INTERCEPTION CACHE
Interception caching has been an enticing idea for ISPs since it allows them to cache Web content in the network transparently to clients. With interception caching, routers or switches on the request path to the server divert the request to a proxy cache, which accepts the connection and send the response as if it were the origin server.
6.1 Problems in Interception caching:
To impersonate the origin server, the cache uses the IP address of the origin server as the source IP address of the response packets. Since this breaks the end-to-end principle of the Internet, interception caching has raised much controversy. The main concerns are the possibility that an interception cache may disrupt TCP
connections and that it deceives clients into assuming they interact with end servers, so that the clients may neglect adding appropriate cache control headers into their requests.
Connections can be disrupted when different packets from the client choose different paths to the server. For example, in Figure 6.1, assume that half way through the download of a page from the cache, TCP acknowledgements from the client will choose the upper path and start arriving at the origin server, which will discard them since it does not recognize the connection. In the meantime, the cache will not receive any acknowledgements and will eventually timeout the connection.
These problems may occur even without route changes, e.g., when OSPF routers forward packets using round-robin load balancing over multiple shortest path routes. In fact, with route load balancing, an inappropriate placement of interception cache in the network may disrupt every Web interaction going through the ISP. In any case, most ISPs are not willing to add extra connection failures, however few. Thus, they limit deployment of interception caches to only network points traversed by all packets from clients. In particular, they typically avoid interception caches in transit backbones and turn off caching for requests arriving from another ISP or from clients known to obtain Internet connectivity from multiple ISPs.
6.2 Interception cache in DHTTP:
DHTTP retains the end-to-end semantics even with interception caching. A DHTTP interception cache will intercept only requests sent over UDP and pass through any requests using TCP. TCP requests which may consists of either data uploads to servers or requests for non-idempotent resources, the kinds of requests that usually cannot benefit from caches anyway.So, restricting interception caches to UDP requests will not reduce cache effectiveness. When a DHTTP cache intercept a UDP request, it will either respond by a UDP packet, or establish a connection back to the client. In either case, it will use its true source IP address. At the transport layer, no IP impersonation occurs and all packets will arrive at each end regardless of routing path properties. Thus,DHTTP would enable a wide integration of interception caches into the Internet fabric, without any consideration of routing policies.
Further, the client is aware it speaks with the cache and not the end server since the response comes from a different IP address. A configuration option can thus be easily added to clients that would allow them to bypass all interception caches or selectively allow or exclude interception caches on certain subnets or for requests to certain Web sites
0 comments:
Post a Comment
Thanks for your Valuable comment