| DannyH 2005-03-07, 5:46 pm |
| Hi,
I would like to know if its possible to configure a child squid to
query parents using ICP, as standard, but to return a HTTP 302
redirect to the end user browser with the URL for the preferred
parent.
The driver behind this is to develop a high bandwidth content delivery
network. Serving more data/bandwidth than any one box could proxy. I
would like to build the below architecture.
WEB-Browser <-> Child Squid (performs ICP request to N parents ,
sends back http Redirect)
WEB-Browser <-> Parent Squid (Parent host serves content direct to
browser without channelling data through Child node)
I am hopeful that the above will allow us to serve +4Gbit/Sec of data.
As an aside the files served will be on average 1Gbyte, around 2,000
of them.
The project is shoe horned into using HTTP as delivery, making use of
the Microsoft BITS service for download management.
The average client will be broadband, read 500Kb-1mb/sec and we expect
a total concurrent load of 5,000 users. Which due to the file size -
will be concurrent.
Other suggestions welcome, initially I had explored large SAN
infrastructure, though factoring IO consumption and number of FC
connections to web servers I'm convinced squid can provide more fully
redundant, better performing solution - backed by a smaller pair of
webservers and large NAS.
The entire solution may end up being replicated in many locations, a
rack per country etc with data-tier updates once per day for upto
10GByte of data.
Thanks for any help,
Danny
|