Chunked Files: Optimize checking if all chunks are there

[guruz]

(not proven 100%)

We've seen benchmarks (TODO: where) where we have seen that for chunked files the later chunks take longer and longer than the first chunks. We're thinking isComplete of OC_FileChunking is to be blamed because it checks the file system for the existance of all chunks for the file. This gets slower (? probably because a succesful stat takes more time)

core/lib/private/filechunking.php

Line 76 in c5a200c

public function isComplete() {

One optimization idea: Have a fast path to check for currentChunk+1 (if it has that many chunks) and if that does not exist then return immediatly as false instead of counting all chunks.

FYI @dragotin @DeepDiver1975 @jturcotte @ogoffart @rullzer

[ogoffart]

Client issue showing that the more chunk, the slower: owncloud/client#4354

[PVince81]

One optimization idea: Have a fast path to check for currentChunk+1 (if it has that many chunks) and if that does not exist then return immediatly as false instead of counting all chunks.

Sounds like a plan.

[rullzer]

PR already in #22602

[rullzer]

One optimization idea: Have a fast path to check for currentChunk+1 (if it has that many chunks) and if that does not exist then return immediatly as false instead of counting all chunks.

Sounds like a plan.

After thinking about it we now just traverse the list from back to front (instead of front to back) and error out the moment we find 1 missing chunk.. But the idea is the same :)

[rullzer]

@PVince81 @cmonteroluque just FYI this helps with a blue ticket on the client.. so would 9.0.1 be an option? (PR is already available).

[PVince81]

done

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.