NoSQL DB Benchmarking with high performance Networking solutions WBDB, Xian, July 2013 © 2013 Mellanox Technologies 1 Leading Supplier of End-to-End Interconnect Solutions Server / Compute Storage Switch / Gateway Front / Back-End Virtual Protocol Interconnect Virtual Protocol Interconnect 56G IB & FCoIB 56G InfiniBand 10/40/56GbE & FCoE 10/40/56GbE Fibre Channel Comprehensive End-to-End InfiniBand and Ethernet Portfolio ICs © 2013 Mellanox Technologies Adapter Cards Switches/Gateways Host/Fabric Software Cables 2 Motivation to Accelerate Data Analytics Data Analysis Requires Faster Network • Hadoop Map Reduce Framework is a network intensive workload - Mapped data is shuffled between nodes in the cluster • Data Replication - A high availability event triggers Multi-Tera of data movement Provide Higher Data Value • Expose SSD’s low latency capabilities • Better server/CPU utilization Big Data Applications Require High Bandwidth and Low Latency Interconnect * Data Source: Intersect360 Research, 2012, IT and Data scientists survey © 2013 Mellanox Technologies 3 Cassandra, Update Latency Cassandra Database enables update capabilities Latency factors • Commit-log settings • Workload © 2013 Mellanox Technologies 4 Cassandra, Read Latency Cassandra Database Read Latency factors • Media used • Workload © 2013 Mellanox Technologies 5 System Used for Cassandra Benchmark 5 Nodes in the Ring 64GB RAM • 8 x 8GB DDR3 1333MHz 2 x E5-2670 • 8 Cores per socket 5 x Seagate® Constellation® ES SATA 6Gb/s 2TB Hard Drive • 7200 RPM NIC: Mellanox Technologies MT27500 Family [ConnectX-3] • 10Gb Ethernet • FW_VER=2.11.500 Switch SX1036 OS: RH 6.3 • MLNX_OFED_LINUX-1.5.3 Apache Cassandra 1.1.12, 2 seeds © 2013 Mellanox Technologies 6 Unlocking the Power of SSDs In Hadoop Environment SSDs Become De-Facto standard in HDFS deployment • Read capability is a critical factor for application performance E-DFSIO, Part of Intel’s HiBench test suite, profiles aggregated throughput on the cluster • 1GbE network impede any performance benefit from SSD deployment E-DFSIO, Showing the Power of SSD @ HDFS © 2013 Mellanox Technologies 7 HBase Benchmarking, Update Latency Updates are made to server memory • Extreme low latency for HBase - Java GC policy hurting on large throughput © 2013 Mellanox Technologies 8 HBase Benchmarking, Read Latency Hitting the media capabilities © 2013 Mellanox Technologies 9 System Used for HBase Benchmarks 4 Region servers, 1 Master, 3 Zookeeper quorum servers 64GB RAM • 8 x 8GB DDR3 1333MHz 2 x E5-2670 • 8 Cores per socket 5 x Seagate® Constellation® ES SATA 6Gb/s 2TB Hard Drive • 7200 RPM NIC: Mellanox Technologies MT27500 Family [ConnectX-3] • 10Gb Ethernet • FW_VER=2.11.500 Switch SX1036 OS: RH 6.3 • MLNX_OFED_LINUX-1.5.3 Apache Hbase 0.94.9, Zookeeper 3.4.5, Apache Hadoop 1.1.2 © 2013 Mellanox Technologies 10 Test Drive Your Big Data EMC 1000-Node Analytic Platform Accelerates Industry's Hadoop Development 24 PetaByte of physical storage Mellanox VPI Solutions Hadoop Acceleration 2X Faster Hadoop Job Run-Time High Throughput, Low Latency, RDMA Critical for ROI © 2013 Mellanox Technologies 11 The Great Things in Hadoop Distributed File System • • • • HDFS is a block storage solution Block size can be modified to provide efficient solutions for very large files Inherent reliability, no need for high end storage solution to make sure data is there! Tuned for Hadoop work loads, write one and read many © 2013 Mellanox Technologies 12 The Less Great Things in HDFS Metadata Server Failure It’s hard to manage the different setting to get the right nodes into the right capabilities. © 2013 Mellanox Technologies Default 3x Replication Small files or latency sensitive Ingress and extraction of data requires additional tools. 13 Local Disks – The Common Practice © 2013 Mellanox Technologies 14 Other Distributed Storage Solution for Hadoop, Really?! © 2013 Mellanox Technologies 15 OrangeFS as Hadoop Storage Solution © 2013 Mellanox Technologies 16 Lustre as Hadoop Storage Solution Source: Map/Reduce on Lustre, Hadoop Performance in HPC Environments, Nathan Rutman, Senior Architect, Networked Storage Solutions, Xyratex © 2013 Mellanox Technologies 17 CEPH as Hadoop Storage Solution Generating lot of Interest since the Ceph kernel client was pulled into Linux kernel 2.6.34 • • • • • Object-based parallel file system Scalable metadata server Each file can specify it’s own striping strategy and object size Automatic rebalancing of data with minimal data movement Hadoop module for integrating Ceph has been in development since 0.12 release Benchmarks on Ceph is still WIP • We are currently working on using running benchmarks on Ceph – Stay tuned!! © 2013 Mellanox Technologies 18 Thank You © 2013 Mellanox Technologies 19