Currently, Time Series Collections have parameters which allocate a fixed amount of memory for the bucket catalog in bytes, which users often need to modify. However, these parameters remain fixed when clusters scale up, i.e. via Autoscale, leading to inefficient resource use. If a user scales down without adjusting the parameters, they risk OOM errors due to reduced overall RAM but extended bucket catalog usage.

Instead of setting a fixed byte value, users should be able to set a percentage that scales with the cluster. The affected parameters are:

• timeseriesSideBucketCatalogMemoryUsageThreshold
• timeseriesIdleBucketExpiryMemoryUsageThreshold

We should also evaluate other parameters that are expressed as percentages but only accept byte values, such as WT Cache, which faces similar scaling challenges.