reef-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Boris Shulman <shulm...@gmail.com>
Subject RE: [DISCUSS] Apache REEF is ...
Date Fri, 01 Apr 2016 19:10:22 GMT
+1 for the latest version.

-----Original Message-----
From: "Markus Weimer" <markus@weimo.de>
Sent: ‎4/‎1/‎2016 11:37 AM
To: "dev@reef.apache.org" <dev@reef.apache.org>
Subject: Re: [DISCUSS] Apache REEF is ...

On 3/31/2016 3:54 PM, Byung-Gon Chun wrote:
> "standard" implies there's consensus among many parties. "common" may
> be more accurate.

Good point. It is a fine line between "standard library in the sense of
stdlib.so" and "the *standard* off all the libraries for the purpose."

How about dropping all qualifiers:

> Apache REEF is a library for building distributed applications on top
> of cluster resource managers (e.g., Apache Hadoop YARN and Apache
> Mesos).

I like this, because it simplifies the statement. And if all goes well,
we can add back `the standard library` in a while, once we've earned it :-)

On 3/31/2016 3:54 PM, Byung-Gon Chun wrote:
> Instead of saying "standard library", perhaps we can say
> "meta-framework".

I'm not a fan of that term. It makes it less clear of what the immediate
utility of REEF is: If you need an AM, you are best served writing it in
REEF.

> Do we want to say "distributed applications"? This's broad.

We initially had "Big Data applications" there. I dropped it, because
for example Azure Stream Analytics scales all the way from tiny to Big
Data. Also, I don't like the hype connotation of "Big Data".

> Mostly we're focusing on "distributed data processing applications".
>  Perhaps we can narrow the scope.

Yes, that is the current focus. At the same time, REEF is currently
sparse in data primitives, and rich in control flow primitives. And
those apply more broadly than to data processing apps. However, I am not
set on this. If we find a concise and precise way to narrow it, I am all
for it. I'd even be OK with dropping the "distributed" as it is implied
by running on resource managers.

Which brings me to a new, extremely concise Option F:

> Apache REEF is a library for application development on cluster
> resource managers (e.g., Apache Hadoop YARN and Apache Mesos).

What do you think?

Markus

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message