[MUD-Dev] Text Parsing

[Petri Virkkula]

>>>>> "Matthew" == Matthew Mihaly <diablo at best.com> writes:

Matthew> On Wed, 9 Jun 1999, J C Lawrence wrote:

>> > We use an interpreted engine (which I realize is inherently
>> > slower)...
>> 
>> That's a dangerous and often massively misleading assumption.
>> Interpreted languages need not be either slow or performance
>> bottlenecks.  LP and BatMUD demonstrate this rather clearly.
>> LambdaMOO demonstrates rather clearly that this _can_ be a problem
>> given poor design and implementation.  BatMUD shows part of the
>> other side: that given good and careful work, the performance hit is
>> a strawman.

Matthew> Yeah, I was a wiz on batmud ages ago. That's what I was
Matthew> thinking of when I was questioning our player problem (as
Matthew> back when I played them, they'd have 250 people online, and
Matthew> people lining up to get in. No idea what it's like now). I
Matthew> just assumed that interpreted languages would be slower I
Matthew> guess becuase of the extra layer.

	Before the summer we (at BatMUD) usually had slightly over 300
	players online during peak hours, I think our record is
	somewhere near 320. We have on average around 170-200 players
	online (Jan: 173; Feb: 182; Mar: 188; Apr: 199; May: 182; Jun
	sofar: 169) [ref:
	<URL:http://www.bat.org/Batmud/statistics/1999/year-summary.html>].

	We have two 360MHz UltraSPARC CPUs, one of them is basically
	100% idling, and with 250-300 players online the another one
	is 50-60% busy. We estimate that we can handle around 400
	simultanous online players with the current hardware, and even
	then one of the CPUs would be almost 100% idle. If we could
	utilize both CPUs then we would perhaps have around 700 online
	players. We could probably have even 800-900 players online if
	we just had a better driver.

	Right now we have total memory available is 768M, the driver
	directly takes 164M of it and other processes take 87M. Still
	there is only 74M free RAM. Rest of missing 443M is used by
	kernel, and there probably mostly by disk cache. I think it is
	reasonable to say that our driver takes currently 650-800M
	RAM, even when eg. top says it takes only around 200M. For
	example, when available memory was reduced from 1280M to 768M
	we clearly noticed a slight slowdown.

	My suggestion is to add more memory, atleast that's what we do
	when we need better performance, and if that does not help,
	then we will buy more CPU power. And with so low memory prices
	we think that is far more cheaper to buy the additional memory
	than spent optimizing our driver (and yes, our mud is bloated
	before somebody start arguing about that). That however how
	you value your time (in our case our free time).


	Petri



_______________________________________________
MUD-Dev maillist  -  MUD-Dev at kanga.nu
http://www.kanga.nu/lists/listinfo/mud-dev