I would believe Facebook does have geographic distribution of their memcached instances as well as all other app and web servers.
Have you tried running a traceroute to their servers to see if there is some identifiable bottleneck between you and their servers? In our instance Facebook is 14 hops away on a fast cable modem on the US East Coast. Hops 11-14 balloon in response times to over 100ms average. If you get same route or end point hops with high latency this might be the issue: 11 FACEBOOK-INC.TenGigabitEthernet6-2.ar1.PAO2.gblx.net (67.17.162.38) 95.380 ms 95.452 ms 110.825 ms 12 ae0.bb01.pao1.tfbnw.net (74.119.76.132) 111.174 ms 111.260 ms 111.108 ms 13 ae5.br02.snc1.tfbnw.net (74.119.76.141) 110.847 ms ae5.br01.snc1.tfbnw.net (74.119.76.139) 110.763 ms ae5.br02.snc1.tfbnw.net (74.119.76.141) 110.832 ms 14 eth-17-17.csw01b.snc2.tfbnw.net (204.15.23.197) 116.560 ms eth-18-1.csw01b.snc2.tfbnw.net (204.15.21.125) 116.446 ms 116.315 ms On Wed, Jan 6, 2010 at 7:02 PM, Martin Bay <[email protected]> wrote: > How come sites like facebook does not place memcached servers around > the world with a live updated copy of their primary memcached servers? > From Europe the facebook website is EXTREMELY slow. >
