When adding Extended Meta to an environment it is recommended that this be done slowly and incrementally. After each incremental change, users should let the new data work its way through the system to see how performance is impacted.
Specific things to look for include:
-
Database sizes and retention - Do database sizes appear significantly larger? Do size rollouts appear to be happening more frequently?
-
Query performance - Are the Extended Meta keys indexed? Are there any queries performing expensive operations, such as contains or regex, on these meta?
-
Aggregation performance - Are Concentrators and Archivers still able to keep up with aggregation? Are any aggregating devices falling behind?
There are two main techniques users can utilize when deploying Extended Meta in their environment:
Incrementally Increasing Extended Meta Sizes
While there is no configuration error in configuring a key from non-extended right to 4,096 (the maximum size of an Extended Meta key), it is better to gradually increase the size and find the minimum size that gives you the most value. Users may first want to increase the size of a single meta to 500, then maybe 1,000 and so on, and noting the impact that each step has on the environment.
Incrementally Adding Extended Meta
There may be multiple keys that users want to extend, however extending all keys at once makes it difficult to pinpoint problems if they may occur. Instead, users should Extended Meta one or two at a time. Thus, if the system is generating a particular piece of Extended Meta in quantities that are affecting performance, the user should know exactly which meta is causing the issue and then can either reduce the maximum size of that meta or remove it entirely.