-5 hardware support for offloaded MPI message demultiplexing. To facilitate further research on the effects of multithreaded MPI on message matching behavior, the source of our two benchmarks is to be included in the next release version of the Sandia MPI Micro-Benchmark Suite.
Proceedings of IPDRM 2020: 4th Annual Workshop on Emerging Parallel and Distributed Runtime Systems and Middleware, Held in conjunction with SC 2020: The International Conference for High Performance Computing, Networking, Storage and Analysis
As network speeds increase, the overhead of processing incoming messages is becoming onerous enough that many manufacturers now provide network interface cards (NICs) with offload capabilities to handle these overheads. This increase in NIC capabilities creates an opportunity to enable computation on data in-situ on the NIC. These enhanced NICs can be classified into several different categories of SmartNICs. SmartNICs present an interesting opportunity for future runtime software designs. Designing runtime software to be located in the network as opposed to the host level leads to new radical distributed runtime possibilities that were not practical prior to SmartNICs. In the process of transitioning to a radically different runtime software design for SmartNICs there are intermediary steps of migrating current runtime software to be offloaded onto a SmartNIC that also present interesting possibilities. This paper will describe SmartNIC design and how SmartNICs can be leveraged to offload current generation runtime software and lead to future radically different in-network distributed runtime systems.
This paper explores key differences of MPI match lists for several important United States Department of Energy (DOE) applications and proxy applications. This understanding is critical in determining the most promising hardware matching design for any given high-speed network. The results of MPI match list studies for the major open-source MPI implementations, MPICH and Open MPI, are presented, and we modify an MPI simulator, LogGOPSim, to provide match list statistics. These results are discussed in the context of several different potential design approaches to MPI matching–capable hardware. The data illustrate the requirements for different hardware designs in terms of performance and memory capacity. This paper's contributions are the collection and analysis of data to help inform hardware designers of common MPI requirements and highlight the difficulties in determining these requirements by only examining a single MPI implementation.
As we approach exascale, computational parallelism will have to drastically increase in order to meet throughput targets. Many-core architectures have exacerbated this problem by trading reduced clock speeds, core complexity, and computation throughput for increasing parallelism. This presents two major challenges for communication libraries such as MPI: the library must leverage the performance advantages of thread level parallelism and avoid the scalability problems associated with increasing the number of processes to that scale. Hybrid programming models, such as MPI+X, have been proposed to address these challenges. MPI THREAD MULTIPLE is MPI's thread safe mode. While there has been work to optimize it, it largely remains non-performant in most implementations. While current applications avoid MPI multithreading due to performance concerns, it is expected to be utilized in future applications. One of the major synchronous data structures required by MPI is the matching engine. In this paper, we present a parallel matching algorithm that can improve MPI matching for multithreaded applications. We then perform a feasibility study to demonstrate the performance benefit of the technique.
Concurrency and Computation: Practice and Experience
Bernholdt, David E.; Boehm, Swen; Bosilca, George; Gorentla Venkata, Manjunath; Grant, Ryan E.; Naughton, Thomas; Pritchard, Howard P.; Schulz, Martin; Vallee, Geoffroy R.
The Exascale Computing Project (ECP) is currently the primary effort in the United States focused on developing “exascale” levels of computing capabilities, including hardware, software, and applications. In order to obtain a more thorough understanding of how the software projects under the ECP are using, and planning to use the Message Passing Interface (MPI), and help guide the work of our own project within the ECP, we created a survey. Of the 97 ECP projects active at the time the survey was distributed, we received 77 responses, 56 of which reported that their projects were using MPI. This paper reports the results of that survey for the benefit of the broader community of MPI developers.
Current proposals for in-network data processing operate on data as it streams through a network switch or endpoint. Since compute resources must be available when data arrives, these approaches provide deadline-based models of execution. This paper introduces a deadline-free general compute model for network endpoints called INCA: In-Network Compute Assistance. INCA builds upon contemporary NIC offload capabilities to provide on-NIC, deadline-free, general-purpose compute capacities that can be utilized when the network is inactive. We demonstrate INCA is Turing complete, and provide a detailed design for extending existing hardware to support this model. We evaluate runtimes for a selection of kernels, including several optimizations, and show INCA can provide up to a 11% speedup for applications with minimal code modifications and between 25% to 37% when applications are optimized for INCA.