Bent, I have been down the road that you are about to travel, only on a smaller scale (26 sites). It was however before TSM offered de-duplication but I am sure that's of little help regarding your initial backup. Some of our sites had 256k lines and a few had single T1's. All of them had some degree of latency issues.
After proposing several solutions to our management team, none of which were attractive for one reason or another the decision was made that we were to kick the initial backups off across the WAN. They could only run after close of business each day and we were held responsible for canceling them during business hours. Fortunately, they could run all weekend long or they may still be running. Eventually they completed but that was only the beginning of the nightmare. Daily backups would often run all night and bleed into business hours at which time management would get a call from either frustrated end users or our network team complaining about the processes bringing the WAN to its knees. I won't even get into the whole nightmare of performing a simple restore! After wrestling with the poor performance, customers at the sites screaming about the backups hogging the WAN, and the TSM Admins reminding how ugly recovery would be, the decision was made to budget for increasing the bandwidth. In the end I think management regretted not procure the funds needed to address WAN speed up front as they had to endure an enormous amount of criticism and bad PR due to the aforementioned issues. While I understand that none of this may be what you want to hear, or the solution to your immediate need, the bottom line is that the key to a competent centralized backup solution is to procure the tools that will make is operate efficiently. Sure increasing bandwidth across 100+ sites is a costly venture, but I will go out on a limb and guess that it's much cheaper and more efficient than buying 100+ backup servers, the associated licensing and storage for each location. Not to mention the man hours to manage them. On a closing note regarding our implementation; once management committed to, and budgeted for, the higher performing WAN links, and seen what it meant to the solution operationally their perspective changed. People began bragging about the solution rather than being constantly beat-up about it. ~Rick -----Original Message----- From: ADSM: Dist Stor Manager [mailto:ADSM-L@VM.MARIST.EDU] On Behalf Of Huebner, Andy Sent: Wednesday, January 16, 2013 12:14 PM To: ADSM-L@VM.MARIST.EDU Subject: Re: [ADSM-L] TSM 1st full backup of remote low-bandwidth nodes The only realistic solution to complete what you are trying to do over the wire is a client side de-duplication solution. In that case you can seed the backup server with local data. I did not understand the quantity of sites you had to deal with. Andy Huebner -----Original Message----- From: ADSM: Dist Stor Manager [mailto:ADSM-L@VM.MARIST.EDU] On Behalf Of Bent Christensen Sent: Wednesday, January 16, 2013 10:25 AM To: ADSM-L@VM.MARIST.EDU Subject: Re: [ADSM-L] TSM 1st full backup of remote low-bandwidth nodes Andy, I do not totally agree with you here. The main issue for us is to get all 107 remote sites converted to TSM reasonably fast to save maintenance and service fees on the existing backup solutions. With the laptop server solution we predict the turn-around time for each laptop to be around 2 weeks, which includes sending the laptop to the remote site, back up all data, send the laptop back to the backup center, export the node. With say 10 laptops this will take at least 6 months. We could buy more laptops but we cannot charge the expenses to the remote sites, and we are stuck with the laptops afterwards ... Disaster restores is a very different ball game. Costs will not be a big issue and we have approved plans for recovering any remote site within 48 hours, which for a few sites includes chartering an aircraft to transport hardware and a technician. - Bent -----Original Message----- From: ADSM: Dist Stor Manager [mailto:ADSM-L@VM.MARIST.EDU] On Behalf Of Huebner, Andy Sent: Tuesday, January 15, 2013 5:17 PM To: ADSM-L@VM.MARIST.EDU Subject: Re: [ADSM-L] TSM 1st full backup of remote low-bandwidth nodes You should use the same method to seed the first backup as you plan to use to restore the data. When you look at it that way a laptop and big external drive is not that expensive. Andy Huebner -----Original Message----- From: ADSM: Dist Stor Manager [mailto:ADSM-L@VM.MARIST.EDU] On Behalf Of Bent Christensen Sent: Tuesday, January 15, 2013 9:37 AM To: ADSM-L@VM.MARIST.EDU Subject: [ADSM-L] TSM 1st full backup of remote low-bandwidth nodes Hi, We are starting up a backup consolidation project where we are going to implement TSM 6.3 clients in all our 100+ remote sites and having them back up over the WAN to a few well-placed TSM backup datacenters. We have been through similar projects with selected sites a few times before, but this time the sites are larger and the bandwidth/latency worse, so there is little room for configuration mishaps ;-) One question always pops up early in the process: How are we going to do the first full TSM backup of the remote site nodes? So far we have tried: - copy data from the new node (include all attributes and permissions) to USB-disks, mount those on a TSM server (as drive X) and do a 'dsmc incr \\newnode\z$ -snapshotroot=X:\newnode_zdrive -asnodename=newnode'. This works OK and only requires a bunch of cheap high capacity USB disks, but our experience is that when we afterwards do the first incremental backup of the new node then 20-40 % of the files get backed up again - and we can't figure out why. - build a temp TSM laptop server, send it to the remote site, direct first full backup to this server, send it back to the backup datacenter and export the node(s). Nice and easy, but requires a lot of expensive laptops (and USB disks, the remote sites typically contain 2 to 10 TB of file data) to finish the project in a reasonable time frame. So how are you guys doing the first full backup of a remote node when using the WAN is not an option? - Bent