Date: Fri, 19 Mar 2004 12:46:29 +0800 From: Erich Dollansky <oceanare@pacific.net.sg> To: Lanny Baron <lnb@FreeBSDsystems.COM> Cc: Olaf Hoyer <ohoyer@ohoyer.de> Subject: Re: Multiprocessor system VS one processor system Message-ID: <405A7B25.8040306@pacific.net.sg> In-Reply-To: <1079670664.33813.72.camel@panda> References: <20040318232348.BE86443D2D@mx1.FreeBSD.org> <20040319013145.P44321@gaff.hhhr.ision.net> <405A6537.2070607@pacific.net.sg> <1079670664.33813.72.camel@panda>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi, Lanny Baron wrote: > On Thu, 2004-03-18 at 22:12, Erich Dollansky wrote: > >> >>People tend to forget that the CPU clock rate of all Sun boxes is >>pretty low but the I/O bandwith is much higher than the memory >>bandwith of Xeon machine. >> >>Little things like changing a CPU while the machine is up and >>running is not known to PC based servers at all. >> >>A PC based server is good when you have to consider the money but >>will increase the risk of down-time. >> > > > That very strongly depends on what kind of pc Server you buy. > > A high quality Server with fully redundant power, dual or quad CPU's, > dual NICs, IPMI compliant, you will have one hell of a time showing me > downtime due to hardware. Yes, if the Server Board blows, it matters not > if it is a Sun<tm> or PC based. But then when it comes to mission > critical, a second lower cost or entry level Server as a backup or > failover will keep the 'lights on'. > No. I have seen no PC based hardware yet which keeps the lights on like the big machines. Even if a server board blows, the lights stay on. Everything is at least double. You can also have it four times in the box. It is just a matter of money. You will read it in the newspapers when people think that they can use PC based server in real mission critical applications. > The thing here is, how well does a company stand behind its product and > how quickly (or slowly) will you get looked after is very important. > The big machines just become slower because the failed hardware is not there anymore but they are still up and running. A good response time of the company behind brings only the performance faster up to normal level. As I said, it is the price you have to pay to get this. Those machines cost much much more than PC based servers but they also deliver what you need for real mission critical applications. You just can't change a motherboard on a PC based machine while the machine is running. You can change any component on the big machines without any problems while the machine is up and running. Erich
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?405A7B25.8040306>