Andrew Purtell wrote:
> > Ya, I wondered about that too, esp given he mentions
> > 
> > "(yes I'm aware that you can run Zookeeper in single server
> > mode from a single JAR file)"
> > 
> > and Noah's stack requires (granted we require java):
> > 
> > Ruby
> > EventMachine/Sinatra/Ohm
> > Redis
> > 
> > Patrick
> 
> That was the dumbest thing I've read in a few weeks.
> 
> Just managing Redis persistence is as or more complex than using ZooKeeper.
> 

Don't be too quick to dismiss this - I think there are useful bits to be
gleaned.  The two big ones that come to mind for me are:

1. There's a lot of fear and hesitation around building (high
performance, critical) infrastructure around Java.

2. Interfaces are very important.

#1 is mostly PR, but also reflects that operating stuff in a JVM adds a
layer of abstraction that is pretty opaque.  Maybe more emphasis on
zookeeper's real-world performance numbers, case studies, and/or
debugging tools is in order.  Personally, I also find the signal to
noise ratio of the zookeeper logging to be very high.  'ruok' generates
INFO and WARNING log entries on my (old, 3.1.1) install, for example.

#2 is something that I think is worth tackling with reference client/API
implementations above and beyond java.  Really, the only truly
supported, asyncronous APIs for zookeeper are Java and C (I'm not
counting the stuff wrapped around the C client), and that limits the
client audience significantly - I think one of the things that Noah gets
right is having a REST interface.

</$.02>

-- 
[email protected] (>^-^)>

Reply via email to