Ungodly packet loss rates

Dalvenjah FoxFire dalvenjah at dal.net
Mon Oct 21 20:41:31 UTC 1996


Edward Vielmetti put this into my mailbox:
> 
> You're transiting two NAPs (MAE-WEST and the DEC exchange (?)) which
> both provide service that's essential on one hand (wide access to
> the greater Internet) and lousy on the other hand (focussed access
> to a particular site).

Hmm. I recently encountered the same problem. I come from CERFnet at
UCSD, and I'm trying to get to a machine off of Sprint through MAE-WEST,
and experiencing the same (30%-40%) ungodly packet loss.
(telnet route-server.cerf.net; ping 206.197.121.4 and you'll see what I mean.
it seems slightly better now, but in the afternoons and evenings it's horrid.)

I phoned Sprint asking them to look into it - and I do have to say they
were very responsive and conducted extensive testing - and the bottleneck
turned out to be, surprise surprise, MAE-WEST.

> If you want good internet service to a particular point on the network,
> you must shop for it specifically and be prepared to continue to shop
> for it whenever the underlying network topology changes.

True. But silly young idealistic me would like to think that all ISPs,
exchange points, and transit providers would try to hold themselves to the
same high level of service that the phone company seems to.

Since all the discussion on what backbone (tier 1) providers can do to
improve interconnection seems to have done something - intra-provider traffic
seems to be much better nowadays than it was last year - can we start doing
the same with major exchange points such as MAE-WEST and MAE-EAST, etc?

If the problem is that the routers/gigaswitches are overloaded, perhaps
some sort of intelligent traffic analysis and division might be in order?

I.e. buy two more gigaswitches and interconnect them. Then, say that the
majority of X's traffic goes through Sprint and AlterNet, but very rarely
goes through MCI or BBN. So, X would go on the same gigaswitch as the Sprint
and Alternet connections, and the MCI and BBN connections would go on another.
If you somehow figure the best possible traffic layout, it might help the
situation till better and faster routers/switches hit the scene. Maybe this
could even be automated - instead of using completely dynamic switches,
use dampened switches that only switch if the traffic pattern changes over
a day or so.

Please forgive me if I'm being simplistic - much as I've learned from watching
this list, I don't know half as much as most of the pros out here about
networking/switching concepts. But I really think we should find a solution
to this problem. If you need a marketing reason, when people find out
that being at an exchange point is next to useless because they're
overloaded, the MAE/NAP operators might not be getting as many
customers they (and investors) might want. It *needs* to be fixed.

-dalvenjah

  Dalvenjah FoxFire, the Teddy Dragon (also known as Sven Nielsen to some :)
                dalvenjah at dal.net    ---     dalvenjah on IRC
     Remember: if you're not on DALnet, you're on the wrong IRC server!!
       (/serv irc.dal.net 7000 or telnet telnet.dal.net to try it out)
--
 ____       _               _      _ "I would've gotten away with it, too, if
|  _ \ __ _| |_ _____ _ _  (_)__ _| |_  it hadn't been for you meddling kids!"
| |_) / _` | \ V / -_) ' \ | / _` | ' \ 
|____/\__,_|_|\_/\___|_||_|/ \__,_|_||_| FoxFire -- dalvenjah at dal.net -- (SN90)
                         |__/





More information about the NANOG mailing list