05-16-2019 03:43 PM
Customer currently holds service metadbs in .mdb. Given problems associated with .mdb (exceeding file locks causing intermitent outages; ongoing corruption etc) we are looking to migrate them off to another database.
A few years ago they tried Oracle and it seemed to be slower than .mdb. Off course that could have been 'environmental' or similar.
Oracle is used as the enterprise db holding a lot of geospatial and non-geospatial data/processes.
Are there any recomendations or experiences whether best off holding service (mainly WMPS, some WMS and WFS) metadb in SQL Server express or Oracle, mainly considering performance and load?
Solved! Go to Solution.
05-16-2019 10:00 PM
What you tell is true, we have several customer where we decided to have metadata on Oracle in order to avoid problems that you comment on mdb. Oracle is slower and has the limitation that after an initialization of the publisher metadata WebMap keep it associated to an active session to oracle, that when tmed out forces to WebMap to do a new initialization and any getmap has the delay that metadata is not taken from cache.
To keep this alive in some customer we have a process monitor that do an initialization of services with oracle metadata every 30 minutes, in this way we try to ensure that all getmap request goes to metadata cached connection.
On other site we have tested with version 2018 SP2 to create metadata in postgres. The process to prepare metadata schema for each service, and populate it from GWS it's faster. We have done limited testing, but we have not ready begin to migrate existing oracle metadata to postgres, this is our idea. I have to test more if in this case the connection with postgres has some time out, and if this can be changed. In the case of Oracle I think the timed out comes from tcpip socket comunication that I think it's not simple to configure, it depends of network configurations.
By the way, looking to google I found this iinteresting input that is mostly related to the same problem that client (in our case WebMap) lost the connection to server. The problem is that WebMap takes some time for caching again metadata conection and read from GeoMedia library, which we try to avoid with the process that do iniitialization every 30 minutos.