[
https://issues.apache.org/jira/browse/CAMEL-13399?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mykhailo Kozik updated CAMEL-13399:
-----------------------------------
Description:
I have a simple route which runs by demand and archives multiple files in one
zip archive.
{code:java}
from(file:/path/to/source)
.aggregate(constant(1), new ZipAggregationsStrategy(true, true))
.completionFromBatchConsumer()
.eagerCheckCompletion()
.to(file:/path/to/target){code}
It works fine when the number of files in source folder is relatively small.
After adding tracing logs to test size of input files / time taken by process,
the following chart could be drawn.
!Screenshot 2019-04-08 18.41.10.png!
That means, to make zip archive from 500mb of files takes over 12 minutes!
Looks like in order to add a file, camel extracts zip archive to input stream,
put file inside it, and build zip archive again. So that becomes near quadratic
complexity, and not acceptable for large folders.
The workaround is to add completionSize or completionPredicate to flush every
100mb, so we got all files archived but splitted into several archives, which
works but not the best choice.
Is there a general solution how to make ZipAggregationStrategy to work in near
linear time, so the process does not become slower with large number of files?
was:
I have a simple route which runs by demand and archives multiple files in one
zip archive.
{code:java}
from(file:/path/to/source)
.aggregate(constant(1), new ZipAggregationsStrategy(true, true))
.completionFromBatchConsumer()
.eagerCheckCompletion()
.to(file:/path/to/target){code}
It works fine when the number of files in source folder is relatively small.
After adding tracing logs to test size of input files / time taken by process,
the following chart could be drawn.
!Screenshot 2019-04-08 18.41.10.png!
That means, to make zip archive from 500mb of files takes over 12 minutes!
Looks like in order to add a file, camel extracts zip archive to input stream,
put file inside it, and build zip archive again. So that becomes near quadratic
complexity, and not acceptable for large folders.
The workaround is to add completionSize or completionPredicate to flush every
100mb, so we got all files archived but splitted into several archives, which
works but not the best choice.
Is there a general solution how to make ZipAggregationStrategy to work in near
linear time, so the process does not become slower with large number of times?
> ZipAggregationStrategy become slower when size of zip grows
> -----------------------------------------------------------
>
> Key: CAMEL-13399
> URL: https://issues.apache.org/jira/browse/CAMEL-13399
> Project: Camel
> Issue Type: Bug
> Components: camel-zipfile
> Affects Versions: 2.23.1
> Reporter: Mykhailo Kozik
> Priority: Major
> Attachments: Screenshot 2019-04-08 18.41.10.png
>
>
> I have a simple route which runs by demand and archives multiple files in one
> zip archive.
> {code:java}
> from(file:/path/to/source)
> .aggregate(constant(1), new ZipAggregationsStrategy(true, true))
> .completionFromBatchConsumer()
> .eagerCheckCompletion()
> .to(file:/path/to/target){code}
> It works fine when the number of files in source folder is relatively small.
> After adding tracing logs to test size of input files / time taken by
> process, the following chart could be drawn.
> !Screenshot 2019-04-08 18.41.10.png!
> That means, to make zip archive from 500mb of files takes over 12 minutes!
> Looks like in order to add a file, camel extracts zip archive to input
> stream, put file inside it, and build zip archive again. So that becomes near
> quadratic complexity, and not acceptable for large folders.
> The workaround is to add completionSize or completionPredicate to flush every
> 100mb, so we got all files archived but splitted into several archives, which
> works but not the best choice.
>
> Is there a general solution how to make ZipAggregationStrategy to work in
> near linear time, so the process does not become slower with large number of
> files?
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
