Last year we started doing very high level work on the feasibility of centralising our Notes infrastructure with our current network. In summary, if we go central, Notes replication traffic - which represents a large part of our network traffic in Mars - will be drastically reduced. However, on the downside, Notes interactive traffic as users access their databases across the WAN would then go up - but by how much? We know the size of our pipes, we know how much bandwidth Notes takes up today. What we do not know is how much traffic would be generated once our thousands of databases have been centralised.
I spoke today to an IBM business partner, Trust Factory, who have developed an algorithm that will calculate based on #user sessions on each server, length of the user sessions & data consumption, how much traffic would be generated if that server was accessed across the WAN. This piece of analysis coupled with the information we already have on our network topology should give us some solid factual information about the feasibility of centralising our infrastructure.
Their customer list was shiny - Philips, Daimler, etc. etc., their demoes were impressive, their pitch just right. If we do decide to do anything along these lines this year, I guess I'll be talking to them.
Subscribe to:
Post Comments (Atom)
Posts
6 comments:
How did they compare to Permessa? Sounds like a very similar company...
They are very similar. In fact, the report they produce after their data gathering also contains the information we would be looking for in a LN DB discovery exercise. On the other hand, their information seems to be much detailed than what I've seen on DYS, although to be fair our version of DYS is as up-to-date as our version of Notes!
Our product knowledge is equally up to date. DYS changed their name to Permessa in May 2007.
Disk space is cheap. Desktop roaming is almost dead. Consolidate onto a couple a large unix clusters and tell people to use local replicas of the databases they use. I bet most people will have mail and maybe 3 other databases at most.
Move all the utility stuff to well designed web apps (room booking, finance, TMS) and we could make this work.
On the other hand, what is the real saving for removing one windows server from each site?
I disagree - disk space is NOT cheap, once you factor in using server grade disks rather than domestic ones, additional disks for RAID type redundancy, more disks if you want to go for SAN replication/clustering and then the cost of tapes and related backup infrastructure.
That's always the danger with centralising. Right now we are delivering reasonable SLAs on standard server hardware with basic (RAID) redundancy. If the server dies you have a site outage.
Centralising gives us the chance to use even less redundant (and therefore cheaper) hardware, cluster at the application level and provide superior SLAs.
The only reason we would look at 'enterprise grade' SAN solutions etc is if it reduces the overall TCO.
Of course our Processing friends may have other opinions on this topic :-)
Flame away...
Post a Comment