I have backups spread during the week (well, my DB backups are full each day, but file system backup is full once per random day in a week). I equally spread the load so I end up with 45TB per day for some 800 clients. We use concept of database hotels (power machines hosting 60+ databases) and if it wasn't for that I would have much more clients. I think when it comes to scalability and performance, your biggest issue is not volume, but rather number of sessions.  I have per day 37k sessions. That's not a small number and unless you can spread your backups around the clock instead of using silly backup window paradigm - then you would have problem. At least I would and most likely I would end up with 2 datazones. In my case server is HPUX 11.31 running on VPAR with 8 CPUs assigned and 32GB. I should probably say that this is dedicated server doing only that role and saving its own data and nothing more. Media servers have much more CPUs of course.

Hi Hrvoje,

Thank you for your answer.

I read somewhere that the number of session per client is 4 by default so if we have 1100 physical clients it gives only 4400 sessions, we should add to this the client for virtual cluster resources and other specific client. In our enviroment we should end with 10k sessions per day.Does it seem correct to you ?

So if NetWorker can handle 37k sessions with one server, can I say that we could have up to 3 to 4 thousand physical clients on a single Neworker server ?

4 is number of parallel sessions by default.  Actually, that used to be the case before, then it was increased to 8 and not sure what is the case with current builds.  What you should keep in mind that this is just pure parallelism value - meaning, it says how many parallel session client will run at any time. By default, you will have more sessions of course per client.  Each session is one file system so, top of my head, on UNIX that would be /, /usr, /var, /tmp, /opt plus some other specific to some UNIX distros and any other mnt point you might have.  On Windows is your drives plus some 5 VSS save sets at least... so if you use parallelism of 4, you will see 4 of them run at the same time and as soon as one is done, the queued one will jump in. If you have databases, you will have archive/transaction log backups.  We do them every 1h and have some 300 various databases. That adds to the count considerably as each log may be one session (I do that what SAP for a reason). In my world, I spread backups around the clock and have no issues if they run during working hours since none of my backups last more than an hour (except few fat databases, but they are backed up through proxy using snapshots so no big deal). If you have 1100 clients which will do pure file system backup, I expect you will have some 6k-7k sessions at least per day. That is not much, but it all depends then how big they are and into what time frame you wish to squeeze them. I personally find that ideal limit above which I would switch to second datazone is around 400 clients. After that you will notice impact on backup server. How much depends on horse power and how this load is presented to server itself - there is simple answer there really. Life is far away from ideal, but if I had 1100 clients I would probably split that between two datazones and that way have enough buffer for future growth.