Most block-based filesystems (block-based meaning the data on the disk is stored in blocks) have latency of around 10μs (microseconds). While this is significantly slower than the throughput of a SSD, this typically is not a bottleneck as most applications don’t read or write that much data each second. OSXFS has a limit throughput of 250 MB/s. Modern systems with SSDs can achieve throughput up to at least a few GBs/second. Latency is the time it takes for a file system call (e.g. Throughput is the rate at which data can be processed. There are two dimensions to filesystem performance: throughput and latency. On Linux, Docker can mount directories and files directly, whereas on Mac, Docker has to pass the request to Mac (osxfs) for each file read/write. The root of the problem lies in the OS file system layer between Docker and Mac.
So when I read in the official roadmap that they will be improving performance, I literally couldn’t wait! Why is it so slow? There are plenty of online topics related to this problem: Docker for Mac has some long-lasting performance issues (ever since the launch in 2016) due to the way volume file sharing works.
0 kommentar(er)
