[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[TIGGE #LGY-600646]: Re: Missing fields from CMA
- Subject: [TIGGE #LGY-600646]: Re: Missing fields from CMA
- Date: Wed, 25 Apr 2007 11:16:55 -0600
Manuel,
> I am sorry, but ... where do you get 12 GB/h from ?
The 12 GB/h value was just an example.
> Let's use the real numbers I presented. We said we currently exchange
> 204 GB/day. We want to reserve enough bandwidth for resends, so we can
> say we exchange 204 GB/12 hours or 408GB/day:
> 408 / 24 = 17 GB/h
>
> So we are looking at a minimum queue size of 17 GBytes, to be reviewed
> each time a new partner starts using LDM or when any model increases
> resolution.
> Please, correct me if I am wrong...
You are right.
The important concept is that the maximum rate of data arrival
over the minimum residency interval is used to ensure that the
age of the oldest data-product in the product-queue never falls
below the minimum residency time. This value is most-easily
determined from the "rtstats" volume plots --- assuming that the
desired minimum residency time is one hour. For example, I
would choose the value 19 GB/h from the plot
http://www.unidata.ucar.edu/cgi-bin/rtstats/iddstats_vol_nc?EXP+tigge-ldm.ecmwf.int
to ensure that the age of the oldest data-product in the
product-queue was always greater than one hour.
The alternative is to accept a smaller minimum residency time
and, hence, permit a smaller queue. If this is done, however,
then the "ldmadmin" "start" and "restart" commands must
alway use the "-m" option to specify the maximum latency
allowed; otherwise, the LDM is likely to accept duplicate
data-products (as we believe we see with CMA due to the
small queue size).
Regards,
Steve Emmerson
Ticket Details
===================
Ticket ID: LGY-600646
Department: Support IDD TIGGE
Priority: Normal
Status: Open