From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dan Cross Message-Id: <200101081942.OAA15636@augusta.math.psu.edu> To: 9fans@cse.psu.edu Subject: Re: [9fans] Re: A question of temperature... In-Reply-To: References: <3.0.5.32.20010106110703.01559620@mail.real.com> Cc: Date: Mon, 8 Jan 2001 14:42:19 -0500 Topicbox-Message-UUID: 4849fd58-eac9-11e9-9e20-41e7f4b1d025 In article you write: >Now consider a little program I load on my Win98 machine that simply >connects to an existing Plan 9 environment as a process server. The client >could define the amount of load it would place on the system. As long as >my machine is on it would be available to run jobs on. This would allow >you to use the processing power of systems which can't be Plan 9 boxes. Yes, but Plan 9 is an operating system, seti@home is an application, with a small and well-defined set of ``services'' (Actually, not really services, but I'm too tired to think of a better word right now): get a block, chunk away on it, send back a result. Plan 9 is, obviously, far more open ended. To get the same set of services while hosted under another OS and not compromising that OS or Plan 9, I'd probably have to run it in some sort of virtual machine. Consider a system call; I could trap it to, eg, Windows, but what do I do with rfork()? It's easier to build a virtual machine than work with the existing OS. This is what Inferno does. That's a pretty hefty screen saver. Also, what do you do when the user comes back? Terminate all the jobs? Migrate them elsewhere? seti@home probably doesn't care because it's embarassingly paraellel. Most applications, however, aren't set up to tolerate the lose of a CPU server. :-( - Dan C.