|
On Thu, May 01, 2003 at 10:50:07AM -0700, ken price wrote:
> Yes, I'm very familiar with Super Sparrow, being one of the only projects of
> it's kind. In fact, our clusters ran Ultramonkey 1.0.2 and 2.0.0 until 6
> months ago - I've since switched everything to Keepalived. Maybe my
> knowledge of BGP routing is the weak point here, which is very possible, but
> won't if be a problem if I have (for example) one datacenter hosted with
> Sprint and the other hosted with AT&T? Super Sparrow seems to accomodate
> loadbalancing, but not real-time failover for thousands of clients hitting a
> single URL. Again, does anyone know if early expiration of a DNS record
> would help here? It would tremendously increase the load on the DNS servers,
> but may be worth it? Thanks!
> "Graham D. Purcocks" <grahamp@xxxxxxxxxxxxx> wrote:
> Ken
>
> I saw something on the UltraMonkey site which did this, Super Sparrow. I
> also found another, but not being too helpful, I can't remember the
> name. However, both projects appeared to start in 1999 and don't appear
> to be be actively developed. Both were Dynamic DNS solutions.
I did quite a lot of work on Super Sparrow a few years ago.
However I have been working on other projects since then.
And there has not been enough interest in Super Sparrow to
allow me to spend much time on it.
The way the system works is to use BGP (or other) information
to determine which of a pool of sites is closest to a
given end-user. It does not require BGP advertistments to
be changed - so you do not need to advertise .uk address space
in .us or stuff like that. Of course when sites go down
their BGP advertisments generally disappear, thus supersparrow
won't send any more traffic their way.
I won't go into more detail for fear of boring the list to tears.
There is an explanation of how the BGP stuff works at
http://www.supersparrow.org/ss_paper/html/node11.html
--
Horms
|
