Cell Relay Archive

[mrityu <mkumar@acceleratednetworks.com>]

In article <3778E6C4.8471EA32@xedia.com>,
  Paul Koning <pkoning@xedia.com> wrote:
> mrityu wrote:
> > VPI/VCI values are actually managed in a switch. I mean what kind of
> > data structures do they use and what is the order of magnitude of
their
> > search time. Linear allocation is simple, but it also will have to
take
> > care of random allocations as in the case of PVC creation.
>
> It would depend on who wrote the code; there's no reason to believe
> there is any consistent pattern.
>
> (Well, it's likely no one is dumb enough to use linear search...)
>
> The obvious choice is by indexing: look up the VPI in one table,
> the VCI in a second table pointed to by the first.  Then again,
> some may use a CAM.  Some may use hash lookup.  Whatever works...
>
>

Hash is fine but I would believe majority of VPI/VCI requests will be
for next available pair(SVC, internal VCs), rather than a particular
choice(PVC). In this situation, won't it be better to leverage the fact
that you can allocate them incrementally, rather than going for hash?
I wanted to know if there is any way to leverage this, I can think of
some bad ones which are severe on memory requirements so they are out.

--
-
Accelerated Networks
Ph: 805 553 9680


Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.