Paul Mach wrote:
> You're right, I'm an idiot. It doesn't have to run the code, it just
> has to look for function headers. But 14000 lines for a simple issue
> assign does seem like unneeded overhead.
I think you should really take a look at the PHP architecture before you
go any further with claims of that nature. A 14000 line (cached)
sequential read is noise against the time of the first query.
I really shouldn't guess how much more likely those static PHP files are
to get served from the buffer cache than any given DB block needed for a
nontrivial issue list, but it's a lot. Then there's lock contention
over the hot DB blocks, IPC overhead, etc. No comparison.
> It is a laptop. But any modern operating system will cache frequently
> accessed files in memory ( I have enought), so HD speed will not be
> the problem. I would attribute it to bus or processor speed if
Unless the bus is already saturated from other sources or you are
full-text searching a list with a few million issues, I can't fathom how
bus contention or saturation could be an issue for a lightweight app
like Eventum. That is indeed less true on a laptop where all resources
are continually squeezed, but you seem to agree running a production
installation on a laptop verges on insanity.
Depending on your OS, all of your postulates should be trivially easy to
test. If your tests show anything significant about the resource
utilization of Eventum, I for one would be very interested in that. As
far as I'm concerned, it can't go too fast.
> I would never run a serious installation off a laptop either. But if
> it is running slow with just little old me using it, what will happen
> when there are lots of people on a legit machine?
I don't know what the largest usage load Eventum has had to date is, but
when we go live with ours this weekend, I'll wager that we'll be making
a pretty valiant run for the title. I am feeling pretty good about that
decision right now.* If my opinion is different on Monday, you'll read
all about it right here.
> Also, I don't have
> the resources to allocate a dedicated machine just for Eventum.
Nor do we.
I agree that Eventum is not without performance issues, but they are
hardly show stoppers. But more to the point, I agree fully with Joao's
observation that any such problems are of a character and complexity
unlikely to be unraveled by blind guesswork.
*Joao and Bryan, I'll be feeling *really* good about it if you guys can
possibly manage to squeak 1.5.1 out tomorrow and it has the email fixes