Dhabaleswar K. Panda

A case for high performance computing with virtual machines

Virtual machine (VM) technologies are experiencing a resurgence in both industry and research communities. VMs offer many desirable features such as security, ease of management, OS customization, performance isolation, check-pointing, and migration, which can be very beneficial to the performance and the manageability of high performance computing (HPC) applications. However, very few HPC applications are currently running in a virtualized environment due to the performance overhead of virtualization. Further, using VMs for HPC also introduces additional challenges such as management and distribution of OS images.In this paper we present a case for HPC with virtual machines by introducing a framework which addresses the performance and management overhead associated with VM-based computing. Two key ideas in our design are: Virtual Machine Monitor (VMM) bypass I/O and scalable VM image management. VMM-bypass I/O achieves high communication performance for VMs by exploiting the OS-bypass feature of modern high speed interconnects such as Infini-Band. Scalable VM image management significantly reduces the overhead of distributing and managing VMs in large scale clusters. Our current implementation is based on the Xen VM environment and InfiniBand. However, many of our ideas are readily applicable to other VM environments and high speed interconnects.We carry out detailed analysis on the performance and management overhead of our VM-based HPC framework. Our evaluation shows that HPC applications can achieve almost the same performance as those running in a native, non-virtualized environment. Therefore, our approach holds promise to bring the benefits of VMs to HPC applications with very little degradation in performance.

High performance RDMA-based MPI implementation over InfiniBand

Although InfiniBand Architecture is relatively new in the high performance computing area, it offers many features which help us to improve the performance of communication subsystems. One of these features is Remote Direct Memory Access (RDMA) operations. In this paper, we propose a new design of MPI over InfiniBand which brings the benefit of RDMA to not only large messages, but also small and control messages. We also achieve better scalability by exploiting application communication pattern and combining send/receive operations with RDMA operations. Our RDMA-based MPI implementation currently delivers a latency of 6.8 microseconds for small messages and a peak bandwidth of 871 Million Bytes (831 Mega Bytes) per second. Performance evaluation at the MPI level shows that for small messages, our RDMA-based design can reduce the latency by 24%, increase the bandwidth by over 104%, and reduce the host overhead by up to 22%. For large messages, we improve performance by reducing the time for transferring control messages. We have also shown that our new design is beneficial to MPI collective communication and NAS Parallel Benchmarks.

EMP: Zero-Copy OS-Bypass NIC-Driven Gigabit Ethernet Message Passing

Modern interconnects like Myrinet and Gigabit Ethernet offer Gb/s speeds which has put the onus of reducing the communication latency on messaging software. This has led to the development of OS bypass protocols which removed the kernel from the critical path and hence reduced the end-to-end latency. With the advent of programmable NICs, many aspects of protocol processing can be offloaded from user space to the NIC leaving the host processor to dedicate more cycles to the application. Many host-offload messaging systems exist for Myrinet; however, nothing similar exits for Gigabit Ethernet. In this paper we propose Ethernet Message Passing (EMP), a completely new zero-copy, OS-bypass messaging layer for Gigabit Ethernet on Alteon NICs where the entire protocol processing is done at the NIC. This messaging system delivers very good performance (latency of 23 us, and throughput of 880 Mb/s). To the best of our knowledge, this is the .rst NIC-level implementation of a zero-copy message passing layer for Gigabit Ethernet.

Understanding the Impact of Multi-Core Architecture in Cluster Computing: A Case Study with Intel Dual-Core System

Multi-core processors are growing as a new industry trend as single core processors rapidly reach the physical limits of possible complexity and speed. In the new Top500 supercomputer list, more than 20% processors belong to the multi-core processor family. However, without an in-depth study on application behaviors and trends on multi-core clusters, we might not be able to understand the characteristics of multi-core cluster in a comprehensive manner and hence not be able to get optimal performance. In this paper, we take on these challenges and design a set of experiments to study the impact of multi-core architecture on cluster computing. We choose to use one of the most advanced multi-core servers, Intel Bensley system with Woodcrest processors, as our evaluation platform, and use benchmarks including HPL, NAMD, and NAS as the applications to study. From our message distribution experiments, we find that on an average about 50% messages are transferred through intra-node communication, which is much higher than intuition. This trend indicates that optimizing intra- node communication is as important as optimizing inter- node communication in a multi-core cluster. We also observe that cache and memory contention may be a potential bottleneck in multi-core clusters, and communication middleware and applications should be multi-core aware to alleviate this problem. We demonstrate that multi-core aware algorithm, e.g. data tiling, improves benchmark execution time by up to 70%. We also compare the scalability of a multi-core cluster with that of a single-core cluster and find that the scalability of the multi-core cluster is promising.

High performance virtual machine migration with RDMA over modern interconnects

One of the most useful features provided by virtual machine (VM) technologies is the ability to migrate running OS instances across distinct physical nodes. As a basis for many administration tools in modern clusters and data-centers, VM migration is desired to be extremely efficient to reduce both migration time and performance impact on hosted applications. Currently, most VM environments use the Socket interface and the TCP/IP protocol to transfer VM migration traffic. In this paper, we propose a high performance VM migration design by using RDMA (Remote Direct Memory Access). RDMA is a feature provided by many modern high speed interconnects that are currently being widely deployed in data-centers and clusters. By taking advantage of the low software overhead and the one-sided nature of RDMA, our design significantly improves the efficiency of VM migration. We also contribute a set of micro-benchmarks and application-level benchmark evaluations aimed at evaluating important metrics of VM migration. The evaluations using our prototype implementation over Xen and InfiniBand show that RDMA can drastically reduce the migration overhead: up to 80% on total migration time and up to 77% on application observed downtime.

Efficient collective communication on heterogeneous networks of workstations

Networks of Workstations (NOW) have become an attractive alternative platform for high performance computing. Due to the commodity nature of workstations and interconnects and due to the multiplicity of vendors and platforms, the NOW environments are being gradually redefined as Heterogeneous Networks of Workstations (HNOW) environments. This paper presents a new framework for implementing collective communication operations (as defined by the Message Passing Interface (MPI) standard) efficiently for the emerging HNOW environments. We first classify different types of heterogeneity in HNOW and then focus on one important characteristic: communication capabilities of workstations. Taking this characteristic into account, we propose two new approaches Speed-Partitioned Ordered Chain (SPOC) and Fastest-Node First (FNF) to implement collective communication operations with reduced latency. We also investigate methods for deriving optimal trees for broadcast and multicast operations. Generating such trees is shown to be computationally intensive. It is shown that the FNF approach, in spite of its simplicity, can deliver performance within 1% of the performance of the optimal trees. Finally, these new approaches are compared with the approach used in the MPICH implementation on experimental as well as on simulated testbeds. On a 24-node existing HNOW environment with SGI workstations and ATM interconnection our approaches reduce the latency of broadcast and multicast operations by a factor of up to 3.5 compared to the approach used in the existing MPICH implementation. On a 64-node simulated testbed, our approaches can reduce the latency of broadcast and multicast operations by a factor of up to 4.5. Thus, these results demonstrate that there is significant potential for our approaches to be applied towards designing scalable collective communication libraries for current and future generation HNOW environments.

Virtual machine aware communication libraries for high performance computing

As the size and complexity of modern computing systems keep increasing to meet the demanding requirements of High Performance Computing (HPC) applications, manageability is becoming a critical concern to achieve both high performance and high productivity computing. Meanwhile, virtual machine (VM) technologies have become popular in both industry and academia due to various features designed to ease system management and administration. While a VM-based environment can greatly help manageability on large-scale computing systems, concerns over performance have largely blocked the HPC community from embracing VM technologies. In this paper, we follow three steps to demonstrate the ability to achieve near-native performance in a VM-based environment for HPC. First, we propose Inter-VM Communication (IVC), a VM-aware communication library to support efficient shared memory communication among computing processes on the same physical host, even though they may be in different VMs. This is critical for multi-core systems, especially when individual computing processes are hosted on different VMs to achieve fine-grained control. Second, we design a VM-aware MPI library based on MVAPICH2 (a popular MPI library), called MVAPICH2-ivc, which allows HPC MPI applications to transparently benefit from IVC. Finally, we evaluate MVAPICH2-ivc on clusters featuring multi-core systems and high performance InfiniBand interconnects. Our evaluation demonstrates that MVAPICH2-ivc can improve NAS Parallel Benchmark performance by up to 11% in VM-based environment on eight-core Intel Clover-town systems, where each compute process is in a separate VM. A detailed performance evaluation for up to 128 processes (64 node dual-socket single-core systems) shows only a marginal performance overhead of MVAPICH2-ivc as compared with MVAPICH2 running in a native environment. This study indicates that performance should no longer be a barrier preventing HPC environments from taking advantage of the various features available through VM technologies.

Beyond block I/O: Rethinking traditional storage primitives

Over the last twenty years the interfaces for accessing persistent storage within a computer system have remained essentially unchanged. Simply put, seek, read and write have defined the fundamental operations that can be performed against storage devices. These three interfaces have endured because the devices within storage subsystems have not fundamentally changed since the invention of magnetic disks. Non-volatile (flash) memory (NVM) has recently become a viable enterprise grade storage medium. Initial implementations of NVM storage devices have chosen to export these same disk-based seek/read/write interfaces because they provide compatibility for legacy applications. We propose there is a new class of higher order storage primitives beyond simple block I/O that high performance solid state storage should support. One such primitive, atomic-write, batches multiple I/O operations into a single logical group that will be persisted as a whole or rolled back upon failure. By moving write-atomicity down the stack into the storage device, it is possible to significantly reduce the amount of work required at the application, filesystem, or operating system layers to guarantee the consistency and integrity of data. In this work we provide a proof of concept implementation of atomic-write on a modern solid state device that leverages the underlying log-based flash translation layer (FTL). We present an example of how database management systems can benefit from atomic-write by modifying the MySQL InnoDB transactional storage engine. Using this new atomic-write primitive we are able to increase system throughput by 33%, improve the 90th percentile transaction response time by 20%, and reduce the volume of data written from MySQL to the storage subsystem by as much as 43% on industry standard benchmarks, while maintaining ACID transaction semantics.

High performance RDMA-based design of HDFS over InfiniBand

Hadoop Distributed File System (HDFS) acts as the primary storage of Hadoop and has been adopted by reputed organizations (Facebook, Yahoo! etc.) due to its portability and fault-tolerance. The existing implementation of HDFS uses Javasocket interface for communication which delivers suboptimal performance in terms of latency and throughput. For dataintensive applications, network performance becomes key component as the amount of data being stored and replicated to HDFS increases. In this paper, we present a novel design of HDFS using Remote Direct Memory Access (RDMA) over InfiniBand via JNI interfaces. Experimental results show that, for 5GB HDFS file writes, the new design reduces the communication time by 87% and 30% over 1Gigabit Ethernet (1GigE) and IP-over-InfiniBand (IPoIB), respectively, on QDR platform (32Gbps). For HBase, the Put operation performance is improved by 26% with our design. To the best of our knowledge, this is the first design of HDFS over InfiniBand networks.

Multidestination message passing in wormhole k-ary n-cube networks with base routing conformed paths

This paper proposes multidestination message passing on wormhole k-ary n-cube networks using a new base-routing-conformed-path (BRCP) model. This model allows both unicast (single-destination) and multidestination messages to co-exist in a given network without leading to deadlock. The model is illustrated with several common routing schemes (deterministic, as well as adaptive), and the associated deadlock-freedom properties are analyzed. Using this model, a set of new algorithms for popular collective communication operations, broadcast and multicast, are proposed and evaluated. It is shown that the proposed algorithms can considerably reduce the latency of these operations compared to the Umesh (unicast-based multicast) and the Hamiltonian path-based schemes. A very interesting result that is presented shows that a multicast can be implemented with reduced or near-constant latency as the number of processors participating in the multicast increases beyond a certain number. It is also shown that the BRCP model can take advantage of adaptivity in routing schemes to further reduce the latency of these operations. The multidestination mechanism and the BRCP model establish a new foundation to provide fast and scalable collective communication support on wormhole-routed systems.