Netware Backup Speeds

[bombcan]

Hi Guys,

What are the fastest speeds that you guys have experienced with Netware and Legato.

Backup and restore. And if possible I would like to know your setup!

Greatly appreciated!

 

[Inforfean1]

You have left out a lot of information. What kind of tape library or drives, network bandwith, how much data you are backing, when you are backing, when you zre backing the data?
Your question depends on these factors and more? If, however, you are ambiguous about using Network, ask one of their sales person to give you presentation. May be it is for you or isn't for you?

 

[605]

NetWorker will not be the bottleneck - it will be the hardware because NW can do serverless backups (at a storage node). However, the network link (to update the databases with the file information at the NW server) and of course a huge number of very small files could be the bottleneck as this speed has to support the backup. Both processes run in parallel.

You can of course run benchmarks to test the behaviour.
For such tests, please read the "Performance Tuning Guide" which you can download from Legato's Support Pages.

 

[phbeauverd]

Our backup speed are as follow :

DATA: level=full, 55 GB 02:45:10 377059 files

 

[Novelli]

Hi,

we get a throughput per clinet of netware with about 18-20 MB/Sec, sometimes 30MB (big files)

Networker is on Solaris 9 and V 7.2.1
Client is 7.2 with combo patch
Netware 6.5 Sp4 + Sp5

Libary is a DL700 form EMC (Virtual Tape Libary)

Cheers
Andreas

 

[bombcan]

Thanks guys for your input. Greatly appreciated. Tell me do you use Netware Client 7.2 in a cluster environment? And does it work

 

[605]

Hi bombcan,

i think this question needs some clarification. Where you you expect a potential problem?

NetWorker does not really care whether the client is clustered, except for licensing .-).

The issue is how the virtual disks are presented to the OS, not to NW. As long as the OS can see it, NW can use it.

The only other necessary task is to start the NW client software (nsrexecd). But this again must be solved by the cluster solution - it can not be done by the NW client software itself.


A cluster scenario can of course be problematic ... for the NW server! But not for the client.

 

[bombcan]

Sorry about the confusion. What I meant is that we are using 7.2 client and our Novell environment is clustered. I beleive it is Version 1.8. As per Legato it is not supported. Now what I wanted to know is, is anyone using it configured for a cluster so that if the volumes fail over networker is aware and backs up as ususal? Hope that was a little clearer

Thanks

 

[JackyZhang]

So fast in your environment, I am trying to get rid of Legato, but I can't

because it is realy slow in NetWare box.

Some time the speed is 2000KB/S, and some time it is only few KB.




Jacky Zhang
CNE,MCSE

 

[Novelli]

@ Jacky,

what is your config ?

I had never eprformance problems with entworker and novell since we have the networker server on solaris.

At the beginnign we started with windows as networker server and 4 Novell backups put the 4 cpus in the backup server to 100% (with 4.22 client)

Now, and since the 7.2 client, we have no perormance problems, but restore problems :-(

 

[Novelli]

Hi again,

for the speed perfomance you should use on the novell side the following:

tsaup18 (at least or Sp5 for 6.5 where newer tsafs and so on are inside)

For tsafs you shoudl use the parameter: LOAD TSAFS.NLM /CacheMemoryThreshold=1 odr disabelign the cache for tsa.

This will ehance the speed, because networker does not use this function !

Add the bsdsock.nlm should be 6.59j and NOT newer !
there is a Bug with the newer versions of bsdsock.nlm in combination with libc.nlm and the networker that the datastream get corrupted and often retreyed.

If you are using solaris as networker server plattform, you should set there the following:

ndd -set /dev/tcp tcp_conn_req_max_q 256
ndd -set /dev/tcp tcp_conn_req_max_q0 2048

or both values higher....default is 128 and 512, which can causes problem if there are many sessions parallel.

For the network card and tcp settings there are also many possibilitys on the novell server..but the points up there should already help.

Cheers
Andreas

 

[JackyZhang]

I would try next week, I have scheduled Network maintenance next week.

Jacky Zhang
CNE,MCSE

 

[bombcan]

Great info Novelli,

Would it be possible to elaborate on the network card specifications on Novell side. As for the bsdsock we tested this in the lab and you are right it works better with 6.59j. Would you know if this is a problem relateds to the Novel NLM or a compatibilty between Novell and Networker. I'm trying to see if Legato needs to do something for it to work with the recent bsdsock.nlm

 

[Novelli]

@ bombcan,

it seams to be, that it has only indirect with the bsdsock.nlm to do, novell says, it has to do with the libc.nlm, nut i got also a new from December the 15. with the same problem in combination with the new bsdsock.nlm 6.63.01

EMC/Legato Eng. is working with novell Eng., together to fix that problem at the moment.

Btw, i got also a new patch from Friday for the saveset recover problem with smsut.nlm
look here, i collected everythign together but use at our own risk !

http://www.eckertweb.de/legato/legato.htm

Andreas

 

[bombcan]

@ novelli,

When you say the saveset recover problem. Was that the speed slowdown? What exactly was the problem?

 

[Novelli]

the problem was, that on a saveseset recover, the smsut.nlm consumed per 20 GB of restored date 100 MB in the ram of the novell server and never gives it free, also not if the memory was enought for the restore and the restore weas finished.

Tha last you sere from the novell server is a abend.

The last lgtps (see my page) solved it