Network proximity calculation

Hello,

I’m trying to do a home grown load balancing solution that needs to
determine which of the two data centers a visitor should be sent to.
I’d like to do some type of a network proximity calculation that
includes latency and number of hops which will allow me to redirect the
user to their “closest/fastest” server.

I know all about CDNs and Appliance Load Balancers (F5); I’m looking for
a custom software solution instead.

Any pointers is much appreciated.

GP.

On Thu, Nov 20, 2008 at 11:00 PM, Grayson P.
[email protected] wrote:

I’m trying to do a home grown load balancing solution that needs to
determine which of the two data centers a visitor should be sent to.
I’d like to do some type of a network proximity calculation that
includes latency and number of hops which will allow me to redirect the
user to their “closest/fastest” server.

I know all about CDNs and Appliance Load Balancers (F5); I’m looking for
a custom software solution instead.

Any pointers is much appreciated.

traceroute -n -w 1 -q 1 $USER | awk ‘{sum += $3} END { printf
(“%8.2f\n”,sum)}’
:wink:

Depending on how much time you can burn up front, one option may be to
query the user’s IP address via a geo-ip-location; perhaps cache the
result based on either the user’s full IP address or class C subnet to
reduce the external API calls. Maybe traceroute and/or ping from each
data center and geo-locate, caching all the results?