It's calculated as follows: grep VmPeak /proc//status Since we’re using a large shared buffer (100GB), we've set the huge pages parameter accordingly so that the entire shared buffer fits into that. Using huge pages reduces overhead when working with large contiguous chunks of memory, as PostgreSQL does, particularly when using large values for shared_buffers. Here are the operating system specifications that we used for the test: Poorly configured OS kernel parameters can cause degradation in database server performance. It is imperative that we configure the OS parameters according to the database server and its workload. The CPU and memory consumption by the pgbench processes are also negligible. Hence, we can assume that the effect of network latency is negligible. We configured the pgbench on the same server as our database server instance. Pgbench -c $threads -j $threads -T 1800 -M prepared postgres We performed 30 minute read-write pgbench tests across different client counts (1, 16, 32, 64, 128, 256, and 512). The scale factor that we used for the test is 10,000, which generates approximately 150GB of data. In our case, we used the default scripts for the pgbench read-write test. However, you can provide your own scripts for the test. By default, pgbench runs a TPC-B-like script that includes simple UPDATE, SELECT, and INSERT statements. The most popular tool for measuring the performance of PostgreSQL is pgbench. In this post, we share the benchmark testing results that we’ve performed across different PostgreSQL versions (v9.5.21 to v12.1) on an AWS m5.metal instance.ģ.1 GHz Intel Xeon Platinum 8175 processors with new Intel Advanced Vector Extension (AVX-512) instruction setĭata is kept in a 300GB SSD with 2000 IOPS, and WAL is kept on another 150GB SSD with 2000 IOPS This article discusses the results of benchmark tests made using pgbench to test the performance of different versions of PostgreSQL on an instance of AWS m5.metal.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |