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

20030218: LDM 6 release candidate tests at OU/CAPS (cont.)

>From: Kevin Kelleher <address@hidden>
>Organization: CAPS
>Keywords: 200302172141.h1HLfK425607 LDM 6 CRAFT

Hi Kevin,

re: 100 radial product size
>  This was tested with the older version of LDM and the BZIP2
>compression.  Might be interesting to see if the new version of LDM
>behaves differently wrt optimizing packet size??

Yes, this is basically the jist of my previous comments.  Sorry to have
been so obtuse.

LDM 6 is _much_ better/more efficient at moving large/larger data
products than LDM5.  This is the main benefit of moving away from use
of blocking RPCs and the 16384 byte BLOCKDATA "chunking" size when
sending large products (in LDM5 this size was 16384 bytes).

Cheers,

Tom

>From address@hidden Tue Feb 18 09:31:10 2003
>CC: Kelvin Droegemeier <address@hidden>, address@hidden,
>   Kelvin Droegemeier <address@hidden>, address@hidden
>Subject: Re: 20030218: LDM 6 release candidate tests at OU/CAPS (cont.)

Tom,

Yes, I was a few emails behind when I sent this, but did get to the
email in which you explained the situation very nicely.  Turns out
Kelvin's point is the data transmission appears to be faster than the
antenna rotation/data collection capability of the radar.  So keeping
the number of radials small is helpful in delivering the data quickly.

I would think we could look at sending a single elevation at a time and
checking the latencies.  For example:  * during a storm the radar scans
14 elevations in 5 min or about 20 seconds per elevation. So it seems
to me waiting for the elevation to finish before sending the packet
would be reasonable.  * during clear air mode (the other extreme) the
radar scans only 5 elevations in 10 min or about 2 min per elevation.
So in this case, the data could be 2 min old.  However, in clear air
there usually isn't much weather to worry about, so does it really
matter??

Kelvin - your thoughts?

Thx.
KK

>From address@hidden Tue Feb 18 10:27:47 2003
>Subject: Re: 20030218: LDM 6 release candidate tests at OU/CAPS (cont.)
>Cc: Unidata Support <address@hidden>,
>   Kelvin Droegemeier <address@hidden>, address@hidden,
>   Kelvin Droegemeier <address@hidden>, address@hidden

Kevin,

It would be great to test these ideas in the context of the new LDM.  A
latency of 10 seconds is viewed by some as being very important, so I
want to be careful here.  I think the best strategy is to test various
possibilities to see what the results look like.

Thanks!!

Kelvin