cancel
Showing results for 
Search instead for 
Did you mean: 
Highlighted

Media Multiplexing

I have a example that I want to share about Media Multiplexing;

PS-MPX - Policy Media Multiplexing                      
TD-MPX - Tape Media Multiplexing

Example

TD-MPX = 30

Policy A - PS-MPX = 2
Policy B - PS-MPX = 8

Policy A - 4 Jobs
Policy B - 10 Jobs

Stage 1 - Policy A kicks off: P(A) - Jobs 1 and 2 are active and Jobs 3 and 4 are queued.
                Policy B kick off 5 minutes later: P(B) - Jobs 1 to 6 are active, Job 7-10 are queued

Status: 8 Jobs active and 6 Jobs queued.


Stage 2 - Policy A - Job 1 completes, but Policy B - Job 7 becomes active. 

Status: 8 Jobs active 5 Jobs queued, 1 Job complete.

Stage 3 - Policy A - Job 2 completes, Jobs 3 - 4 will be still queued.  Policy B - Job 8 becomes active.

Status: 8 Jobs active 4 Jobs queued 2 Jobs complete.

Stage 4 -  Policy B - Jobs 1-6 completes - Jobs 9 and 10 become active.  Policy A - Jobs 3 and 4 are queued.

Status: 4 Jobs Active 2 Jobs queued and 8 Jobs complete.

Stage 5 - Policy B - Jobs 7-9 completes.  Policy A - Job 3 becomes active and Job 4 is queued.

Status: 2 Jobs Active, 1 Job queued and 11 Jobs complete

Stage 6 -
Policy B - Job 10 completes, Policy A - Job 3 still running and Job 4 becomes active.

Status: 2 Jobs Active, 0 jobs queued, 12 Jobs complete.

Stage 7 - Policy A - Jobs 3 and 4 complete.

Status: 0 Jobs Active, 0 jobs in the queue, 14 Jobs complete.

The question I ask, how is the media multiplexing working on the tape? 

The policy that hows the greater PS-MPX 90 % of the time will always finish first.  The lower PS-MPX have a difficult time in competing for tape resources.

I am looking forward to any response on this?



 

4 Replies
Highlighted

Bump

Have you got anywhere with this Ashleigh?

I've 'bumped' it as it was not visible due to this issue:

Thread Title Bug
Highlighted

I take it there was no job

I take it there was no job priority on this? Are you sure you have this config correct because Policy B should have had 8 jobs start straight away if your TD-MPX is set to 30.

This is odd!

Have you tried creating a new tape storage unit with one tape drive in it and trying again?

Another thought..... are all the jobs within the policy different clients? If you have not set the Max jobs per client on your master server high enough this could be it!
Highlighted

Yes, it looks like something

Yes, it looks like something is not configured correctly here if you have queued jobs. Max jobs per client would need to be set high if there are multiple jobs on single clients ...are you writing to a single volume pool ? How many tape drives do you have ? Have you correctly set the multiplexing value in the schedules ? I think we are missing information here to get a full picture and give an answer as there are multiple factors which affect how many jobs are queued and how many sent to tape simultaneously.
Highlighted

Two things

First - see my blog on just how it Writes to tape.
www-secure.symantec.com/connect/blogs/understanding-how-netbackup-writes-tape

Now with that understanding your question about jobs wanting a tape drive.

You have different options on how to set up the use of your tape drives.
When you setup a Storage Unit, you tell it how many tape drives to use at once.
For example we are going to say 2 drives.
You also say how many Streams can go to the tape drive at once.
For example we are going to say 10 streams.

So far we have 2 tape drives with 10 streams per drive - so right now we can have 20 active jobs at once.

10 jobs streaming to driveA and 10 jobs streaming to driveB

Now a job streaming to driveA completes.  A job that is queued will now to active and stream to driveA so it has 10 streams again.

However - you could be restricting this someplace else as well.

Look at the Policy you can limit the number of jobs a Policy can run at once.
If your policy says 5, and is using your Storage group with 2 drives - even though the storage group will allow 20 jobs to run, you Policy limits you to 5.

So now you are using 1 drive with 5 streams going to the drive.  (even thought the storage unit will allow you 15 more jobs if you wanted)

But let us say your policy has no limit set.......

Lets Look at the Schedule - here you say Media Multiplexing is set to 4.

So your storage unit would all 20 jobs
you are saying that any tape drive in use by this policy/schedule is only allowed to send 4 streams to a tape drive at once.

So now you are using 2 drives with 4 streams going to the drive. For a total of 8 streams.

However this can still be limited more some place else.

Lets look at the properties of the Master Server.
A) Under Global Attributes you have "Maximum jobs per client" and you have it set to 6.
B) your storage unit is set to 2 drives 10 jobs each for a total of 20 jobs.
C) You have no limit on your policy, and your schedule also says  20 jobs.
D) You have 5 clients in your policy, and each client will kick of 6 jobs for a total of 30 for jobs/streams that need to run.

client 1 will have 6 jobs go active (see A).
client 2 will have 6 jobs go active (see A).
client 3 will have 6 jobs go active (see A).
client 4 will have 2 jobs go active (see B & C)
client 5 will have all jobs queued (see A & B & C)

So that takes us to 20 jobs running.  As a job stream finishes, client 4 will have another job go Active.
If at any point client 4 gets 6 jobs active at the same time but you have less then 20 jobs running, client 5 will then have a job go active.

This "maximum jobs per client" is a way to spread the load.  If you have a server that might kick off 20 streams, you don't want all 20 running at once as it might bring the server to its knees.  So instead of having 1 server beating itself to death and all the other servers waiting, you can have a bunch of servers doing a bit of work and none of them overloaded.

But, there is still one more place you can limit this.
Say by default you do want each server to run 6 streams at once, but you have 1 very very small client, and 6 streams at once on this server would just kill it.  You can have the properties set to 6 for Global (see A) but you can limit this one server to just 1 stream at a time.

Again go to the properties of the master server, and look at Client Attributes.
Here you would add the name of your slow client and on the General tab, check Max data streams and set it to 1.
So all other clients would have a max of 6 streams running, and this small client would only run 1 stream at a time.

I hope this was helpfull.