Optimize fallocation

Summary:
Based on my recent findings (posted in our internal group), if we use fallocate without KEEP_SIZE flag, we get superior performance of fdatasync() in append-only workloads.

This diff provides an option for user to not use KEEP_SIZE flag, thus optimizing his sync performance by up to 2x-3x.

At one point we also just called posix_fallocate instead of fallocate, which isn't very fast: http://code.woboq.org/userspace/glibc/sysdeps/posix/posix_fallocate.c.html (tl;dr it manually writes out zero bytes to allocate storage). This diff also fixes that, by first calling fallocate and then posix_fallocate if fallocate is not supported.

Test Plan: make check

Reviewers: dhruba, sdong, haobo, ljin

Reviewed By: dhruba

CC: leveldb

Differential Revision: https://reviews.facebook.net/D16761
This commit is contained in:
Igor Canadi
2014-03-17 21:52:14 -07:00
parent ae25742af9
commit f26cb0f093
6 changed files with 86 additions and 37 deletions

View File

@@ -49,8 +49,6 @@ struct EnvOptions {
// construct from Options
explicit EnvOptions(const Options& options);
EnvOptions AdaptForLogWrite() const;
// If true, then allow caching of data in environment buffers
bool use_os_buffer = true;
@@ -61,13 +59,21 @@ struct EnvOptions {
bool use_mmap_writes = true;
// If true, set the FD_CLOEXEC on open fd.
bool set_fd_cloexec= true;
bool set_fd_cloexec = true;
// Allows OS to incrementally sync files to disk while they are being
// written, in the background. Issue one request for every bytes_per_sync
// written. 0 turns it off.
// Default: 0
uint64_t bytes_per_sync = 0;
// If true, we will preallocate the file with FALLOC_FL_KEEP_SIZE flag, which
// means that file size won't change as part of preallocation.
// If false, preallocation will also change the file size. This option will
// improve the performance in workloads where you sync the data on every
// write. By default, we set it to true for MANIFEST writes and false for
// WAL writes
bool fallocate_with_keep_size = true;
};
class Env {
@@ -260,6 +266,16 @@ class Env {
// Generates a unique id that can be used to identify a db
virtual std::string GenerateUniqueId();
// OptimizeForLogWrite will create a new EnvOptions object that is a copy of
// the EnvOptions in the parameters, but is optimized for writing log files.
// Default implementation returns the copy of the same object.
virtual EnvOptions OptimizeForLogWrite(const EnvOptions& env_options) const;
// OptimizeForManifestWrite will create a new EnvOptions object that is a copy
// of the EnvOptions in the parameters, but is optimized for writing manifest
// files. Default implementation returns the copy of the same object.
virtual EnvOptions OptimizeForManifestWrite(const EnvOptions& env_options)
const;
private:
// No copying allowed
Env(const Env&);