oddur at ccpgames
Aug 2, 2012, 8:07 AM
Post #1 of 1
Ranged Backend Requests for WAN acceleration
We are evaluating varnish to build an internal content delivery network for our company.
We have big http resources which we need to accelerate the delivery of from backend which have really high latency. To work around the latency issue we´d like to use ranged http requests to request the resource in multiple ranged chunks.
Would it be possible to have varnish request the resource from the backend in multiple concurrent chunks from the backend ?
I´ve found this way of doing this: http://devblog.seomoz.org/2011/05/how-to-cache-http-range-requests/ which would work for us if the http client always requests the same ranges, which we cannot guarantee.
The workflow I´m looking for is:
1. Client makes a request to varnish for a resource
2. In the case of a cache miss varnish does a vcl_fetch
3. The vcl_fetch uses multiple ranged requests to the backend for the resource and merges them into one cache entry and serves to the client.
I should note that we are using the streaming support as well.
Oddur Snær Magnússon
Senior Programmer | Reykjavík | Iceland
[Description: Description: Description: Description: cid:DB5C0F43-387B-4CF8-BC6F-B17512BB8251 [at] ccp]