Filters








153 Hits in 3.0 sec

Jdib: Java Applications Interface to Unshackle the Communication Capabilities of InfiniBand Networks

Wan Huang, Hongwei Zhang, Jin He, Jizhong Han, Lisheng Zhang
2007 2007 IFIP International Conference on Network and Parallel Computing Workshops (NPC 2007)  
Using existing TCP/IP emulation layers over InfiniBand networks in Java applications presented rather poor communication performance and heavy overhead in host CPU.  ...  In this paper, we propose Jdib (Java Direct InfiniBand), a Java application acceleration technique to directly exploit RDMA (Remote-Direct-Memory-Access) capability of In-finiBand network hardware.  ...  It has extremely low latency and low CPU overhead.  ... 
doi:10.1109/npc.2007.111 fatcat:faqdr3o5pbecpkeuktktzzijgm

Jdib: Java Applications Interface to Unshackle the Communication Capabilities of InfiniBand Networks

Wan Huang, Hongwei Zhang, Jin He, Jizhong Han, Lisheng Zhang
2007 2007 IFIP International Conference on Network and Parallel Computing Workshops (NPC 2007)  
Using existing TCP/IP emulation layers over InfiniBand networks in Java applications presented rather poor communication performance and heavy overhead in host CPU.  ...  In this paper, we propose Jdib (Java Direct InfiniBand), a Java application acceleration technique to directly exploit RDMA (Remote-Direct-Memory-Access) capability of In-finiBand network hardware.  ...  It has extremely low latency and low CPU overhead.  ... 
doi:10.1109/icnpcw.2007.4351550 fatcat:nqucxb7eszenrg45cnru3mqvbi

Towards Low-Latency Byzantine Agreement Protocols Using RDMA

Signe Rusch, Ines Messadi, Rudiger Kapitza
2018 2018 48th Annual IEEE/IFIP International Conference on Dependable Systems and Networks Workshops (DSN-W)  
We investigate the usage of RDMA in the Reptor BFT protocol for low-latency agreement and show first steps towards an RDMA-enabled consensus protocol.  ...  With the availability of Remote Direct Memory Access (RDMA) in data centers, message exchange latency can be reduced compared to TCP, as RDMA enables kernel bypassing and thereby avoids intermediate data  ...  RDMA enables low-latency communication for applications by reducing the number of intermediate data copies, thereby alleviating the network overhead as well as taking advantage of the offered bandwidth  ... 
doi:10.1109/dsn-w.2018.00054 dblp:conf/dsn/RuschMK18 fatcat:olcgouoit5go7kree62flpyblu

A Benchmark to Evaluate InfiniBand Solutions for Java Applications [article]

Stefan Nothaas, Fabian Ruhland, Michael Schöttner
2019 arXiv   pre-print
Low-latency network interconnects, such as InfiniBand, are commonly used in HPC centers and are even accessible with todays cloud providers offering equipped instances for rent.  ...  In this paper, we present the "Java InfiniBand Benchmark" to evaluate the currently available (and supported) "low-level" solutions to utilize InfiniBand in Java.  ...  INTRODUCTION RDMA capable devices have been providing high throughput and low-latency to HPC applications for several years [18] .  ... 
arXiv:1910.02245v1 fatcat:5her5jnf4fbgzafydilptylhiu

High-Performance Design of HBase with RDMA over InfiniBand

Jian Huang, Xiangyong Ouyang, Jithin Jose, Md. Wasi-ur-Rahman, Hao Wang, Miao Luo, Hari Subramoni, Chet Murthy, Dhabaleswar K. Panda
2012 2012 IEEE 26th International Parallel and Distributed Processing Symposium  
These interconnects provide advanced network features, such as Remote Direct Memory Access (RDMA), to achieve high throughput and low latency along with low CPU utilization.  ...  The High Performance Computing (HPC) domain has exploited high performance and low latency networks such as InfiniBand for many years.  ...  In this paper, we enabled RDMA communication channel in a Java based system. VII.  ... 
doi:10.1109/ipdps.2012.74 dblp:conf/ipps/HuangOJWWLSMP12 fatcat:owbotqyffndlrebgbrscn5kepm

Data centric computing for internet scale enterprises

Yuqing Gao
2013 Proceedings of the ACM/SPEC international conference on International conference on performance engineering - ICPE '13  
 Global secondary index for HBase (Hadoop Database)  RDMA Enabled Cache  Polyglot runtime systems & challenges for benchmarking community Big Data is about analyzing data at scale along the dimensions  ...  systems  Global secondary index for HBase (Hadoop Database)  RDMA Enabled Cache  Polyglot runtime systems & challenges for benchmarking community Categories of NoSQL Use-case Patterns ■ Rapid development  ...  Peers directly read and write the data at the end of the pointers using RDMA Remote pointers & one-sided RDMA  A server can export a remote pointer that refers to a data record in a pinned & registered  ... 
doi:10.1145/2479871.2479924 dblp:conf/wosp/Gao13 fatcat:74ooahyt5rannjuywj3nv3zwoq

High performance RDMA-based design of HDFS over InfiniBand

N. S. Islam, M. W. Rahman, J. Jose, R. Rajachandrasekar, H. Wang, H. Subramoni, C. Murthy, D. K. Panda
2012 2012 International Conference for High Performance Computing, Networking, Storage and Analysis  
The existing implementation of HDFS uses Javasocket interface for communication which delivers suboptimal performance in terms of latency and throughput.  ...  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  ...  Implementations of MPI, such as MVAPICH2 [5] , achieve low one-way latencies in the range of 1-2µs.  ... 
doi:10.1109/sc.2012.65 dblp:conf/sc/IslamWJRWSMP12 fatcat:z3qahxdfdjcsrd73uikhlzhcya

A study of hardware assisted IP over InfiniBand and its impact on enterprise data center performance

Ryan E. Grant, Pavan Balaji, Ahmad Afsahi
2010 2010 IEEE International Symposium on Performance Analysis of Systems & Software (ISPASS)  
As such high performance socket networks are likely to be deployed or converged with existing Ethernet networking solutions, the performance of such technologies is important to assess.  ...  SDP has excellent performance but limited utility as only applications relying on the TCP/IP sockets API can use it and other IP stack uses (IPSec, UDP, SCTP) or TCP layer modifications (iSCSI) cannot  ...  Other work on the Java communication stack has been done in [10] , where RDMA operation functionality was added for Java through a specialized library.  ... 
doi:10.1109/ispass.2010.5452035 dblp:conf/ispass/GrantBA10 fatcat:xw6gj2fayneuzlmocpuxiv5jfq

vRead

Cong Xu, Brendan Saltaformaggio, Sahan Gamage, Ramana Rao Kompella, Dongyan Xu
2015 Proceedings of the 16th Annual Middleware Conference on - Middleware '15  
By doing so, vRead can significantly avoid device virtualization overhead, resulting in improved I/O throughput as well as CPU savings for Hadoop workloads and other applications relying on HDFS.  ...  In fact, data read from an HDFS datanode VM must go through virtual devices multiple times -incurring non-negligible virtualization overhead -even though both client VM and datanode VM may be running on  ...  To connect to remote vRead daemons on other machines with low latency and low CPU cost, we use RDMA interfaces (declared in rdma/rdma_cma.h and infiniband/arch.h) instead of TCP/IP APIs to exchange data  ... 
doi:10.1145/2814576.2814735 dblp:conf/middleware/XuSGKX15 fatcat:3aj3d3ei6bamzefnljbmiabufq

Scalable RDMA performance in PGAS languages

M. Farreras, G. Almasi, C. Cascaval, T. Cortes
2009 2009 IEEE International Symposium on Parallel & Distributed Processing  
threads competing for the same network device any improvement in network device access time is magnified fourfold.  ...  In hybrid execution mode the network device is shared by all UPC threads running on a blade; the improvement caused by smaller CPU overheads was augmented by smaller network device overheads; with four  ... 
doi:10.1109/ipdps.2009.5161025 dblp:conf/ipps/FarrerasACC09 fatcat:pufvpxgwdza5ho2pgvbohiolfe

FastMPJ: a scalable and efficient Java message-passing library

Roberto R. Expósito, Sabela Ramos, Guillermo L. Taboada, Juan Touriño, Ramón Doallo
2014 Cluster Computing  
advantage of high-speed cluster networks (e.g., InfiniBand) to provide low-latency and high bandwidth communications; (3) including a scalable collective library with topology aware primitives, automatically  ...  ., productivity, portability, multithreading) of Java for parallel programming, its poor communications support has hindered its adoption in the HPC community.  ...  The observed point-to-point communication efficiency of xxdev devices allows FastMPJ to provide low-latency and high-bandwidth communications for MPJ parallel applications, both on high-speed networks  ... 
doi:10.1007/s10586-014-0345-4 fatcat:jxpqqaj3frbrfendjhmims72ty

Peak performance

Hannes Mühleisen, Romulo Gonçalves, Martin Kersten
2013 Proceedings of the Ninth International Workshop on Data Management on New Hardware - DaMoN '13  
The recent broad availability of the zero-copy data transfer protocol RDMA over low-latency and highthroughput network connections such as InfiniBand prompts us to revisit the long-proposed usage of memory  ...  In this paper, we present a solution to make use of remote memory without manipulation of the operating system, and investigate the impact on database performance.  ...  the "standard system tools" approach and Sjoerd Mullender for his work on MonetDB internals.  ... 
doi:10.1145/2485278.2485287 dblp:conf/damon/MuhleisenGK13 fatcat:piuwm6idovcjndlvuvgu23r2oa

Java Fast Sockets: Enabling high-speed Java communications on high performance clusters

Guillermo L. Taboada, Juan Touriño, Ramón Doallo
2008 Computer Communications  
(DSM) middleware on clusters and (4) low-level libraries on high-speed networks.  ...  JFS overcomes these performance constraints by: (1) enabling high-speed communication on cluster networks such as Scalable Coherent Interface (SCI), Myrinet and Gigabit Ethernet; (2) avoiding the need  ...  For instance, Jdib [25] accesses Mellanox Verbs Interface (VAPI) on Infiniband through a low-level API which directly exploits RDMA and communication queues.  ... 
doi:10.1016/j.comcom.2008.08.012 fatcat:jtribpsapzdklcw2fkmvnto3qy

Hardware Trends

Tim Harris
2015 ACM SIGACT News  
., computation versus network communication), the reliability of operations (e.g., that memory accesses are reliable, but network communication is not), and even some of the basic properties of the system  ...  (e.g., that the contents of main memory are lost on power failure).  ...  Some network devices provide low-latency hardware distribution of data to multiple machines at once (for instance, the ExaLINK matrix switches advertise 5ns latency multicasting data from an input port  ... 
doi:10.1145/2789149.2789165 fatcat:77jnr3wpkrcctegvqqg24xnyhy

Design and evaluation of benchmarks for financial applications using Advanced Message Queuing Protocol (AMQP) over InfiniBand

Hari Subramoni, Gregory Marsh, Sundeep Narravula, Ping Lai, Dhabaleswar K. Panda
2008 2008 Workshop on High Performance Computational Finance  
, which place less processing load on the host.  ...  Our results indicate that in order to achieve the high scalability requirements demanded by high performance computational finance applications, we need to use modern communication protocols, like RDMA  ...  This coupled with the fact that the communication using RDMA incurs very low overhead on the host CPU makes this an ideal choice for the underlying protocol. Figure 4.  ... 
doi:10.1109/whpcf.2008.4745404 fatcat:yv4vge4w3bgzbpjsulcpczefeu
« Previous Showing results 1 — 15 out of 153 results