You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@couchdb.apache.org by Fredrik Widlund <fr...@qbrick.com> on 2010/04/19 09:52:44 UTC

RE: Ezra doesn't recommend CouchDB_


Anybody using a replicated setup, with say at least 5-10 nodes or so? If so can you state an average load in terms of ops/sec and cpu?

Kind regards,
Fredrik

-----Original Message-----
From: James Hayton [mailto:theboss@purplebulldog.com]
Sent: den 18 april 2010 23:26
To: user@couchdb.apache.org
Subject: Re: Ezra doesn't recommend CouchDB

I know this guy is a founder of engine yard, but I really question if he
knows what he is talking about here.  He said Mongo was durable and unless I
am missing something, Mongo openly admits that they trade durability for
speed.  CouchDB is probably the most durable solution out there I think (I
regularly just kill an amazon ec2 instance in the middle of something and
never had a problem...), which is something he says he values when he
recommends Tokyo, Mongo and Redis.  I am really curious why he says he
wouldn't even consider using CouchDB.   I am using CouchDB with a rails 3
app right now and I have not noticed any problems.

I would love to hear what everyone else thinks.  Is there some big problem
hanging out there that I don't know about?

James

On Sun, Apr 18, 2010 at 12:53 PM, Niket Patel <ne...@me.com> wrote:

> That is good to know. I assume you mentioned couchapp use case.
>
> We are not using couchapp but I think since 0.10 we are seeing good enough
> performance only know problem is dynamic queries ( Same problem that
> Raindrop have ) for our use case couchdb is good fit.
>
> I believe need to understand Map/Reduce upfront is biggest problem for any
> newbie. Performance is never a problem.
>
> There are other small things like very high disk use.
>
>
> On 18-Apr-2010, at 11:53 PM, Anh <7z...@gmail.com> wrote:
>
>  That seems unlikely IMHO, we've been perf testing Couch for our usage and
>> it outperforms a decently tuned Java webapp running on Resin (known for its
>> good performance) by about 2x.   It handles the top-end really well, serving
>> 50 or 60 times our current load, with slower response, but not knocking
>> over, unlike the Java webapp which starts dropping connections
>>
>