Alex Vandiver wrote: > On Thu, 2014-01-30 at 10:57 +0100, Michelle Sullivan wrote: > > >> My frontends are on completely different networks (and indeed different >> data centers) from the database servers for security (mainly DDoS type) >> reasons >> >> [snip discussion of optimization of column_info()] >> > > RT simply isn't designed to operate in an environment where every query > has tens to hundreds of ms of latency. Attempts to optimize this > particular set of queries will merely cause the limiting factor to > become some other set. While we're absolutely interested in patches > that help decrease the number of queries that RT runs (as you've noted, > there are several queries run multiple times), I expect you will find > this to be a frustrating game of whack-a-mole. If at all possible, > altering your network topology to remove the latency is most probably a > more straightforward way to have a performant RT instance. >
A thought (and a non too simple one I'll admit) .. how about developing a 'light-weight' RT front end talking to a heavy backend.. (might be a good reason to develop out the the REST interface ;) ) In the mean time I'm going to have to re-think the network design and hope that my employer doesn't just go 'we'll use a different interface' and force me onto CTS as I'm trying to get it deployed company wide and up-root the Salesforce and CTS Ticketing systems that are being currently used (which needless to say would result in a sizeable support/devel contract coming BPs way... If you want to know more about that message me privately/directly.) Michelle -- Michelle Sullivan http://www.mhix.org/