On Sat, 2005-01-01 at 08:59 -0600, Rich Adamson wrote: > > I'm not seeing these problem with X101P, nor does any of my (not so > > many) clients. And all that's boils down to is that: a. I'm lucky and b. > > I've have helped my luck by using only one card per machine, choosing a > > good MB and making sure the card don't share IRQ with anything. > > > > But the point still remains - any software or hardware that needs to be > > rebooted every 3 days to work is broken and should not be used. Period. > > > > If you use such software or hardware you should find out what's wrong > > and fix it or switch to something else. > > I think you've really nailed it on the head with the above statement. > The problem is that everyone keeps harping on shared interrupts, repeating > statements that others have made, etc, but no one is actually making any > attempt to identify the root problem. And, primarily because this seems > to be a very technical issue with digium cards only, where the majority > of developers and those with the skills necessary to identify the root > cause don't use digium cards.
WRONG. Those developers with the skills are the usually running Digium cards. The difference is we tend to use T1 interfaces and with risk mitigation, we don't put more than 1 card in a single machine. You will find many of us trying to talk people out of using more than 2 cards into a single machine regardless of the technology. > These systems (motherboards and all) don't have a problem with 99.9% of > other vendor's cards and drivers, but yet the * resolution constantly boils > down to things like swapping motherboards instead of identifying what the > issue really is in terms of certain digium cards. And 99.9999% of the cards out there don't require the low latency timing nor the quantity of interupt services. Maybe you should spend some time with people doing high power computing and listen to the grief they deal with in motherboards so they can get the most out of the Myrinet cards. People doing HPC know that anytime they have to cross the network they are losing a lot of CPU time especially when all nodes have to checkpoint before they can move forward. If you take anything from conversations with them, listen to the chipsets on the motherboards they use. One of the groups I talk to mentioned that by spending a few bucks more for a specific ServerWorks chipset than another Serverworks chipset meant a pretty significant difference for the cluster's performance. You might have noticed that many have noted here that Via is not a good chipset to use. You might find that Via chipsets are used in a lot of low end machines. Back when I had to do windows support, you could bet any machine with chronic crashes was on a Via chipset put together by someone trying to save money on every component. Please understand that at anytime you put a system together for something as important as your phones, you shouldn't scrimp on the hardware. So maybe the advice shouldn't be so much to swap the motherboard but to quit scrimping on it. -- Steven Critchfield <[EMAIL PROTECTED]> _______________________________________________ Asterisk-Users mailing list Asterisk-Users@lists.digium.com http://lists.digium.com/mailman/listinfo/asterisk-users To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-users