20001005: queuesize for NOAAPort system

>From: address@hidden (Rodger R. Getz)
>Organization: UCAR/Unidata
>Keywords: 200010051415.e95EFR103306

>We just installed a NOAAPort system from Planetary Data Systems and are
>receiving the NWSTG channel. Since the queuesize is so critical now with ldm
>5.1.2, what do you advise as the proper size of the queue. We are getting
>the NWSTG feed on our own system and not dealing with anything via IDD other
>than a WSI feed of two NEXRAD sites.
>
>Also, will there be any further details on how to use pqmon to finetune the
>queuesize?
>
>Thanks.
>
>Rodger R. Getz, President and CEO
>AWIS Weather Services, Inc.
>1735 East University Drive, Suite 101
>P.O. Box 3267
>Auburn, AL      36831-3267         http://www.awis.com
>ph: (334) 826-2149 ext 104 (voice)  (334) 826-2152 (FAX) 
>

Roger,

We currently filter the NEXRAD (SDUS5x) products out of the IDD (since they
are encrypted at present. These are pretty constant at about 150MB per hour.
The peak volume of grids is 200MB an hour. Looking at the rest is about 10MB
per hour. So from that, if you wanted to keep an hours worth of NWSTG 
plus a couple of WSI, I start between 450MB and 600MB. I don't think
this is particularly critical (unless you are short of disk space)
since the search, insertion and deletion of products is now so much faster.

With pqmon, you see an output such as:
nprods nfree  nempty      nbytes  maxprods  maxfree  minempty    maxext  age
33321   253  161738   798839200     38414      367    156854     23880 10157

The thing you probably want too look at first is age, which is the age
of the oldest product currently in the queue. If that is greater than
3600 consistently, then you know your queue is big enough to hold
at least an hours worth of data.

The next thing to look at is minempty. That is the minumum number of 
slots available in the queue. If that gets down to 1, then the LDM will have
to expire data to create storage slots in the data base- instead of
expiring data when space in the queue is needed. If that were the case, then
the queue would have to be created with the -S option for more slots.
Since the default takes in to account that the LDM is used or NOAAPORT data,
you should not run into any problems here. nprods + nempty is the
number of product slots that the queue could hold (= maxprods + minempty).
If minempty was very large, then the queue could be created with fewer slots.



Steve Chiswell