For best performance, mHUB should run entirely in memory without any disk access. Sufficient RAM is a key factor, as is the number of processor cores available for processing. See the Requirements chapter, above, for further details.
Any RDBMS in use could significantly have an impact on mHUB's performance, so the RDBMS should be tuned for maximum performance. Data should be hosted on fast disks, so solid state disks (SSDs) are highly recommended.
If data is being accessed across a LAN or WAN then performance can be significantly impacted. It is recommended that remote data is accessed only when necessary. If possible, periodically replicate data to the local machine instead of directly accessing the remote data.
Only load necessary data. For instance, if a database table has 30 columns but only 10 of these are actually used for matching, then loading the data from these 20 columns into a mHUB engine could be unnecessary and impact both memory usage and performance.
Should a process that uses mHUB unexpectedly crash, it's possible to configure the engine to create a debug log of all data loaded and all operations performed on the data. This can help identify problematic data, which can then be provided to 360Science for analysis, or cleaned/purged from the original data source. Refer to the Configuration Guide for details.
When a process that uses mHUB unexpectedly crashes, a minidump file ("matchITHub.dmp") will be created in the current folder. This file can be compressed (zipped) and submitted to 360science for analysis. Even a compressed minidump file can be many MBs in size, so please contact Syniti support and open a ticket to determine the most appropriate way the file can be provided. Note that no confidential data is stored within the file, it only provides information such as the location of the crash, the state of all threads and stacks, and the contents of local variables and processor registers.