This post is more than 5 years old
20 Posts
0
3715
Limiting Number of Clients that Run Concurrently
Hi There - Please bear with me as I'm new to Networker having come from a Netbackup and Dataprotector background. I've spent a few days reading through config guides and whitepapers and, whether or not its the terminology, or my inability to spot it, ive been looking for an option for a policy whereby I can limit the number of clients that can run concurrently. i.e.
If I create a policy which contains 100 servers that all use the same directive and have a window that allows backups to run between 18:00 03:00 can I limit the total amount of servers that can run at any one time within that window to say, 25?
My ultimate goal, within this current environment is to lessen the "policy sprawl" and create fewer policies with more clients as a pose to what is happening at the moment, many policies with fewer clients.
your input is appreciated.
Regards,
Storage_Monkey (aka BackupBoy)
LilianaW
40 Posts
0
January 20th, 2015 04:00
Hello BackupBoy,
Number of policies in NetWorker is one of the impacts for people coming from NetBackup world where you used to define policies for OS type for example and finished with few ones. NetWorker have many flexible options and you need to learn how to efficiently use them to acquire administrator and site needs.
Specifically to your request you can use "parallelism" attribute which come at several levels ( server, client, group...)
I would pay attention to best practices on parallelism and number of clients on policies. take in consideration also the recovery needs when you test and configure parallelism and multiplexing.
You didn't supply the NetWorker version but you can find all the relevant information on configuration and best practices in "NetWorker Administration Guide" and "NetWorker Performance Optimization Planning Guide " documentation.
Valuable documents can be found at support.emc.com as well.
Regards,
--Liliana
Storage_Monkey
20 Posts
0
January 21st, 2015 07:00
Hi Liliana - many thanks for your swift response and pointing me in the right direction. To clarify (and I will get around to reading those recommendations with specific focus on parallelism) if I have 50 servers in a group and I set 10 to run in parallel the other 40 will not run until those 10 complete? or 10 will run in parallel and the other 40 will not?
Rgds
LilianaW
40 Posts
0
January 22nd, 2015 03:00
Hello BackupBoy,
I supposed your question is when you are setting group parallelism to 10 save streams simultaneous. All others will wait for a slot to run.
Proper groups parallelism values ensure optimal operating system performance.
Group parallelism ( Configuration > Group Properties > Advanced > Savegrp parallelism) defines the number of simultaneous save streams that a scheduled backup can generate for a NetWorker group. The default value for this attribute is "0", which means that the attribute has no effect on other parallelism settings.
When Savegrp parallelism attribute is set to "0", the savegroup process may start as many clients in the group as possible, up to the maximum number allowed by the server parallelism setting defined for the NetWorker server.
If the value is greater than zero, the parallelism value will override any other parallelism considerations the backups in the group would use. You need to find the best approach for optimal performance.
Talking about Server parallelism simple rules for setting server parallelism are difficult to develop. Server parallelism is the maximum number of savestreams that can be backup simultaneously in a data zone including server and client initiated backups.
For a client file system backup particularly for example, the master daemon nsrd (The "Big Boss" ) monitors the number of server processes engaged in managing incoming savestreams . As the group backup operation executes , savegrp continues to prepare worklists and start nsrexec daemons, one per client machine , to manage server/client communications . The nsrd compares each “save” request for a nsrexec with the number of incoming savestreams. If the current number of savestreams is less than the value of the server’s Parallelism attribute , nsrd grants the save request and savegrp start another nsrexec daemon.
Extending the performance discussion to the multiplexing area ,there is a relationship between the value of server parallelism and the degree to which an administrator can exploit the backup efficiencies of multiplexing. Since server parallelism defines the number of client savestreams that are accepted at one time , it also defines the number of savestreams available for multiplexing to devices.
Like an example you could set Server Parallelism value = Total number of target sessions (for all devices).
Proper client parallelism values are important as well. For regular clients, use the lowest possible parallelism settings to best balance between the number of save sets and throughput. For the backup server, set highest possible client parallelism to ensure that index backups are not delayed.
Parallelism is a general term within the NetWorker software for a number of configurable options that allow you to adjust the volume of data that a system processes, which can improve the performance of servers, storage nodes, and
devices. There are several standard NetWorker performance features that can be configured to enhance performance in any environment. You need to identify the trade-offs at your site and select optimal values for these NetWorker performance features. Know your site needs, understand how NetWorker works and change gradually
Additionally to admin guides, you can follow the following link for valuable information as well:
http://nsrd.info/blog/?s=parallelism
Hope that help
Regards,
--Liliana
Storage_Monkey
20 Posts
0
January 23rd, 2015 08:00
Liliana - Thank you for taking the time to make such an informed and detailed answer.
Regards,
Chris