On Mon, Mar 03, 2003 at 01:32:01PM +0200, Costas Dokolas wrote: > So, the answer to the questions "has that been done before" and "is it > possible with Java" are yes and yes, yes? > > Anyway, I envisioned a simulated network running simulated nodes each on a > thread, but that seems VERY resource hungry, especially when you must run at > least a few thousand nodes to get some results (even worse, assuming each > request/insert uses a thread). Did Serapis use this scheme or something > similar?
No, Serapis was a single threaded time-step simulator. It pushed messages from one virtual node to another, then allowed each an opportunity to act. It simulated network latency and throughput by scheduling the arrival of messages based on latency and message size. > > Doc > > > -----Original Message----- > > From: Scott Miller [mailto:scgmille at freenetproject.org] > > Sent: Saturday, March 01, 2003 12:47 AM > > To: devl at freenetproject.org > > Subject: Re: [freenet-dev] Simulations (related to the Stanford video) > > > > > > On Fri, Feb 28, 2003 at 01:27:10PM -0800, Tracy R Reed wrote: > > > On Fri, Feb 28, 2003 at 05:08:20PM +0200, Costas Dokolas > > spake thusly: > > > > So the question is, if I wanted to run some simulations > > how would I go about > > > > doing it? How was it done initially? Can it be done in > > Java? Is there > > > > something built for this purpose already? If there is > > nothing available, > > > > > > I have pondered using something like User Mode Linux, > > creating maybe a > > > hundred instances with maybe a gig of ds each (easy with > > 200G IDE disks > > > nowadays) and seeing what happens. Cpu speed, RAM, and disk > > IO would be > > > the problems but at least this would eliminate network > > bandwidth issues. > > > But there may not even be a need to run each fred in its own kernel > > > instances. Since it runs on random ports etc it seems like > > maybe a lot > > > could be run in parallel as long as they don't step on each > > other too much > > > resource-wise. > > > > > > > could something be done in Java, or is it not practical > > for some reason > > > > (like speed, or capacity)? > > > > > > <troll> > > > Java is not practical for reasons like speed or capacity > > but who's gonna > > > let that stop them from trying? ;) > > > </troll> > > > > > > <snip lots of good simulation ideas> > > > > > > What sort of simulation efforts have been attempted > > already? Seems like I > > > recall Ian or someone writing code to simulate certain > > situations but that > > > was more of a mathematical analysis, not a simulation > > running a bunch of > > > actual nodes. > > Actually the opposite. Serapis was the Java simulator back > > in the day, > > and it did precicely run a bunch of (virtual) nodes. We did > > manage to > > simulate a few thousand nodes. > > > > Scott > > > > _______________________________________________ > devl mailing list > devl at freenetproject.org > http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl > -- -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20030303/0ebfc664/attachment.pgp>
