Hi everyone, About a month back I said I would email the list with some measurements of RAM usage for tor in embedded systems running in the wild. These preliminary numbers might be of interest. Here's what I did. I ran tor in a ramdisk environment with only 3 binaries (busybox, tor and openntpd, statically linked against uclibc) on 1) a i686 box, 4 x 2.80GHz Xeon with 4GB ram (image at http://opensource.dyc.edu/pub/tor-ramdisk/images/tor.uclibc.i686.20090131.iso) and 2) a MIPS board (Mikrotik RB433AH) with a 680 MHz Atheros AR7161 and 128MB ram (image at http://opensource.dyc.edu/pub/tor-mips-ramdisk/images.ar7161/tor-mips-ramdisk.elf). After booting, I waited until the systems established themselves as relay only and directory server nodes in the network. I then monitored ram usage as time went on. Here's what I found: 1) node "simba" = i686 box with BandwidthRate 150KB BandwidthBurst 200KB Day Total(MB) Disk(MB) 7 246 30 9 247 31 12 249 31 16 255 31 19 258 33 21 261 33 Here total = total ram usage including paging and ramdisk, while disk = ramdisk only (mostly due to DataDirectory files) I did not systematically measure CPU usage, but it was very small. 2) node "mufasa" = mips with BandwidthRate 40KB BandwidthBurst 80KB Day Total(MB) 1 45 2 56 3 56 4 56 5 56 6 61 7 56 8 56 Given the different way in which the ramdisk was set up on the MIPS, there was no easy way to seperate paging from disk memory. Again, I did not systematically measure CPU, but watching top occasionally, I never saw loads over 0.1 or cpu usage over 10%. I realize these numbers are rough and incomplete, but they give a ball park of what's needed. I'm going to repeat these measurements, but would like some feedback from the community regarding what you'd like to see. -- Anthony G. Basile, Ph.D. Chair of Information Technology D'Youville College Buffalo, NY 14201 USA (716) 829-8197
Attachment:
signature.asc
Description: OpenPGP digital signature