This doesn't add much performance per se but is great for minimizing tail latencies. If you outsource this to the OS, you have no idea how close or far you are from the limits, it doesn't really work. In some versions, you can switch internal execution strategies to take advantage of currently under-utilized resources. This is basically a fine-grained form of internal micro-backpressure, dynamically prioritizing subsets of the workload that minimize impact on the resources under pressure at any moment in time. This enables the database to dynamically adapt its behavior and scheduling decisions in response to specific resource pressures or excess resource capacity. By implication, the code knows exactly how much of which resources are instantaneously available at all times since it is directly managing those resources. There is a less obvious benefit in more schedule-driven database architectures. The obvious benefits to doing things this way are wide-ranging. Allocating all required memory during startup has been idiomatic for database architectures as long as I can remember.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |