Professional Documents
Culture Documents
Legal Notice
Copyright 2010 Symantec Corporation. All rights reserved.
Symantec, the Symantec Logo, and NetBackup 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.
The product described in this document is distributed under licenses restricting its use,
copying, distribution, and decompilation/reverse engineering. No part of this document
may be reproduced in any form by any means without prior written authorization of
Symantec Corporation and its licensors, if any.
THE DOCUMENTATION IS PROVIDED "AS IS" AND ALL EXPRESS OR IMPLIED CONDITIONS,
REPRESENTATIONS AND WARRANTIES, INCLUDING ANY IMPLIED WARRANTY OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT,
ARE DISCLAIMED, EXCEPT TO THE EXTENT THAT SUCH DISCLAIMERS ARE HELD TO
BE LEGALLY INVALID. SYMANTEC CORPORATION SHALL NOT BE LIABLE FOR INCIDENTAL
OR CONSEQUENTIAL DAMAGES IN CONNECTION WITH THE FURNISHING,
PERFORMANCE, OR USE OF THIS DOCUMENTATION. THE INFORMATION CONTAINED
IN THIS DOCUMENTATION IS SUBJECT TO CHANGE WITHOUT NOTICE.
The Licensed Software and Documentation are deemed to be commercial computer software
as defined in FAR 12.212 and subject to restricted rights as defined in FAR Section 52.227-19
"Commercial Computer Software - Restricted Rights" and DFARS 227.7202, "Rights in
Commercial Computer Software or Commercial Computer Software Documentation", as
applicable, and any successor regulations. Any use, modification, reproduction release,
performance, display or disclosure of the Licensed Software and Documentation by the U.S.
Government shall be solely in accordance with the terms of this Agreement.
Symantec Corporation
350 Ellis Street
Mountain View, CA 94043
http://www.symantec.com
Contents
Chapter 1
Introduction
........................................................................... 5
Chapter 2
Chapter 3
Chapter 4
Index
.................................................................................................................... 31
Contents
Chapter
Introduction
This chapter includes the following topics:
Introduction
About this tech note
Chapter
Support for the AIX 5.3, 6.1, and 7.1 operating systems for deduplication servers
and for client-side deduplication.
NetBackup now reserves 4 percent of the storage space for the deduplication
database and transaction logs rather than 10 percent.
Chapter
10
Use case
Deduplication type
Table 3-1
Use case
Deduplication type
The data that are provided in this document are averages derived from testing.
The data provide information you can use to help determine the requirements
for your deduplication environment and do not represent a commitment by
Symantec to specific performance objectives.
See Client deduplication example on page 11.
See Media server deduplication example on page 13.
11
12
Figure 3-1
Network bottleneck
NetBackup
clients
Bottleneck
NetBackup media
server
Figure 3-2
NetBackup
deduplication
clients
No bottleneck!
Figure 3-3 shows that client deduplication may be a better solution in a Gigabit
Ethernet (GigE) network than media server deduplication. The figure shows
average performance with 80% deduplication. With client deduplication, backups
for 12 clients are three times as fast as using media server deduplication (282
MB/s versus 94 MB/s).
Figure 3-3
13
14
Figure 3-4
Figure 3-5
16 GB RAM
Client pool:
4 GBs RAM
8 GB RAM
15
16
Table 3-2
Ingest rate
115 MB/s
460 MB/s
Clock speed and number of cores. The higher the individual clock speed, the
faster an individual backup stream. The more cores you have, the higher the
aggregated backup throughput.
The deduplication rate, which depends on the type of data. The higher the
deduplication rate, the lower the CPU affect; less data needs to be transferred,
encrypted, or compressed.
The speed at which you stream the data to the server. The faster the stream,
the higher the CPU utilization per core. Each stream is handled by one core.
The number of simultaneous backup streams. The more streams, the more
parallel deduplication calculations are performed and the greater the CPU
affect.
17
18
The following tables provide throughput information for a test environment that
uses Intel Architecture hosts with 2.8-GHz quad core CPUs. AMD and Intel
processor performance is similar.
Table 3-3 shows the average throughput per core.
Table 3-3
Average backup throughput per core with 2.8-GHz quad core CPU
Cores
0% deduplication
100% deduplication
50 MB/s
90 MB/s
Table 3-4 shows the average throughput per quad core CPU both for a storage
server and a load balancing server. (A NetBackup load balancing media server
provides performance similar to a NetBackup media server that hosts the PureDisk
Deduplication Option agent.)
Table 3-4
Quad core
CPUs
Storage server
Storage server
0% deduplication
100% deduplication
140 MB/s
275 MB/s
200 MB/s
360 MB/s
275 MB/s
550 MB/s
400 MB/s
720 MB/s
The speed of the throughput for a single stream is an important factor for
application database backups such as Oracle.
Table 3-5 shows the peak throughput per stream both for a storage server and a
load balancing server. (A NetBackup load balancing media server provides
performance similar to a NetBackup media server that hosts the PureDisk
Deduplication Option agent.) Backing up an Oracle database with a single stream
consumes one core of a 2.8-GHz quad core CPU. If the server has eight cores, the
stream consumes 12.5% of the processor.
Usually, Oracle is backed up by multiple parallel streams. On a single quad core
2.8-GHz CPU host with four streams, a stream rate of 90 MB/s saturates all four
cores. The host can back up an Oracle database with four streams at a rate of 360
MB/s (best case). If the media server has two quad core CPUs, the Oracle backup
consumes 50% of the processor but the backup finishes twice as fast as with a
single stream. (As a reminder, these numbers differ for each environment and
depend on the CPU speeds.)
Table 3-5
Storage server
100% deduplication
0% deduplication
100% deduplication
200 MB/s
115 MB/s
230 MB/s
115 MB/s
The Sun T2000 processor has slower CPU clocks but more cores than Intel or
AMD. The single stream rate is lower, and therefore more streams are required
to achieve throughput similar to Intel and AMD processors. The following table
shows the average performance on a load balancing server.
Table 3-6
Processor
8 streams
32 streams
Sun T2000
115 MB/s
460 MB/s
19
20
100 TB
10%
Backup types
24 hours
10 hours
Retention period
four weeks
Daily incremental
Because 1 NIC supports up to 115 MB/s (after a 10% performance deduction for
noise or overhead), the following are required for the example environment:
You can spread the NICs across the deduplication servers or use faster HBAs on
the storage server.
See Data ingest rate on servers on page 16.
See Media server deduplication sizing example on page 20.
See Performance test results on page 22.
Backup type
Ingest rate
Weekly full
1213 MB/s
Daily incremental
291 MB/s
You can use a combination of storage servers and loading balancing servers to
achieve the performance requirements. A storage server with two quad core
processors provides 280-MB/s throughput . A load balancing server with two quad
core processors provides 400-MB/s throughput. In this example, two storage
servers and two load balancing servers provide enough throughput (1360 MB/s)
to accommodate the ingest rate for the weekly full backup.
See Media server CPU and deduplication on page 17.
See Media server deduplication sizing example on page 20.
See Performance test results on page 22.
21
22
Table 3-8
Backup
Data
Deduplication
Data transferred
Window
Minimum network
capacity required
Weekly full
100 TB
93-95%
5-7 TB
24 hours
60-85 MB/s
Daily incremental
10 TB
80-90%
1-2 TB
10 hours
29-58 MB/s
The maximum effective transfer rate possible for a 1 Gbit/s NIC is 115 MB/s.
Therefore, a 1 gigabit Ethernet is sufficient bandwidth between the storage servers
and the load balancing servers.
See Media server deduplication sizing example on page 20.
See Performance test results on page 22.
One storage server with 2 quad core processors, 24 GBs of RAM, and 3 NICs.
One load balancing server with two quad core processors, 4 GBs of RAM, and
3 NICs.
Note: Each storage server and load balancing server pair is a separate deduplication
node. Deduplication between the two nodes does not occur.
In a Gigabit network, the following table shows the possible performance.
Table 3-9
1 GigE performance
Backup
Backup throughput
Weekly full
1266 MB/s
23 hours
Daily incremental
812 MB/s
3.59 hours
Table 3-10
10 GigE performance
Backup
Backup throughput
Weekly full
1344 MB/s
21.7 hours
Daily incremental
1344 MB/s
2.2 hours
Avoid client deduplication for slow processors or for processors that rely on
more than two streams for maximum performance.
CPU utilization
Number of cores
Percent used
50%
23
24
Table 3-11
Number of cores
Percent used
25%
12.5%
16
< 10%
The backup speed depends on the CPU type and clock speed. Intel and AMD
perform similarly. The new SPARC CPU performs similarly to the AMD and Intel
processors. The following table displays examples of backup speeds for these CPU
types .
Table 3-12
Backup speed
0%
80 MB/s
22 MB/s
55 MB/s
90%
120 MB/s
45 MB/s
NA
100%
130 MB/s
70 MB/s
120 MB/s
Average RAM usage on a Linux client is approximately 500 MB. The following
table shows the RAM usage per number of streams; two is the optimal number of
streams for a client.
Table 3-13
Number of streams
430
470
580
740
16
1080
32
1870
Deduplication rate
One stream
Two streams
0%
80 MB/s
114 MB/s
90%
120 MB/s
173 MB/s
100%
130 MB/s
~ 200 MB/s
25
26
Chapter
Revision history
This chapter includes the following topics:
Revision history
Revision
Date
Description
February 1, 2010
Initial version.
February 8, 2010
28
Revision history
Revision history table
Table 4-1
Revision
Date
Description
April 1, 2010
Added CIFS and NFS as files systems that are not supported.
Added information about CPU attributes that are desirable for deduplication to
the following section in the deployment planning chapter: About deduplication
media server requirements.
Added the following to the deployment planning chapter: About deduplication
storage requirements, About deduplication storage capacity, About optimized
duplication of deduplicated data, and About NetBackup deduplication port
usage.
Added the following to the configuration updates chapter: Configuring optimized
duplication of deduplicated data.
Added the following to the operational updates chapter: Optimized duplication
fixes.
Added the following to the troubleshooting chapter: Duplication configuration
script log, Server not found, Database system error (220), Media open error
(83), Media write error (84), Cannot delete a disk pool, and Restoring files
at a remote site.
Also made various edits throughout the document.
Revision history
Revision history table
Table 4-1
Revision
Date
Description
September 24,
2010
The number of this document changed from Tech Note 338123 to TECH77575.
Removed sections from various chapter; they either no longer apply to the
NetBackup 7.0.1 release or were added to the 7.0.1 version of the NetBackup
Deduplication Guide.
Removed the following topics from the deployment planning chapter: About
deduplication storage requirements, About deduplication storage capacity,
About optimized duplication of deduplicated data, and About NetBackup
deduplication port usage, About deduplication and SAN Client, and
Deployment best practices.
Removed the following topics from the configuration chapter: Using an
alternate network interface for the deduplication storage, Configuring
optimized synthetic backups for deduplication, Throttling optimized
deduplication traffic, Enabling compression or encryption, Configuring
optimized duplication of deduplicated data,Setting deduplication storage
server attributes, Clearing deduplication storage server attributes, Viewing
deduplication storage server attributes, Setting deduplication pool
attributes, Clearing deduplication pool attributes, Viewing deduplication
pool attributes, About the deduplication storage server configuration file,
Setting NetBackup configuration options by using bpsetconfig, and
Reconfiguring the deduplication storage server and storage paths.
Removed the following topics from the operational updates chapter:
Optimized duplication fixes, About storage unit groups for deduplication,
About NDMP deduplication, About Windows System State deduplication,
About the optimal number of backup streams, About storage unit groups
for deduplication Performing maintenance manually, Resizing the storage
partition, Deleting a deduplication host configuration file, Resetting the
deduplication registry,Restoring files at a remote site, and Specifying the
restore server.
Removed the following topics from the troubleshooting chapter: Client
deduplication fails, Duplication configuration script log, Server not found,
Database system error (220), Media open error (83), Media write error
(84), About the deduplication host configuration file, Server not found
error, Storage full conditions, Cannot delete a disk pool, and
Deduplication fails after services are restarted or a domain controller is
restarted.
Deleted the troubleshooting chapter (no topics remain).
February 16, 2011 Added a new topic: About storage management for NetBackup deduplication.
29
30
Revision history
Revision history table
Table 4-1
Revision
Date
Description
Removed the reference to a fix in the Using an alternate network interface for
the deduplication storage topic.
7.1
February 10, 2012 Changed the revision numbering to match the NetBackup release level.
7.5A
February 10, 2012 Removed all information that applied to NetBackup 7.0, 7.0.1, and 7.1. That
information either does not apply to 7.5 or was added to the NetBackup
Deduplication Guide for release 7.5. Added new feature information for 7.5.
Index
B
backup
speed 24
C
client deduplication
guidelines 11
D
data ingest rate
example 20
guidelines 16
deduplication
CPU usage 23
RAM usage 24
sizing guidelines 23
storage management 10
F
file system
Veritas File System for deduplication storage 10
M
media server
sizing guidelines 15
write speed guidelines 19
media server deduplication
CPU example 21
CPU guidelines 17
guidelines 13
V
volume manager
Veritas Volume Manager for deduplication
storage 10