[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Packetshaper and LDM/IDD data flows



Gerry Creager N5JXS wrote:
> 
> The Sitara system is hardware and software, and must be, because it has
> a fail-to-wire state.  If the box goes down, even a power failure, it
> fails so that traffic is passed untouched.
> 
> There are several alternatives surrounding determining whether some form
> of policing or shaping is enforced.  None of the "simple" ones like
> Packeteer are readily discernable...  About the only way to tell is if
> the system first remarks the IP TOS bits and then filters on those.
> Sitara does this normally; Packeteer does not.
> 
> One of the methods of policing Peer-2-peer traffic involves guessing
> what ports are involved and rte-limiting those ports.  I've got to check
> IANA and see if IDD/LDM has registered its ports.  I'd be surprised if
> not, but, if not, we need to get on the stick.  OK, so I should have
> checked:  388 is registered.  It should be a simple matter to get the
> network geeks to not block/filter/shape port 388 udp/tcp and we should
> be OK.
> 
> It will, however, be an education process.
> 
> If it'd help, I can serve as a liason, answering questions about LDM for
> the campus network guys... especially those involved in Internet-2.
> I'll also bring this up to the I2 networking folks at the Fall Members
> meeting in LA.  I suspect this can be managed.
> 
> One issue here is that most of the LDM users are computer-literate
> meteorologists.  I'm a somewhat meteorology-literate network engineer
> (and PI for the Texas Mesonet)... and one of the network geeks in
> Internet-2.
> 
> Gerry
> 

Gerry, 

Thanks for all the information!  This area of traffic shaping is new to
me.   I know nothing about the alternatives.  Could an end user tell if
Sitara is remarking bits?

We should be in good shape since, as you pointed out, we only really
care about traffic on port 388.  But, the general problem of shaping
traffic that uses a variety of dynamically allocated ports sounds hard. 
Reminds me of the problem of spam filtering - the risk of a false
positive (identifying something as spam that isn't) is often too high to
make it useful.   At least with the LDM we have some idea how traffic
should flow so if there's a problem it would probably be noticed.

If you would bring this up at the I2 Members Meeting that would great. 
And, if you're willing to serve as a liason, that's super!  I'll keep
that offer in my back pocket.  I'm waiting to see where this goes and if
more effort will be required on our (the UPC) part.

Anne  
-- 
***************************************************
Anne Wilson                     UCAR Unidata Program            
address@hidden                 P.O. Box 3000
                                  Boulder, CO  80307
----------------------------------------------------
Unidata WWW server       http://www.unidata.ucar.edu/
****************************************************