Community
IBM Community Home
Business Analytics
Business Automation
Cloud Pak for Data
Data Science
DataOps
Hybrid Data Management
IBM Japan
IBM Z & LinuxONE
Internet of Things
Middleware
Power Systems
Public Cloud
Security
Storage
Supply Chain
Watson Apps
Log in
Announcements
Blogs
Groups
Discussions
Events
Glossary
Site Content
Libraries
on this day
between these dates
Posted by
Announcements
Blogs
Groups
Discussions
Events
Glossary
Site Content
Libraries
on this day
between these dates
Posted by
Skip to main content (Press Enter).
Sign in
Skip auxiliary navigation (Press Enter).
Hybrid Data Management
Topic groups
HDM Global
Data Server Manager
Database Conversion Workbench
Db2 (On Premises and Cloud)
Db2 Analytics Accelerator for z/OS
Db2 Data Management Console
Db2 Early Access Program
Db2 Event Store
Db2 DevOps Experience Sponsor User Program Community
Db2 for z/OS
Db2 Tools for z/OS
Db2 Warehouse (On Premises and Cloud)
IBM Data Virtualization Manager for z/OS
IBM Query Management Facility (QMF)
Informix
Integrated Analytics Systems
Netezza Performance Server
Open Source Offerings
User groups
Events
Upcoming Hybrid Data Management Events
On Demand Webinars
IBM Expert TV
Virtual Community Events
All IBM Community Events
Participate
Post to Forum
Share a Resource
Blogging on the Community
Connect with Hybrid Data Management Users
All IBM Community Users
Resources
IBM Support
IBM Cloud Support
IBM Champions
Demos
Marketplace
Marketplace
IBM Hybrid Data Management Community
Connect with Db2, Informix, open source, and other data experts to gain value from your data, share insights, and solve problems.
Join / Log in
Skip main navigation (Press Enter).
Toggle navigation
Content types
Announcements
Blogs
Groups
Discussions
Events
Glossary
Site Content
Libraries
Date range
on this day
between these dates
Posted by
User Groups
User Group Tag Test
Netezza Performance Server
DB2
Data and AI on IBM Z
View Only
Group Home
Discussion
102
Library
29
Blogs
10
Events
0
Members
275
Back to Blog List
Improved Backups on Netezza Performance Server
By
MICHAEL DEROY
posted
Tue November 03, 2020 12:00 PM
Options Dropdown
4
Recommend
Vaze
MICHAEL DEROY
Shruthi Subbaiah Machimada
Sowjanya Repala
We heard your feedback
. You wanted faster backups and we've delivered. For years, NPS has used the zlib library to compress your backups, but since NPS 11.0.4.1 (as a tech preview) we've added support for a modern compression library called zstd. Zstd was open sourced in 2015, and is now included in various file-systems, databases, and even the Linux kernel itself. Using zstd, you can expect improved backup speeds (up to 40%), Reduced SPU CPU Utilization (Up To 25%) and even better compression (Up to 10%). This functionality is currently limited to backup and restore from systems of the same size, but a future release will enable it to work between systems of all sizes. We plan on continuing to support the zlib compression format in the future to maintain compatibility with older backups.
Lets take the new zstd backup functionality for a spin with a simple test. In this example I'll be backing up a single table with just one backup thread. Since 11.0.4.1, you can now specify the compression library for performing backups by passing the '-compressionLib <zlib, zstd>' flag to the nzbackup command. Lets run a backup for both zlib and zstd libraries and compare the results.
ZLIB
ZSTD
Improvement
Backup Time
13.3s
8.3s
40%
Backup Size
292M
277M
5%
Backup time was reduced by 40% and the backup was 5% smaller in size!
ZSTD produces better compression, faster, and with less CPU utilization during the backup (check _vt_system_util during both backups to observe the reduced spu CPU utilization).
A bonus feature of using zstd is that we allow you to specify the level of compression you want when performing a backup by passing the optional argument '-compressionLvl <lvl>'. When running a zstd backup with a compression level of 5, the backup was slightly slower than our default compression level but produced a smaller backup. Be careful though; A compression level of 10 may result in a 20% smaller backup but be 24% slower than even our old zlib compression.
ZLIB
ZSTD
ZSTD Lvl 5
ZSTD Lvl 10
Backup Time
13.3s
8.3s
9s
17s
Backup Size
292M
277M
259M
240M
Our default zstd compression level performs backups faster than zlib, with a smaller backup and uses less CPU in the process. Give ZSTD compression a spin and let us know your feedback.
Relevant Documentation:
nzbackup -
https://www.ibm.com/support/knowledgecenter/SSTNZ3/com.ibm.ips.doc/postgresql/admin/r_sysadm_nzbackup_syntax.html
external tables -
https://www.ibm.com/support/knowledgecenter/SSTNZ3/com.ibm.ips.doc/postgresql/load/r_load_compress.html
*Note that the improvements mentioned in this blog are based on an exercise in a controlled environment in our lab, and compression ratios and rates will vary depending on the data.
0 comments
31 views
Permalink
Hybrid Data Management
Topic groups
HDM Global
Data Server Manager
Database Conversion Workbench
Db2 (On Premises and Cloud)
Db2 Analytics Accelerator for z/OS
Db2 Data Management Console
Db2 Early Access Program
Db2 Event Store
Db2 DevOps Experience Sponsor User Program Community
Db2 for z/OS
Db2 Tools for z/OS
Db2 Warehouse (On Premises and Cloud)
IBM Data Virtualization Manager for z/OS
IBM Query Management Facility (QMF)
Informix
Integrated Analytics Systems
Netezza Performance Server
Open Source Offerings
User groups
Events
Upcoming Hybrid Data Management Events
On Demand Webinars
IBM Expert TV
Virtual Community Events
All IBM Community Events
Participate
Post to Forum
Share a Resource
Blogging on the Community
Connect with Hybrid Data Management Users
All IBM Community Users
Resources
IBM Support
IBM Cloud Support
IBM Champions
Demos
Marketplace
Marketplace
Copyright © 2020 IBM Data Science Community. All rights reserved.
Powered by Higher Logic