Locality of Reference...Routers - TCP-IP

This is a discussion on Locality of Reference...Routers - TCP-IP ; Hi, I havent got a a good answer for a question I have. If I design a network processor for a router, can I safely say there is lot of locality of references in dst ip addresses to put a ...

+ Reply to Thread
Results 1 to 3 of 3

Thread: Locality of Reference...Routers

  1. Locality of Reference...Routers

    Hi,
    I havent got a a good answer for a question I have. If I design a
    network processor for a router, can I safely say there is lot of
    locality of references in dst ip addresses to put a data cache? How
    does the answer change as we move from core to edge, if at all it
    does.

    If there are papers on this can somebody please point it out.

    GKN


  2. Re: Locality of Reference...Routers

    In article <1173307690.596123.70180@30g2000cwc.googlegroups.co m>,
    "gkn" wrote:

    > Hi,
    > I havent got a a good answer for a question I have. If I design a
    > network processor for a router, can I safely say there is lot of
    > locality of references in dst ip addresses to put a data cache? How
    > does the answer change as we move from core to edge, if at all it
    > does.


    Intuitively it seems like caching should benefit, and most
    high-performance routers make use of it. When you open a connection you
    typically send lots of packets as you upload or download data. These
    will be rapid-fire much of the time, so caching the routing decisions
    will save quite a bit of searching through the routing table for each
    packet.

    > If there are papers on this can somebody please point it out.


    Have you searched ACM SIGCOMM and IEEE Communications articles and
    conference proceedings? This seems like something that would have been
    researched quite a bit in the past couple of decades.

    --
    Barry Margolin, barmar@alum.mit.edu
    Arlington, MA
    *** PLEASE post questions in newsgroups, not directly to me ***
    *** PLEASE don't copy me on replies, I'll read them in the group ***

  3. Re: Locality of Reference...Routers

    On Mar 7, 2:48 pm, "gkn" wrote:

    > I havent got a a good answer for a question I have. If I design a
    > network processor for a router, can I safely say there is lot of
    > locality of references in dst ip addresses to put a data cache? How
    > does the answer change as we move from core to edge, if at all it
    > does.


    The industry has moved away from this type of design. In the real
    world, on the Internet, you need your maximum performance to deal with
    hostile traffic. Hostile traffic can easily use a large enough variety
    of addresses that caching is useless. So you need your maximum
    performance in a situation where caching is least effective. As a
    result, it's generally considered better to design a system that is as
    fast as possible over the entire routing table.

    Knowing this reasoning, you can see that there are some cases where
    this doesn't apply. For example, applications that don't have to deal
    with attack traffic.

    DS


+ Reply to Thread