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

Re: LDM bandwidth

>To: address@hidden
>From: Joe Van Andel <address@hidden>
>Subject: Re: 20010531: LDM bandwidth
>Organization: ATD
>Keywords: LDM bandwidth

Hi Joe,

> We are considering using LDM to send S-Pol radar data for the IMPROVE-2
> project, this fall.
> 
> S-Pol will be located in Oregon, and will send data via a T1 line to the
> nearest university site in Oregon.  From there, we want to send the data
> to University of Washington, and also back to NCAR.  What kind of
> bandwidth can we expect over Abilene?  
> 
> That is, what's the typical transfer rate in kbytes/second between LDM's
> connected to Abilene?

First, it may be that the T1 line will be the bottleneck, not Abilene.
Two LDMs connected with a T1 line between NWS and NASA Goddard
transfer about 500 Mbytes/hour, limited by the T1.  From
motherlode.ucar.edu to Illinois or Penn State, we sometimes transfer
800 to 1000 Mbytes/hour over Abilene, but that's not limited by the
LDM, it's limited by the amount of data requested.  From FL4 to ML, we
have transferred 1.5 Gbytes/hour.

Another factor is RPC latency, since each LDM product requires at
least one round trip for the RPC call and return.  So while an LDM can
send 2000 products/second locally, the speed of light limits us to
around 30 products/second across the country, even using Abilene.  So
to get high bandwidth across the country, product granularity must be
sufficiently large.

I'm CC:ing Anne Wilson, our local LDM expert, so if I've got any of
these wrong, she'll correct me ...

--Russ

_____________________________________________________________________

Russ Rew                                         UCAR Unidata Program
address@hidden                     http://www.unidata.ucar.edu