Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Basic NetBackup Performance TuningBill Roth - Technical Product Manager, NetBackupRay Schafer – Technical Product Manager, NetBackupJune 12, 2007
This forward-looking indication of plans for products is preliminary and all future release dates are tentative and are subject to change. Any future release of the product or planned modifications to product capability, functionality or feature are subject to ongoing evaluation by Symantec, and may or may not be implemented and should not be considered firm commitments by Symantec, and should not be relied upon in making purchasing decisions.
Symantec Vision 2007 2
Agenda
Opening 11
Examining the challenge 22
The data transfer path 33
Bottleneck identification 44
Closing 55
Symantec Vision 2007 3
The Situation
• Complaints about slow backups? – Backups don't complete in time and impact production workload– Backups don't complete within a 24 hour window, impacting the next
days backups– Fast tape drives aren't being fully utilized
• Issues with long restores– Database recovery takes days instead of minutes– Full volume recovery takes much longer than expected– File recovery takes a long time to complete
Symantec Vision 2007 4
The Situation
• Complaints…– Hello, Has anyone use a LTO Ultrium 2 autoloader tape drive that has
decent backup time? I have one and I can't figure why it take so long just to backup 280GB locally.
• The tape drive is connect directly to the server, a Dell PowerEdge 2950 and all I’m backing up is the C and D drive, about 280g and it takes 6:30:00 to complete, suppose to take about two hours only.
• I call HP up and go through some testing and eventually was told, "the device is working fine, not sure why it takes so long", by Hp Tech support.
Symantec Vision 2007 5
The Complication
• Advice is wonderful, but it can complicate the equation…– LTO-2 tend to do 32 MB/s with compression and 16 MB/s without
compression. That means;• 1 hr at 32megs per second you get 115200mb or 115gig per hour• 1 hr at 16megs per second you get half that at 57.6gig.
– Assuming the optimal number it would take 2.4 hours to backup. Figure the same amount of time for verify. This should put your backup at about 5 hours.
• "suppose to take about two hours only“• Can you see now where your calculations are incorrect? Odds are you were not
including verify time nor delays caused by open files. Your tape drive should also be on a SCSI card all by itself.
Symantec Vision 2007 6
The Complication
• More advice adds to the complication…
– I have found if you’re backing up a lot of small files it takes a lot longer. I have a DLT S4. It says it will backup 800GB uncompressed (one tape) in 10 hours ...
• It takes 24 usually for me.
Symantec Vision 2007 7
The Implication
• The need for basic performance tuning skills is clear– Unless you’re “lucky” considerable time will be consumed, possibly
without positive results– There is a great likelihood that you may face multiple data protection
performance issues, each with a different root cause
Symantec Vision 2007 8
Agenda
Examining the challenge 22
Symantec Vision 2007 9
The Challenge is Clear
• But where should I start?– Are SLA's for backup and restore in place...
• Some sites have general guidelines stating that all full backups should complete during a weekend window
• Some sites have more granular guidelines that apply to specific data types– These sites may have characterized data according to business value
or importance• Some sites have no documented plan, and wait for complaints and problems
to surface• Some of us may have inherited a data protection infrastructure that has been
ignored or mismanaged in the past
– How do you define "slow backups" and "long restores"?• Unless SLA's have been defined, "slow" or "long" may be a perception issue• Does anybody actually measure and record performance?
Symantec Vision 2007 10
The Challenge is Clear
• Before trying to address the challenge it might be a good idea to better understand a few key areas:– What backup and restore performance you're trying to achieve?
• Basing this determination on business requirements is recommended• Segregating or characterizing requirements based on data business value is
also recommended
– What are current performance levels?• Measure and record backup and restore performance on a daily, weekly, or
monthly basis• Track both full and incremental backups• Track all restore requests• Symantec recommends restore testing on a frequent basis, measure and
record these tests as well
Symantec Vision 2007 11
The Challenge is Clear
• Once the challenge is accurately defined it becomes possible to evaluate the situation using a logical sequence of steps– Performance goals / SLA's are known and understood– Current performance is known and non-compliant data protection
operations have been identified
Symantec Vision 2007 12
Agenda
The data transfer path 33
Symantec Vision 2007 13
Data Transfer Path
• What is the data transfer path for this slow client?
Remote Office 1 Remote Office 2
WAN
Main Campus 1 Main Campus 2
SAN
Backup Infrastructure
• Defining a typical data transfer path
Symantec Vision 2007 14
Data Transfer Path
Backup Server NetBackup media server, or master & media server combination
Storage Device Secondary storage used for backup images
TCP/IP network path between the client and the NetBackup server
NetBackup client with data protection requirements
Network
Client
• Defining a typical data transfer path
Symantec Vision 2007 15
Data Transfer Path
Symantec Vision 2007 16
Data Transfer Path
• Defining the data transfer path– Begin by trying to isolate the data transfer path relative to a
specific client from the bulk of the data protection infrastructure– The basic high level data transfer path includes the client, a
network, a media server, and a secondary storage device
Backup Server Storage DeviceNetworkClient
Symantec Vision 2007 17
Data Transfer Path
• Defining the data transfer path– For those of us that don’t appreciate block diagrams:
NetBackup Client
Disk Storage
Network
Tape & Disk
Storage Units
NetBackup Media Server
Symantec Vision 2007 18
Data Transfer Path
• Understand the client component of the data transfer path:
– 2x 3.2 GHz Intel Xeon CPU– 2 GB RAM– 2x Intel PRO/1000 MT NIC– LSI PCI-X Ultra320 SCSI Adapter– Emulex LP9000 HBA– Seagate ST373207LC SCSI Disk– 10x Hitachi SCSI Disk (SAN Connected)
MS Windows Server 2003 SP2MS Exchange 2003 SP2NetBackup 6.0 MP4 Client
Client
Symantec Vision 2007 19
Data Transfer Path
• Understand the client component of the data transfer path:
3.2 GHz CPU
10/100/1000 NIC
3.2 GHz CPU
10/100/1000 NIC
1 Gb HBA
2 GB RAM
Ultra 320 SCSI
73 GB
6.9 GB6.9 GB
6.9 GB6.9 GB
6.9 GB
6.9 GB6.9 GB
6.9 GB6.9 GB
6.9 GB
NetBackup Client
Symantec Vision 2007 20
Data Transfer Path
• Understand the client component of the data transfer path:– Examine additional parameters such as disk drive specific
information– Vendor web pages often provide great detail:
Symantec Vision 2007 21
Data Transfer Path
• Understand the client component of the data transfer path:– Examine additional parameters…– LSI PCI-X Ultra320 SCSI Adapter
• View vendor web pages to obtain additional information• View industry web pages for additional information
– www.t10.org– www.scsita.org
– Ultra320 is defined as being able to provide up to 320 MB/s data transfer rates
Symantec Vision 2007 22
Data Transfer Path
• Understand the network component of the data transfer path:
Client NIC configurationNetwork
• Parameters– Local settings are easily determined– Other parameters may be more difficult to
determine
Symantec Vision 2007 23
Data Transfer Path
• Understand the network component of the data transfer path:– Understand the network path being used to perform backups /
restores• Switches• Hops• Port settings• Other network traffic
NetBackup Client
Network
NetBackup Media Server
Symantec Vision 2007 24
Data Transfer Path
• Understand the media server component of the data transfer path:
– Attributes:• CPU• Memory• HBAs• NICs• Other adapters, disk drives, etc.
NetBackup 6.0 MP4 media serverBackup Server
NetBackup Media Server
Symantec Vision 2007 25
Data Transfer Path
• Understand the storage device component of the data transfer path:
– Attributes:• Device type• Rated speed• Capacity
Disk & Tape based secondary storageStorage Device
Symantec Vision 2007 26
Agenda
Bottleneck identification 44
Symantec Vision 2007 27
Bottleneck Identification
• Before trying to identify a bottleneck:– Characterize performance
• Take a granular look at an individual client relative to the backup infrastructure• Break down the infrastructure into individual components
– Measure & document current performance attributes• Determine & record current data transfer rates, backup & restore duration
– Have a goal, SLA, or performance metric that you’re trying to achieve
Symantec Vision 2007 28
Bottleneck Identification
• Bottlenecks & current limitations– There’s always a bottleneck in the data transfer path
• It may be a very fast bottleneck, but it exists non-the-less
– Use a variety of tool and utilities to evaluate current performance– Pinpoint performance bottlenecks
Symantec Vision 2007 29
Bottleneck Identification
• Tools and utlities– On the client, you can use a variety of tools and utilities that can
assist is understanding how fast data can be extracted from the source disk or array:
• Bonnie++ (Unix)• Tiobench (Linux)• HDSpeed (Windows)• Many others available
– Additionally, you can execute the NetBackup bpbkar process in standalone mode
• This is the NetBackup process that is used during backups• The “NetBackup Planning and Performance Tuning Guide” contains
instructions for using this tool
Symantec Vision 2007 30
Bottleneck Identification
• Tools and utlities– Client data extraction example – IOMeter:
– This utility is able to read file system data at about 54 MB/s
Symantec Vision 2007 31
Bottleneck Identification
• Tools and utilities– Network tools and utilities can be used to determine if network
bottlenecks are impacting backup performance• Copying data across a network connection may provide a ballpark
performance figure, but the results can be skewed by source and destination disk drives
– IPERF• Can be used to test and isolate network issues• Can measure throughput in either direction• Supported on Unix, Linux, and Windows• Testing does not involve disk subsystems
Symantec Vision 2007 32
Bottleneck Identification
• Tools and utilities– Network performance test example, IPERF
– Test results indicate an approximate 42 MB/s data transfer rate
Symantec Vision 2007 33
Bottleneck Identification
• Tools and utilities– What about the media server?
• A variety of tools and utilities are available to assist in bottleneck identification• NetBackup debug logs such as bptm, bpdm, and bpbkar• OS resident utilities that track CPU and memory utilization, etc.
– OS resident logs are helpful in a variety of cases• Windows Event logs, and additionally, viewing the BIOS messages during the
boot process• Unix var/adm/messages
Symantec Vision 2007 34
Bottleneck Identification
• Tools and utilities– NetBackup media server
Symantec Vision 2007 35
Bottleneck Identification
• Tools and utilities– NetBackup media server
Symantec Vision 2007 36
Bottleneck Identification
• Tools and utilities - Buffers– The NetBackup media server uses shared memory to buffer data
between the network and the tape or disk drive• Or between the disk and the tape drive if the NetBackup media server and
client are the same system
– The number and size of these shared data buffers can be configured on the NetBackup media server
– The size and number of the tape and disk buffers may be changed so that NetBackup optimizes its use of shared memory
– Changing the default buffer size may result in better throughput for high-performance tape drives
• These changes may also improve throughput for other types of drives
– Buffer settings are for media servers only and should not be used on a pure master server or client
Symantec Vision 2007 37
Bottleneck Identification
• Tools and utilities - Buffers
Symantec Vision 2007 38
Bottleneck Identification
• Tools and utilities - Buffers
• Tools and utilities – Buffers– The block diagram representation
* NetBackup 6.5 uses bptm with disk, prior versions use bpdm
Diskbpdm(child)
SharedBuffers
bpdm/bptm*(parent)
bptm(child)
SharedBuffers
bptm(parent)
Tape
Symantec Vision 2007 39
Bottleneck Identification
Symantec Vision 2007 40
Bottleneck Identification
• Tools and utilities – Buffers– There are two schools of thought surrounding the use of buffers
to accomplish the goal of improved performance– The first uses NetBackup logs to view buffer wait and delay
counters• This approach is more time consuming but provides insight into what’s going
on internally within NetBackup• Logs provide buffer usage details on a process by process basis• These are generally used to provide input for bottleneck identification and
additional tuning
– The second takes a more generalized approach of just setting them to an “optimal” value
• We’ve run through an exhaustive series of steps and have come to the conclusion that in our environment, a predetermined quantity of buffers and predetermined buffer size gives us the best performance
Symantec Vision 2007 41
Bottleneck Identification
• Tools and utilities – Performance Tuning Guide
Symantec Vision 2007 42
Bottleneck Identification
• Tools and utilities – Performance Tuning Guide
Symantec Vision 2007 43
Bottleneck Identification
• Tools and utilities– Storage Devices should be matched to HBA’s or SCSI adapters
• Be cognizant of the number of drives and the throughput per drive compared to the speed of the FC/SCSI connection
• Understand rated throughput capabilities of tape drives
– Tape streaming is a good thing• Proper communication bandwidth is one of many factors that can assist in
eliminating “shoe-shining”
– Data Compression• Tape based compression is generally favorable
– Preferred over client side compression
Symantec Vision 2007 44
Bottleneck Identification
• Tools and utilities– Storage Devices - understand current limitations…– Use “dd” or “tar” to test without reading data
from a disk source• UNIX:• dd if = /dev/zero of = /dev/rmt/0cbn
– Windows:• dd if = NUL of = \\.\TAPE0• Note: Windows may require the installation of a utility or
other software to enable these commands…
NetBackup Master / Media
Server
Tape Storage
Unit
Symantec Vision 2007 45
Bottleneck Identification
• Tools and utilities– Storage Devices - understand current
limitations…– UNIX measurement tools
• iostat• SAR
– Windows measurement tools• NetBackup includes four NetBackup related
performance Counters:– Disk / Tape Read (2)– Disk / Tape Write (2)
Symantec Vision 2007 46
Agenda
Closing 55
Symantec Vision 2007 47
Closing
• Improving performance– Develop a plan of attack
• Bottleneck identification should yield a list of areas identified as potential bottlenecks
– Prioritize bottleneck elimination• It often makes sense to eliminate major bottlenecks first• It may not make sense to purchase and deploy faster tape drives if the current
tape drives are not the main bottleneck in the data path, for example
– Execute a methodical plan of attack• As the configuration is altered for bottleneck elimination, make a minimal
number of simultaneous changes and record new performance metrics• Confirm that the appropriate bottleneck has been eliminated• Move on to address subsequent bottlenecks as need to meet business
requirements
Symantec Vision 2007 48
Closing
• Additional reference– Advanced NetBackup Performance Tuning
• Session S287204• Today following this session
– “Backup & Recovery”• By W. Curtis Preston• Available on amazon.com
– “Implementing Backup and Recovery: The Readiness Guide for the Enterprise”
• By David B. Little and David A. Chapa• Available on amazon.com
– “Veritas NetBackup™ Backup Planning and Performance Tuning Guide”
• Available on support.veritas.com
Copyright © 2007 Symantec Corporation. All rights reserved. Symantec and the Symantec Logo are trademarks or registered trademarks of Symantec Corporation or its affiliates in the U.S. and other countries. Other names may be trademarks of their respective owners.This document is provided for informational purposes only and is not intended as advertising. All warranties relating to the information in this document, either express or implied, are disclaimed to the maximum extent allowed by law. The information in this document is subject to change without notice. This forward-looking indication of plans for products is preliminary and all future release dates are tentative and are subject to change. Any future release of the product or planned modifications to product capability, functionality or feature are subject to ongoing evaluation by Symantec, and may or may not be implemented and should not be considered firm commitments by Symantec, and should not be relied upon in making purchasing decisions.
Thank You