When talking OSD (Operating System Deployment): we speak about "DS" (Altiris Deployment Solution)...
But when speak about DS, must ask for "6.9 or 7.1" ?
About DS 6.9:
We already build with success a single DS 6.9, with 60 remote sites (or perhaps 55 only, but 2 times). So we can afford 60 secondary PXEs, connected to a single central DS.
Where is the maximal limitation ?
Well, in theory, only a total of 5'000 machines can connect the central DS server. I am pretty sure we can upper this number using Windows 2008.
Regarding numbers of remote sites maximum: a secondary PXE will only connect the DS central a single TCP session...
We often got a problem using DS 6.9 because of any change preboot files, immediate impact thousand remote sites, and overload WAN. The workaround is to "stop" the PXE config helper on each site, to start only during "maintenance hours" (out of business).
Notice the Symantec Management Platform Server can easier manage hundred of servers to automate start, stop services, on demand, or scheduled.
So with some care, not afraid to higher more 50 sites... But need to test carefully, and no idea where we should stop... (for sure 500! but for 250?)
About DS 7.1 (sp1):
The jump is not an upgrade, but a full "new" solution: because no more simple 2 tiers SQL/win32 application, with ability to manage 5'000 clients only.
The Symantec Management Platform with the Deployment Solution 7.1, is a huge 3-tiers MS-DotNet platform, can manage 20'000 to 50'000 (depending which source we read...), connected to a single Database/Management platform... I do not say, a single server there: for sure need additionnal servers, for external SQL store, package services, task services, and in particular for OSD the boot services (see http://www.symantec.com/docs/DOC3464)
Of course, the sizing will also depends of the other Altiris services/solutions used, or not (patch, delivery, inventory, etc...).
So the question here is, how many sites (and clients), before we need to up to a hierarchy of "SMP" servers. What can we do, using a single SMP (alias NS) ? For OSD !
At a 1st approach, it seems we got a bump up, being able to manage 5 or 10 times more...
At a second detailed view: We need to deploy a TASK (TS)+PACKAGE(PS) server on each remote site (to get a Boot Server: BS). We can deploy hundred of package servers. But only max 100 task servers, to connect a single NS... And the recommend is 10 max ???
So we can't build more than 50, and very max 100 remote sites, for a single CMDB/SMP/NS platform. Not more sites than DS 6.9!
Each BS (the site server with TS+PS), can manage 500 clients (1000 very max?): So I can in theory up to 25'000 computers, if was 500 on each sites...
A single NS if Less 50 sites to manage, if less 500 computers per site !
e.g. 15'000 computers: 1 HQ 5'000 (10 DS), 20 sites 500 computers (20 DS)...
If it is possible to consolidate a single TS for multiple sites, it is not really possible if need to deploy BS (OSD on PXE), must have 1 each site local.
In the real world, most of the time, numbered branch offices are small : eg. a central with 5'000, and 500 sites with 10 to 300 computers, with a total 10'000 remote computers, got a full total 15'000 computers. That is meaning, we MUST build a hierarchy with 10 NS, and a central report one, all the same only 15'000 clients, (normally can be a single NS, max 3 NS). Happily: Hierarchy is big improvement in 7.1 :)
More 50 sites to manage, need multiple NS in a hierarchy,
all the same only 15'000 computers to manage...
e.g.: HQ 5'000 clients (10 DS), 500 sites of 200 clients (10 NS, 500 DS)
If need to make large OSD, we must quickly go back to a more conventional preboot folder or USB or DVD instead of PXE, or huge hierarchy build of multiple NS...
Or... install 5 DS 6.9, and deploy 500 secondary PXEs !?
I think I will have to talk some guys at Barcelona, next month... :)