ok i'll put my side of calculations, maybe they are wrong... but lets c. (firstly, B=Bytes, b=bits)
these are 10Mbit cards.. =10 Mb/sec =(10/8) MB/sec (byte = 8 bits..) =1.2x MBytes/sec which i think should then be the upper limit on a 10Mbps card...right? therefore a 1000KB/s transfer rates are about 83%. they are not fantastic agreed, but they arent that bad. but then how do you get 3/4 MBps?? or are you talking about 3/4 Mbps? if its 4 Mbits/sec i think the capacity utilization is darn low... if its 4 MBytes/sec... you've just set the world record. either way, you have lots of work to do (maybe set a patent for the combination of cards used ;-) just kidding, but well please anyone get me correct if i bundled up majorly somewhere.. affly robins On Mon, 2003-07-07 at 09:25, Jasmeet S. Virdi wrote: > Point 3/5: Am not sure whether u got the figures correct. But on my > win-win LAN, with 10 Mbps cards we get an effective rate of 3-4 Mbps > with intel Pros. R u sure about the lin-lin rates ? How did u go about > measuring this performance ? I am interested, mayb my calculations were > wrong !! > > -js > > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On > Behalf Of Robins Tharakan > Sent: Sunday, July 06, 2003 10:53 PM > To: The Linux-Delhi mailing list > Subject: Re: [ilugd] Software RAID performence > > > > yikes!! thats a *lo[t|g]* of questions.... > i'll try to comment on a few... > > 1. i think the celeron would be too slow to be able to extract the > performance (although i have *no* idea about the add-on card > performance). > > 2. i dont presume ram should affect, if the system only does raid. > although ofcourse ram gives better performance (but thats more of an > obvious point. more ram ofcourse means better caching/performance etc..) > > 3. 100M cards.. hmm.. i think that equates to a ~10Mbyte/s (per > interface) are you sure thats the performance you are aiming at? i mean, > i agree burst speeds are relatively better, but for an average speed, i > think this would be a bottleneck. > > 4. although i am unclear as to what is the effective purpose (dataset > calculations ok, but basically is it scattered data, or backup sort of > sequential data?) frankly, if heat is a trouble ( a single 40Gb 7200 rpm > is giving me worries at home!) i think you could easily go for a 4800rpm > (adding to which is cost savings). if ofcourse you are looking at > scattered data, 7200 is the way to go. > > 5. i have found good data rates of about 1000kb/s on 10Mbps (compex etc) > cards when transferring from linux-linux systems. (ofcourse win-linux > are generally lower [win9x-linux are far lower than win2k-linux]), so i > believe the low cost ones arent that bad for normal use. unless ofcourse > you have a chance of data corruption (bad wiring, lot of disturbance > etc...) > > 6. i used ext3 mainly because i didnt want to take a chance, and frankly > i hadnt worked on anything else as yet (then). but i have heard good > reviews (and dangerous warnings) about jfs and reiser... but i have no > idea.. (sorry just adding up my experience..!) > > affly > robins > > On Sun, 2003-07-06 at 22:35, Ambar Roy wrote: > > Hi, > > > > _______________________________________________ > ilugd mailing list > [EMAIL PROTECTED] > http://frodo.hserus.net/mailman/listinfo/ilugd > > > _______________________________________________ > ilugd mailing list > [EMAIL PROTECTED] > http://frodo.hserus.net/mailman/listinfo/ilugd > _______________________________________________ ilugd mailing list [EMAIL PROTECTED] http://frodo.hserus.net/mailman/listinfo/ilugd
