Evaluation of Objectivity/AMS on the Wide Area Network

Slides:



Advertisements
Similar presentations
Martin Suchara, Ryan Witt, Bartek Wydrowski California Institute of Technology Pasadena, U.S.A. TCP MaxNet Implementation and Experiments on the WAN in.
Advertisements

All Rights Reserved, Copyright(C) 2007, Hitachi, Ltd. 1 Transport-layer optimization for thin-client systems Yukio OGAWA Systems Development Laboratory,
TCP transfers over high latency/bandwidth network & Grid TCP Sylvain Ravot
LOGO Transmission Control Protocol 12 (TCP) Data Flow.
The Effects of Wide-Area Conditions on WWW Server Performance Erich Nahum, Marcel Rosu, Srini Seshan, Jussara Almeida IBM T.J. Watson Research Center,
Web Server Benchmarking Using the Internet Protocol Traffic and Network Emulator Carey Williamson, Rob Simmonds, Martin Arlitt et al. University of Calgary.
Fundamentals of Computer Networks ECE 478/578
1 End to End Bandwidth Estimation in TCP to improve Wireless Link Utilization S. Mascolo, A.Grieco, G.Pau, M.Gerla, C.Casetti Presented by Abhijit Pandey.
Ahmed El-Hassany CISC856: CISC 856 TCP/IP and Upper Layer Protocols Slides adopted from: Injong Rhee, Lisong Xu.
Copyright © 2005 Department of Computer Science 1 Solving the TCP-incast Problem with Application-Level Scheduling Maxim Podlesny, University of Waterloo.
Advanced Computer Networking Congestion Control for High Bandwidth-Delay Product Environments (XCP Algorithm) 1.
Vijay Vasudevan, Amar Phanishayee, Hiral Shah, Elie Krevat David Andersen, Greg Ganger, Garth Gibson, Brian Mueller* Carnegie Mellon University, *Panasas.
Congestion Control Tanenbaum 5.3, /12/2015Congestion Control (A Loss Based Technique: TCP)2 What? Why? Congestion occurs when –there is no reservation.
CS215 TCP Westwood Control Model Development and Stability Analysis Hu, Kunzhong Dong, Haibo Mentor: Wang, Ren Professor:
High-performance bulk data transfers with TCP Matei Ripeanu University of Chicago.
Network Traffic Measurement and Modeling CSCI 780, Fall 2005.
On Efficient On-line Grouping of Flows with Shared Bottlenecks at Loaded Servers by O. Younis and S. Fahmy Department of Computer Sciences, Purdue University.
Reduced TCP Window Size for VoIP in Legacy LAN Environments Nikolaus Färber, Bernd Girod, Balaji Prabhakar.
Analysis of Active Queue Management Jae Chung and Mark Claypool Computer Science Department Worcester Polytechnic Institute Worcester, Massachusetts, USA.
Copyright © 2005 Department of Computer Science CPSC 641 Winter Tutorial: TCP 101 The Transmission Control Protocol (TCP) is the protocol that sends.
TCP Behavior across Multihop Wireless Networks and the Wired Internet Kaixin Xu, Sang Bae, Mario Gerla, Sungwook Lee Computer Science Department University.
Introduction 1 Lecture 14 Transport Layer (Congestion Control) slides are modified from J. Kurose & K. Ross University of Nevada – Reno Computer Science.
KEK Network Qi Fazhi KEK SW L2/L3 Switch for outside connections Central L2/L3 Switch A Netscreen Firewall Super Sinet Router 10GbE 2 x GbE IDS.
1 A Basic R&D for an Analysis Framework Distributed on Wide Area Network Hiroshi Sakamoto International Center for Elementary Particle Physics (ICEPP),
Transport Layer 4 2: Transport Layer 4.
Experiences in Design and Implementation of a High Performance Transport Protocol Yunhong Gu, Xinwei Hong, and Robert L. Grossman National Center for Data.
TCP Throughput Collapse in Cluster-based Storage Systems
Raj Jain The Ohio State University R1: Performance Analysis of TCP Enhancements for WWW Traffic using UBR+ with Limited Buffers over Satellite.
Maximizing End-to-End Network Performance Thomas Hacker University of Michigan October 26, 2001.
POSTECH DP&NM Lab. Internet Traffic Monitoring and Analysis: Methods and Applications (1) 4. Active Monitoring Techniques.
Principles of Congestion Control Congestion: informally: “too many sources sending too much data too fast for network to handle” different from flow control!
TCOM 509 – Internet Protocols (TCP/IP) Lecture 04_b Transport Protocols - TCP Instructor: Dr. Li-Chuan Chen Date: 09/22/2003 Based in part upon slides.
TCP Transport Control Protocol Information management 2 Groep T Leuven – Information department 2/35 Introduction UDP provides the connection.
UDT: UDP based Data Transfer Yunhong Gu & Robert Grossman Laboratory for Advanced Computing University of Illinois at Chicago.
Chapter 12 Transmission Control Protocol (TCP)
UDT as an Alternative Transport Protocol for GridFTP Raj Kettimuthu Argonne National Laboratory The University of Chicago.
High-speed TCP  FAST TCP: motivation, architecture, algorithms, performance (by Cheng Jin, David X. Wei and Steven H. Low)  Modifying TCP's Congestion.
Rate Control Rate control tunes the packet sending rate. No more than one packet can be sent during each packet sending period. Additive Increase: Every.
Transport Layer3-1 TCP throughput r What’s the average throughout of TCP as a function of window size and RTT? m Ignore slow start r Let W be the window.
Performance Engineering E2EpiPEs and FastTCP Internet2 member meeting - Indianapolis World Telecom Geneva October 15, 2003
CS2910 Week 9, Class 1 Today –Quiz on SMTP –TCP: Flow control vs. Congestion Control –TCP startup, error recovery, and shut-down Before 8am Thursday: –Watch.
The Macroscopic behavior of the TCP Congestion Avoidance Algorithm.
Prentice HallHigh Performance TCP/IP Networking, Hassan-Jain Chapter 13 TCP Implementation.
TCP transfers over high latency/bandwidth networks & Grid DT Measurements session PFLDnet February 3- 4, 2003 CERN, Geneva, Switzerland Sylvain Ravot
EDC Intenet2 AmericaView TAH Oct 28, 2002 AlaskaView Proof-of-Concept Test Tom Heinrichs, UAF/GINA/ION Grant Mah USGS/EDC Mike Rechtenbaugh USGS/EDC Jeff.
Final EU Review - 24/03/2004 DataTAG is a project funded by the European Commission under contract IST Richard Hughes-Jones The University of.
TCP as a Reliable Transport. How things can go wrong… Lost packets Corrupted packets Reordered packets …Malicious packets…
1 Internet Traffic Measurement and Modeling Carey Williamson Department of Computer Science University of Calgary.
@Yuan Xue A special acknowledge goes to J.F Kurose and K.W. Ross Some of the slides used in this lecture are adapted from their.
Window Control Adjust transmission rate by changing Window Size
Approaches towards congestion control
The Transport Layer (TCP)
Satellite TCP Lecture 19 04/10/02.
Reddy Mainampati Udit Parikh Alex Kardomateas
TCP Vegas: New Techniques for Congestion Detection and Avoidance
Limited Transmit for TCP
The transfer performance of iRODS between CC-IN2P3 and KEK
Khiem Lam Jimmy Vuong Andrew Yang
Magda El Zarki Professor, ICS UC, Irvine
TCP Sequence Number Plots
Lecture 19 – TCP Performance
CS Lecture 2 Network Performance
Cisco Prime NAM for WAN Optimization Deployment
Sven Ubik TCP performance Sven Ubik
“Promoting the Use of End-to-End Congestion Control in the Internet”
Project-2 (20%) – DiffServ and TCP Congestion Control
CSE 4213: Computer Networks II
TCP Interactive Data Flow
When to use and when not to use BBR:
Presentation transcript:

Evaluation of Objectivity/AMS on the Wide Area Network SATO Hiroyuki KEK Computing Research Center

Introduction Measured the performances of Objectivity/AMS : AMS write/read access (5.1 vs 5.2) AMS over LAN and WAN Resource utilization of the AMS server is monitored for 5.1 and 5.2 on LAN CPU utilization TCP/IP packet was monitored for AMS transaction over LAN and WAN. Transfer rate

Testbed Configuration Monitored CPU/Packet/…

AMS/Write-Read Performance Test on LAN and WAN Objectivity page size is 8192 bytes Test object (40 bytes) : 175~176 objects / page #’objects for the write/read : 1,000,000 objects for LAN : 570 pages 100,000 objects for WAN : 57 pages

CPU Utilization for Write/Read for V5.1 and V5.2 on LAN 20 40 60 (sec) 80 100 30 10 CPU Utilization (%) Time V5.1 V5.2 AVE:12% AVE:18% AVE:4% AVE:8%

Measurements on the WAN Characteristics of the network between KEK and CERN RTT (Round Trip Time) is about 300 msec The bandwidth is 2Mbps 10 20 30 500 400 300 (msec) Day in 2000 Round Trip Time between KEK and CERN

Data Transfer for Write on LAN and WAN 20 40 60 80 (sec) Time 1000 200 400 600 800 1200 Transfer Rate (kB/s) Server (AMS) at KEK Client at KEK/CERN 52 bytes Control Transfer Phase (CTP) 8236 bytes 8280 bytes 36 bytes Data Transfer Phase (DTP) 8280 bytes Bulk Transfer

Data Transfer for Read on LAN and WAN 25 50 75 100 (sec) Time 500 200 300 400 Transfer Rate (kB/s) 150 Server (AMS) at KEK Client at KEK/CERN Control Data Transfer Phase (CDTP) 52 bytes RTT 8236 bytes 1 2 Time since #’4M is sent (sec) 4000 4100 4200 x 103 Seq. Number Write Read

Comparison with AMS Write and ftp Write performance over WAN is faster than ftp ! 20 40 60 80 100 120 140 Transfer Rate (kB/s) AMS ftp 20 40 60 80 Time (sec)

Window Size in TCP/IP Monitored with tcpdump AMS : 33580 bytes ftp : 24820 bytes Window size affects the transfer rate It can be changed with “setsockopt” function (standard max : 65535 bytes / optional 232 bytes) Bulk data transfer rate for various received window size is measured

Bulk Data Transfer Rate for Various Window Size (KEK - CERN) 64240 bytes (Test Program) 33580 bytes (Objectivity) 24820 bytes (ftp) Window size 6 12 18 24 (hour) Time 50 100 150 200 250 Data Transfer Rate (kB/s) 02/Feb/2000

Effective Bandwidth for RTT = 280 ms (KEK - CERN) 10 -2 10 -1 1 10 1 10 2 ftp AMS Efficiency (%) 2Mbps Efficiency = Window Size RTT Bandwidth 20Mbps Needs the window scale option 200Mbps 10 3 10 7 10 6 10 5 10 4 Window Size (bytes)

Comparison with network of surface and satellite CERN 1 2 3 4 5 6 7 8 9 KEK 100 150 200 1 2 3 Page Number Transaction Time / Page (sec) surface (RTT=300ms) satellite (RTT=655ms) Read 1 2 3 4 5 6 7 8 TCP Timeout LOST! 10 11 ack2 10 11 ack8 12 13 ack4 12 13 14 15 16 ack6 14 15 16 ack8 9 ack16 8236 = 536*15+196 ==> 16 segmets MSS for satellite The congestion window is initialized after this transaction.

Summary Write/Read operation of Objectivity 5.2 works on the WAN AMS server 5.2 is multi-threaded Consumes more CPU cycles than 5.1 Does not matter when network is a bottleneck Write performance over WAN is faster than ftp more speedup is expected with a larger window size (Optimal window size may vary with the available buffer size of the network router) Read performance is poor due to the hand-shaking Oscillation of congestion window is observed in satellite network — further study needed