[Pdl] Custom queuing in PDL

Michael Hearne mhearne at usgs.gov
Thu Sep 6 15:08:22 UTC 2012


Group - After consultation with the Product Distribution development team, I will be taking on the development of a custom queuing module in PDL, which will be similar in functionality to the current DefaultIndexerListener.

The primary driver is the product for which I am responsible - PAGER.  For this product, I need something that can sort a long list of ShakeMap products in the desired order and then call PAGER with the products in that queue.  Recently we had an issue where a long list of ShakeMap products were stacked up, with some significant events in the back of some much smaller less significant ones.  It is my hope that this new module will solve my problem, and hopefully be useful to other product consumers as well. 

My strawman algorithm is this:

Sort all products P into two piles:
1) Events older than X minutes or hours
2) Events newer than X minutes or hours

The purpose for this first split is to ensure that older (and presumably less pressing) products are processed last.

Sort pile #2 in descending order by "impactfulness" (magnitude, maxmmi, etc.), and if there are multiple versions of a product for an event, select only the most recent version, under the assumption that more recent information is better information.

This module will be *completely* optional to use, and I will attempt to make it as configurable as possible, including:
 - product type P
 - threshold time X (used to separate products into two piles)
 - boolean to select only most recent version, or all versions.
 - impactfulness properties used (not sure about this one, but will see what I can do).

If anyone has any questions, suggestions, or concerns, please feel free to get in touch with me directly, or continue the discussion on this list.

Thanks,

Mike Hearne 

   


More information about the PDL mailing list