I am currently using TMM 2.1.6, to perform an auto-find mappings between a source and a target. The application was able to automatically find 200+ mappings within a few seconds, but when I try to save those mappings, it is taking approximately 30+ minutes to complete. Is there anyway for us to speed up the performance when the mappings are being saved?
For reference, our TMM repository housed on a network database and we will be unable to convert that to a local database.
15 seconds per mapping save is unusually slow.
The controllable things that can affect save performance are:
1) Do not run Teradata in a virtual machine (e.g. TD Express for VMware). This is the biggest problem that most people encounter.
2) Network latency (i.e. ping times) - public internet access or VPN access is typically slow. If possible, place the TMM client PC on the same network segment as the Teradata server. Although we have not certified any remote desktop configurations, Microsoft RDP appears to work very well when the user is geographically remote from the server, at least for a single user (non-shared instance of TMM) - where TMM is running on a Windows machine on the same ethernet segment as the TD server (or close as possible) and a user uses RDP to remotely connect to the TMM Windows machine.
3) Concurrent TMM users for the same repository database. This is usually not much of an issue unless someone is importing a lot of large files, or if you have a lot of very active users. TMM write operations to the repository start a transaction and lock appropriate tables in the repository for the duration of the operation (not including user response wait time).
4) Amount of metadata. This is usually not much of an issue, but if you keep all versions of all data representations and associated mappings such that you have 100K or more mappings in a repository, then that might start adding a second or two to each save (TMM has to check for duplicates before saving each mapping, where both combinations of data_element_id and related_data_element_id have to be checked in the Java code)... Archive old DRs and sets of mappings then delete them if your policies allow that.
Thank you Clark. I think our problem is probably network latency because our PC may not be on the same network segment as the Teradata server, and we won't be allowed access to use RDP into another machine. Do you have any other tips that we could use to speed it up?
Import of mappings uses a totally diffferent method to load the repository and is usually a lot faster than saving from an auto-match window if you have a lot of mappings. If you have enough mappings to save, you might try copying the mappings (or export them if that function is available in the auto-match window) to a TMM mapping import file and then import them.