Small number of IOVs – no server stress

Slides:



Advertisements
Similar presentations
1 Chapter 16 Tuning RMAN. 2 Background One of the hardest chapters to develop material for Tuning RMAN can sometimes be difficult Authors tried to capture.
Advertisements

External Sorting The slides for this text are organized into chapters. This lecture covers Chapter 11. Chapter 1: Introduction to Database Systems Chapter.
Lecture 8: Memory Hierarchy Cache Performance Kai Bu
Some More Database Performance Knobs North American PUG Challenge
LOAD BALANCING IN A CENTRALIZED DISTRIBUTED SYSTEM BY ANILA JAGANNATHAM ELENA HARRIS.
G. Alonso, D. Kossmann Systems Group
Review of Relative Density Principles v Relative Density principles apply to compaction of relatively clean, coarse- grained soils. v Relatively clean.
Database Management Systems, R. Ramakrishnan and J. Gehrke1 External Sorting Chapter 11.
Chapter 2 Analytical Models 2.1Cost Models 2.2Cost Notations 2.3Skew Model 2.4Basic Operations in Parallel Databases 2.5Summary 2.6Bibliographical Notes.
1 External Sorting Chapter Why Sort?  A classic problem in computer science!  Data requested in sorted order  e.g., find students in increasing.
OS Fall ’ 02 Performance Evaluation Operating Systems Fall 2002.
B + -Trees (Part 1) COMP171. Slide 2 Main and secondary memories  Secondary storage device is much, much slower than the main RAM  Pages and blocks.
1 External Sorting Chapter Why Sort?  A classic problem in computer science!  Data requested in sorted order  e.g., find students in increasing.
1 External Sorting for Query Processing Yanlei Diao UMass Amherst Feb 27, 2007 Slides Courtesy of R. Ramakrishnan and J. Gehrke.
OS Fall ’ 02 Performance Evaluation Operating Systems Fall 2002.
May 21, 2009 Stata for micro targeting using C++ and ODBC Masahiko Aida.
External Sorting Chapter 13.. Why Sort? A classic problem in computer science! Data requested in sorted order  e.g., find students in increasing gpa.
Average Session Load (ASL) The Golden Metric ? Kyle Hailey
Zabbix Performance Tuning
Database Project Team 4 Group c v Menna Hamza Mohamad Hesham Mona Abdel Mageed Yasmine Shaker.
Index tuning-- B+tree. overview © Dennis Shasha, Philippe Bonnet 2001 B+-Tree Locking Tree Traversal –Update, Read –Insert, Delete phantom problem: need.
Physical Database Design & Performance. Optimizing for Query Performance For DBs with high retrieval traffic as compared to maintenance traffic, optimizing.
Chapter 20 Other Memory Management Topics
ATLAS Metrics for CCRC’08 Database Milestones WLCG CCRC'08 Post-Mortem Workshop CERN, Geneva, Switzerland June 12-13, 2008 Alexandre Vaniachine.
PROC SQL Phil Vecchione. SQL Structured Query Language Developed by IBM in the early 1970’s From the 70’s to the late 80’s there were different types.
Today  Table/List operations  Parallel Arrays  Efficiency and Big ‘O’  Searching.
1 Wenguang WangRichard B. Bunt Department of Computer Science University of Saskatchewan November 14, 2000 Simulating DB2 Buffer Pool Management.
BW Know-How Call : Performance Tuning dial-in phone numbers! U.S. Toll-free: (877) International: (612) Passcode: “BW”
Applications hitting a wall today with SQL Server Locking/Latching Scale-up Throughput or latency SLA Applications which do not use SQL Server.
Embedded System Lab. 오명훈 Memory Resource Management in VMware ESX Server Carl A. Waldspurger VMware, Inc. Palo Alto, CA USA
Sorting.
Achieving Scalability, Performance and Availability on Linux with Oracle 9iR2-RAC Grant McAlister Senior Database Engineer Amazon.com Paper
Large Data Operations Joe Chang
© 2014 Carl Lund, all rights reserved A First Course on Kinetics and Reaction Engineering Class 12.
Processes Introduction to Operating Systems: Module 3.
1 Database mini workshop: reconstressing athena RECONSTRESSing: stress testing COOL reading of athena reconstruction clients Database mini workshop, CERN.
OPERATING SYSTEMS CS 3530 Summer 2014 Systems with Multi-programming Chapter 4.
10/2005COOL read workload1 COOL verification Reconstruction stress testing David Front, Weizmann Institute It works well! Doesn’t it?
1 Components performance modelling - Outline of queue networks - Mean Value Analisys (MVA) for open and close queue networks.
Lecture 08: Memory Hierarchy Cache Performance Kai Bu
1 Chapter Seven. 2 Users want large and fast memories! SRAM access times are ns at cost of $100 to $250 per Mbyte. DRAM access times are ns.
Roy Ernest Database Administrator Pinnacle Sports Worldwide
Little’s Law & Operational Laws. Little’s Law Proportionality relation between the average number of jobs (E[N]) in a system and the average system time.
Introduction to Database Systems1 External Sorting Query Processing: Topic 0.
CPSC Why do we need Sorting? 2.Complexities of few sorting algorithms ? 3.2-Way Sort 1.2-way external merge sort 2.Cost associated with external.
Copyright © 2009 Rolta International, Inc., All Rights Reserved Michael R. Messina, Management Consultant Rolta-TUSC, Oracle Open World 2009 (60 min) ID#:
Bigtable: A Distributed Storage System for Structured Data
Database Management Systems, R. Ramakrishnan and J. Gehrke1 External Sorting Chapter 11.
External Sorting. Why Sort? A classic problem in computer science! Data requested in sorted order –e.g., find students in increasing gpa order Sorting.
ATLAS FroNTier cache consistency stress testing David Front Weizmann Institute 1September 2009 ATLASFroNTier chache consistency stress testing.
AMGA-Bookkeeping Carmine Cioffi Department of Physics, Oxford University UK Metadata Workshop Oxford, 05 July 2006.
External Sorting Jianlin Feng School of Software SUN YAT-SEN UNIVERSITY courtesy of Joe Hellerstein for some slides.
F453 Module 8: Low Level Languages 8.1: Use of Computer Architecture.
 INDEX  Overview.  Introduction.  System Requirement.  Features Of SQL.  Development Process.  System Design (SDLC).  Implementation.  Future.
Morten Kromberg CXO, Dyalog Ltd
CPS216: Data-intensive Computing Systems
Basic Performance Parameters in Computer Architecture:
David Front Weizmann institute May 2007
OCR AS Level F451: Data transmission
Chapter 3: Process-Concept
Alejandro Álvarez on behalf of the FTS team
CSE Social Media & Text Analytics
20 Questions with Azure SQL Data Warehouse
Lecture 28: Index 3 B+ Trees
Basic Cache Operation Prof. Eric Rotenberg
External Sorting.
Performance And Scalability In Oracle9i And SQL Server 2000
Cache - Optimization.
Presentation transcript:

Small number of IOVs – no server stress Explaining differences in COOL stress test results of Andrea and David: Small number of IOVs – no server stress

Summary When the number of IOVs in a COOL IOVs table is 1000, selection via COOL API stresses the server considerably This stress does not happen if table only has 50 IOVs The extra stress caused by querying the last vs. the first IOV in a table is independent to the above assertions Without the index, re-computing statistics causes the server stress to be reduced but just a little With 1000 IOVs, adding index on (IOV_since, IOV_until), causes the server stress to reduce considerably

Testing parameters base line 100 folders 100 channels per folder ~100 Bytes per IV record Statistics is calculated after inserting a few records At test, per folder, bulk read one IOV of all channels Read first IOV Per test, the same data is read by all clients No index on (IOV_since, IOV_until) Number of clients, repeated for: [20,30,40,50] Clients spawned once every 5 seconds Before test start, server memory is not cleaned

Parameters under test Num of IOVs at each IOV table: 50 (Andrea tested with 60 IOVs) vs. 1000 (David tested with 1000 IOVs) Statistics computer recently or not? Existence of index on (IOV_since,IOV_until)

50 IOVs numClients 20 30 40 50 Avg (clientMem%) avg (clientsCP%) (load5) Max (elapsed) Min Test Value Name time 3 73 56 26 33 20 numClients 2005-10-26_09:38 91 1 27 30 2005-10-26_09:40 90 32 40 2005-10-26_09:43 50 2005-10-26_09:47 Test: Small number of IOVs, read first IOV Server: Quite happy: high IO at beginning, mainly CPU later Clients: ~ 90% CPU Avg. elapsed time: steady 26-33 Conclusion: (Library) latch contention is negligible when number of IOVs at table is ~50

50 IOVS, read last IOV numClients 20 30 40 50 Avg (clientMem%) avg (clientsCP%) (load5) Max (elapsed) Min Test Value Name time 3 82 15 13 14 20 numClients 2005-10-26_11:53 79 30 2005-10-26_11:55 1 18 40 2005-10-26_11:58 83 16 50 2005-10-26_12:02 Test: Small number of IOVs, read last IOV (# 49) Server: Very happy: mainly CPU Clients: ~80% CPU Avg. elapsed time: fast ~14 sec Conclusion: When table has only 50 IOVs, reading the last vs. the first IOV does not cause extra server stress

1000 IOVs, no New Index numClients 20 30 40 50 Avg (clientMem%) avg (clientsCP%) (load5) Max (elapsed) Min Test Value Name time 3 6 1 482 362 412 20 numClients 2005-10-26_14:03 11 311 165 222 30 2005-10-26_14:09 278 152 224 40 2005-10-26_14:13 10 326 169 254 50 2005-10-26_14:21 Test: 1000 IOVs per table, no new index, read first IOV Server: High IO and library latch contention Clients: ~10% CPU Avg. elapsed time: slowe ~200-400 sec Conclusion: When table has 1000 IOVs, not having new index on (IOV_since, IOV_until), causes server to do much IO and have library latch contention

1000 IOVs, no New Index,statistics numClients 20 30 40 50 Avg (clientMem%) avg (clientsCP%) (load5) Max (elapsed) Min Test Value Name time 3 16 193 135 166 20 numClients 2005-10-26_14:49 13 1 237 156 194 30 2005-10-26_14:53 7 690 173 472 40 2005-10-26_14:59 8 383 186 288 50 2005-10-26_15:15 Test: 1000 IOVs per table, no new index, read first IOV, after re-compute statistics Server: High IO and library latch contention Clients: ~10% CPU Avg. elapsed time: slowe ~200-400 sec Conclusion: Re-computing statistics did lower library latch contention only for lower number of clients

1000 IOVs, with New Index numClients 20 30 40 50 Avg (clientMem%) avg (clientsCP%) (load5) Max (elapsed) Min Test Value Name time 3 92 31 25 26 20 numClients 2005-10-26_12:59 95 1 27 30 2005-10-26_13:02 91 40 2005-10-26_13:05 90 28 50 2005-10-26_13:09 Test: 1000 IOVs per table, with new index, read first IOV Server: Very happy: mainly CPU Clients: ~90% CPU Avg. elapsed time: fast 26 sec Conclusion: When table has 1000 IOVs, using new index on (IOV_since, IOV_until), server is happy