Why are we upgrading servers again?I find myself asking this question as I look at the latest performance figures on Ford, the server due for upgrade Thursday morning.
I see we are currently getting a lock wait ratio of 1:10116 on the database, a lock wait means a query to the database cannot instantly be processed as it is waiting on someone else's query to unlock a table.
Any number over 5000 is considered to be a very good result
It is pretty clear that with a bit of load balancing between systems, we could provide very fast forums for a very long time before growth forced an upgrade.
But what I am thinking is that knowledge is worth paying for. $2000 on more memory hurts a lot, but sooner or later we would have to spend that money, and by spending it sooner we will get the opportunity to see three servers that are quite similar in power aside from the memory capacity work side by side with 8gb,12gb and 16gb of memory.
This should mean that as load increases, we can make a more intelligent decision on whether to buy another server, and if we do, what memory to use in it, or indeed whether we are better off throwing memory at existing servers.
Anything that stops us buying a new server saves big bucks on monthly colocation and engineering support costs, as well as of course the cost of the server itself.
When seen in this light $2000, does not seem so much money