Beware of armchair critics...

dandellion posts the rebuttal to the armchair architecture rant about the upcoming changes to decentralize the asset server infra.

I won't comment on the validity of the rant - as I do not know the details of how the internals are done, I do not know whether it is true or false. However, if the content (the stuff that is present only on that sim) is brought closer to the sim itself, it can only benefit the things. Why ?

Because the sim needs to deliver the content to those connected anyway. (that is, until there is a usage of caps for it - which seems to be underway) - so the traffic anyway will pass through the sim server. So, moving the content closer to it can only improve the things. Ok, maybe it will add some load on the sim itself (uhm - it needs to read the files!), but then let's look at the way the sims are used currently:

There's an interesting sentence about updating 3 racks of servers, approximately 500 regions, which led me to do some math...

The biggest rack that I know of, is around 44U. "U" is a standard rack unit.

Assuming a simplistic 1U servers, this means approximately that one server currently hosts around 4 regions. (44*3 = 132, which when multiplied by 4 gives 528 - which is close enough to 500 regions).

So, moving one region off the server, and putting an asset server "piece" on it would give a huge net win in performance.

Of course, you do not have to put the asset server piece on *every* server - so one can quite easily get just a net performance improvement, despite of what the critics say :-)

Of course, the 1U servers is a bit of an assumption - there are bladeservers that should allow a better density - I'd be interested to see some samples, as I am a bit lazy to search now.