Mailing-List: contact dev-help@couchdb.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@couchdb.apache.org
Received-SPF: neutral (nike.apache.org: local policy)
MIME-Version: 1.0
In-Reply-To: 
 <CAN-3CBL2UVTtKsrT7RMXZ9rcvpA9G7zXNL=AXi3GOwgFSH8FLg@mail.gmail.com>
References: 
 <CAN-3CB+TreyyZgOst9GT8x-TFT3C+5fP2fMbiZ1mG7BS_sF3+Q@mail.gmail.com>
	<CAPaJBx4mV=4LC0kKBnXbhQwwdEA88gw9CMfJxpV2oitESo46yA@mail.gmail.com>
	<CAN-3CB+zSC+By8ttue6qtepYEdVr7cmMj7hW1SY3b5rVNmmKKw@mail.gmail.com>
	<F1F4C8FB-4050-4F79-8E49-8ED5C618DDF4@cloudno.de>
	<CAN-3CBL2UVTtKsrT7RMXZ9rcvpA9G7zXNL=AXi3GOwgFSH8FLg@mail.gmail.com>
Date: Tue, 19 Feb 2013 13:40:28 +0100
Message-ID: 
 <CAJ_D0VAagw1jgB+6KP2mEWJhbcs2mFBUOr1qsPtb6ebprtrJsg@mail.gmail.com>
Subject: Re: Node.js npm December usage
From: Hans J Schroeder <hs@cloudno.de>
To: "dev@couchdb.apache.org" <dev@couchdb.apache.org>
Content-Type: multipart/alternative; boundary=e89a8ff1c5c682bedf04d613253e

--e89a8ff1c5c682bedf04d613253e
Content-Type: text/plain; charset=ISO-8859-1

Thanks for the explanation. Again I am impressed with what can be done with
CouchDB. Iriscouch's service is awesome, performance-wise and
availability-wise.

- Hans

On Tuesday, February 19, 2013, Jason Smith wrote:

> Hi, Hans! I just emailed asking if you were on this list. I guess I should
> pay better attention myself!
>
> Yes we use, if you will, wide-area load balancing and also local-area load
> balancing.
>
> Locally, we have CouchDB replicas and reverse-proxies.
>
> Globally, we run a content distribution network, primarily using a custom
> Node.js DNS server: https://github.com/iriscouch/dnsd. We use geolocation
> to route to the best data center.
>
> Both systems also provide high-availability features. If a couch is down,
> we do not (well, ideally!) route to it. If the reverse-proxies are down, we
> do not include its IP address in the DNS response.
>
>
>
> On Tue, Feb 19, 2013 at 2:39 AM, Hans J Schroeder <hs@cloudno.de> wrote:
>
> > Hi Jason,
> >
> > These stats are totally impressive. Especially because it is real world
> > data and no result of a synthetic benchmark.
> >
> > I am interested how the three data centres are used with standard
> couchdb.
> > A combination of load balancing and master-master replication?
> >
> > - Hans
> >
> > On Feb 18, 2013, at 5:01 AM, Jason Smith <jhs@iriscouch.com> wrote:
> >
> > > Sure!
> > >
> > > Not mentioned in that email (and pardon me for banging on about it) is
> > that
> > > usage grows 15% monthly, i.e. doubling every 5 months. February is a
> > short
> > > month but we will probably hit 130M queries, a 1/3 growth since I wrote
> > > that email. Pretty exciting!
> > >
> > > We are working on publishing reports and stats about individual
> packages
> > > and things, so this is a good time to work on this.
> > >
> > > Next steps? Maybe I'll start scribbling down ideas on the wiki?
> > >
> > > On Mon, Feb 18, 2013 at 5:25 AM, Noah Slater <nslater@apache.org>
> wrote:
> > >
> > >> Man, this is totally great.
> > >>
> > >> Perhaps we could write it up as a case-study and promote it on our
> > >> homepage?
> > >>
> > >> Does that sound like a good idea? Something you could help with?
> > >>
> > >>
> > >> On 1 January 2013 05:32, Jason Smith <jhs@iriscouch.com> wrote:
> > >>
> > >>> Hi, all. Sorry to be distant from the community recently. No excuse.
> > >>>
> > >>> I thought I might share December stats from one of Apache CouchDB's
> > most
> > >>> well-known deployments and killer apps: the Node.js npm registry.
> > >>>
> > >>> ## Facts
> > >>>
> > >>> * Zero downtime
> > >>> * Three data centers: SoftLayer, EC2, Joyent
> > >>> * 99,327,470 HTTP queries served = 37/sec
> > >>>
> > >>> * Slowest minute: Dec 08 09:35, 578 queries = 9.6/sec
> > >>> * Busiest minute: Dec 20 18:43, 19,776 queries = 329/sec
> > >>>
> > >>> * Slowest second: (many), 0 queries
> > >>> * Busiest second: Dec 20 18:43:03, 932 queries/sec
> > >>>
> > >>> ## Reflections
> > >>>
> > >>> This is only the public registry. Our customers and also independent
> > >> third
> > >>> parties run their own replicas. We do not or cannot (respectively)
> > >> publish
> > >>> their usage stats.
> > >>>
> > >>> Think about that. Isaac owns the registry. We run the registry. Yet
> > >> neither
> > >>> of us can even **know** its entire function, much less do anything
> > about
> > >>> it. That is empowerment. That is why I joined CouchDB. CouchDB is
> free
> > >>> software for free data. It carries the ideals of the Free Software
> > >> movement
> > >>> into the 21st century.
> > >>>
> > >>> Plenty of sites can produce more impressive numbers than these. There
> > are
> > >>> even larger CouchDB sites out there. But I am still proud. This is
> not
> > a
> > >>> multi-million dollar venture-capitalized eyeball something something.
> > We
> > >>> run standard, orthodox Apache CouchDB. That is encouraging. I did not
> > >>> deliver these numbers. Apache CouchDB did. These are not benchmarks.
> > >> These
> > >>> are production logs. That is nine-hundred thirty-two satisfied
> > customers
> > >> in
> > >>> one second! (Well, a true sysadmin would say "not-yet disappointed
> > >>> customers" which is all one can ask for.) It shows that anybody can
> > wield
> > >>> CouchDB to similar effect.
> > >>>
> > >>> There are general-purpose programming languages, and there are
> > >>> domain-specific programming languages. Nobody gets upset because you
> > >> can't
> > >>> write a web server --
> Iris Couch
>

--e89a8ff1c5c682bedf04d613253e--