[MapServer-users] Speeding up msDrawMap() for shapefiles

[Matthew Graber]

Hi Jukka,

Thanks for the advice. I was able to confirm through local testing that
using "DATA" is consistently faster than using the OGR connection (was
getting over 500 ms for OGR vs just over 300 ms for DATA). My results are
also suggesting that performing the shapefile optimizations may make my
request several hundredths of a second faster, I'm a little less sure on
that one. But regardless, it appears that the bulk of the problem with the
slow S3 connection is likely S3's overhead, so I don't think there's
anything more I can do with MapServer to speed it up.

Thanks again,
Matt

On Fri, Dec 8, 2023 at 11:58 AM Rahkonen Jukka <
jukka.rahkonen at maanmittauslaitos.fi> wrote:

> Hi,
>
>
>
> You should test the same shapefiles locally for understanding where the
> time gets spent. Normal time with simple styles is perhaps up to 300 ms.
> The optimization hints were written long before there were S3 or Amazon
> (the company).
>
> Unfortunately I don’t know how to optimize access to S3.
>
>
>
> -Jukka Rahkonen-
>
>
>
> *Lähettäjä:* MapServer-users <mapserver-users-bounces at lists.osgeo.org> *Puolesta
> *Matthew Graber via MapServer-users
> *Lähetetty:* perjantai 8. joulukuuta 2023 21.47
> *Vastaanottaja:* mapserver-users at lists.osgeo.org
> *Aihe:* [MapServer-users] Speeding up msDrawMap() for shapefiles
>
>
>
> Hi,
>
>
>
> I'm trying to speed up the time required for msDrawMap() to execute for
> rendering shapefiles. Currently, according to the mapserver logs, it's
> taking 2.5-3.5 seconds, sometimes up to 5 seconds for it to execute:
>
>
>
> [Wed Dec  6 18:43:47 2023].699791 CGI Request 53 on process 2641
> [Wed Dec  6 18:43:47 2023].717216 msDrawMap(): rendering using
> outputformat named png (AGG/PNG).
> [Wed Dec  6 18:43:52 2023].299580 msDrawMap(): Layer 1010
> (Shapefile_Layer), 4.582s
> [Wed Dec  6 18:43:52 2023].299792 msDrawMap(): Drawing Label Cache, 0.000s
> [Wed Dec  6 18:43:52 2023].299803 msDrawMap() total time: 4.584s
> [Wed Dec  6 18:43:52 2023].310853 msSaveImage(stdout) total time: 0.011s
>
> [Wed Dec  6 18:43:52 2023].310883 mapserv request processing time
> (msLoadMap not incl.): 4.611s
>
>
>
> The mapfile is currently set up using "CONNECTIONTYPE OGR" and
> "CONNECTION" to access the shapefiles from an S3 bucket as follows:
>
>
>
> CONNECTIONTYPE OGR
> CONNECTION    '/vsis3/s3_bucket_path/my_shapefile%.shp'
>
>
>
> I noticed at https://mapserver.org/input/vector/shapefiles.html that it
> says "Shapefile access is built directly into MapServer. It is also
> available through OGR, but direct access without OGR is recommended and
> discussed here."
>
> So I tried removing "CONNECTIONTYPE OGR" and changing "CONNECTION" to
> "DATA" as follows:
>
>
>
>  DATA    '/vsis3/s3_bucket_path/my_shapefile%.shp'
>
>
>
> I haven't been able to observe any definitive speed improvements by doing
> this yet, although I can't say for sure that there aren't any.
>
>
>
> Based on https://mapserver.org/optimization/vector.html, I also tried
> using the shptree command to create a spatial index and the coshp command
> to sort the shapefile and index. Even after doing that, I still see load
> msDrawMap() times of up to 5 seconds.
>
>
>
> Does anyone know of any other optimizations that could be made here? Or if
> not, would using shptree and coshp on the shapefile and then accessing it
> in the mapfile using "DATA" to take it from S3 be the optimal way to set
> things up?
>
>
>
> Thanks!
>
> Matt
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/mapserver-users/attachments/20231218/11603d09/attachment.htm>