Are pull replications really still faster than push?

[Jens Alfke <je...@couchbase.com>]

The wiki documentation of the _replicate API says:

> As of CouchDB 0.9, pull replication is a lot more efficient and resistant to errors, and it is suggested that you use pull replication in most cases, especially if your documents are large or you have large attachments.
(http://wiki.apache.org/couchdb/Replication)

This was probably written circa 2009. Is it still true? I know from experience that the pull replication protocol has a big bottleneck in that it has to fetch revisions using individual GET requests, whereas push replication can send them in bulk via a POST to _bulk_docs. In practice I have seen this have a big effect on pull-replication performance.

—Jens

Re: Are pull replications really still faster than push?

[Adam Kocoloski <ko...@apache.org>]

On Jun 20, 2013, at 1:32 PM, Jens Alfke <je...@couchbase.com> wrote:

> The wiki documentation of the _replicate API says:
> 
>> As of CouchDB 0.9, pull replication is a lot more efficient and resistant to errors, and it is suggested that you use pull replication in most cases, especially if your documents are large or you have large attachments.
> (http://wiki.apache.org/couchdb/Replication)
> 
> This was probably written circa 2009. Is it still true? I know from experience that the pull replication protocol has a big bottleneck in that it has to fetch revisions using individual GET requests, whereas push replication can send them in bulk via a POST to _bulk_docs. In practice I have seen this have a big effect on pull-replication performance.
> 
> —Jens

It's no longer true.

Adam