nu.kanga.list.mud-dev

10672: Re: [MUD-Dev] memory and speed

>>>>> "Miroslav" == Miroslav Silovic <silovic@zesoi.fer.hr> writes:

Miroslav> Thus it stands to reason that if your playerbase is large, certain
Miroslav> aspects of VR (such as rooms, exits etc) should be coded in C. That
Miroslav> sacrifices flexibility, but... lag is no#1 most popular way of losing
Miroslav> players.

	In my experience lag does not cause players to stop playing
	your mud. Below is history of machines BatMUD has been running
	on:

Years	    Machine            RAM    CPU		   Memory  Peak
                                                           Usage   Players
==========================================================================
1990-199?   Sony NWS-1530      16M    Motorola 68030?      5M?     10
199?-1991   Sony NWS-1530      12M    Motorola 68030?      5M?     10-20
1991-1991   ?                  ?      ?                    11M     30-40?
1991-1992   Sun SS 2           56M    Sparc something      15M?    60?
1992-1993   Sun SS 2           64M    Sparc something      18M?    70?
1993-1994   Sparc Classic      96M    Sparc something      50M?    80?
1994-1995   SS10/51 clone      160M   SuperSparc 50MHz     90M?    120?
1995-1997   SS10/71 clone      160M   SuperSparcII 75MHz   110M    160?
1997-1998   Sun Ultra 2 clone  320M   UltraSPARC 200MHz    150M    240
1998-1999   Sun Ultra 60       1280M  2*UltraSPARC 360MHz  200M    290
1999-       Sun Ultra 60       768M   2*UltraSPARC 360MHz  220M    310


	During 1997 when we were switching to the faster machine, we
	had set our player limit to 120, but still got additional 40
	players in our incoming player queue. Players spent over 60
	minutes in the queue before they were allowed to play, they
	are allowed only use our chat channel while being in the
	queue. So our peak player count was not affected by lag, in
	fact players were willing to wait long times for their turn to
	play.

	Similarily during 1998 we had 240 online players PLUS 20-40
	players on queue. Wait time in the queue was usually over 40
	minutes.

	Peak player count raised immediately with the Ultra 60 to near
	300. But because our machine was very unstable, we started to
	loose players. Peak player count dropped to near 250. Then our
	machine was fixed by Sun, and we changed 1024MB unnamed memory
	to 512MB Sun's memory (memory sold by Sun has higher price tag
	on it).

	Now that the machine is stable again, our peak player count is
	over 300 again, though due the summer it dropped dramatically
	again. I expect to see around 340 simultanous online players
	after the summer in BatMUD.


	Petri


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