This is a continuation of the earlier posts starting with this one: http://ravinote.blogspot.com/2020/09/best-practice-from-networking.html
Random writes perform just as well as sequential writes on SSD as long as the data are comparable. If the data size is small and the random writes are numerous, it may affect performance.
The use of garbage collector sometimes interferes with the performance of the networking server. The garbage collector has to be tuned for the kind of workloads.
In Solid State Drives as per descriptions online, there is a garbage collection process inside the SSD controller which resets dirty pages into renewed pages that can take the writes. It is important to know under which conditions the garbage collection might degrade performance. In the case of SSD, a continuous workload on small random writes puts a lot of work on the garbage collection.
A garbage collector will find it easier to collect aged data by levels. If there are generations in the pages to be reclaimed, it splits the work for the garbage collector so that the application operations can continue to work well.
Similarly, aggregated and large pages are easier for the garbage collector to collect rather than multiple spatially and temporally spread out pages If the pages can be bundled or allocated in clusters, it will signal the garbage collector to free them up at once when they are marked
Among the customizations for garbage collector, it is helpful to see which garbage collector is being worked the most. The garbage collector closer to the application has far more leverage in the allocations and deallocations than something downstream.
No comments:
Post a Comment