Mailing-List: contact dev-help@accumulo.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@accumulo.apache.org
Received-SPF: pass (athena.apache.org: domain of busbey@cloudera.com
 designates 209.85.216.46 as permitted sender)
MIME-Version: 1.0
In-Reply-To: <1964235200.13497615.1403320887710.JavaMail.root@comcast.net>
References: <1341675922.13455905.1403312644690.JavaMail.root@comcast.net>
 <1760482492.13459235.1403313150097.JavaMail.root@comcast.net>
 <1472683858.13465692.1403314205595.JavaMail.root@comcast.net>
 <1180482440.13467947.1403314515456.JavaMail.root@comcast.net>
 <CAF1jEfAfxgHKs3SO5DbdbLTCHv7Wir7QQKoDMAXFNuuNshsFEw@mail.gmail.com>
 <CAGHyZ6JDgtM30YFuyQRjfFjqoLrf9=pKfCfy1m6O3B6rAnWVqQ@mail.gmail.com>
 <1964235200.13497615.1403320887710.JavaMail.root@comcast.net>
From: Sean Busbey <busbey@cloudera.com>
Date: Mon, 23 Jun 2014 21:55:11 -0500
Message-ID: 
 <CAGHyZ6+Oc=FMWsbBYEvJE5h4s8khaTc_pLA=dabfBrNt6-QQ+w@mail.gmail.com>
Subject: Re: [VOTE][BLOG] New Blog Entry - "Scaling Accumulo with Multi-Volume
 Support"
To: "dev@accumulo apache. org" <dev@accumulo.apache.org>
Content-Type: multipart/alternative; boundary=001a1133453cc428e804fc8c17c6

--001a1133453cc428e804fc8c17c6
Content-Type: text/plain; charset=UTF-8

paragraph 3: link [3] should be to the 1.6 version of the bulk ingest
information.

paragraph 5-7: It would be nice if  you could show some kind of chart to
back up the NameNode having scale issues.

you could cite the usenix article from 2010 on NameNode write rates[1], but
that was only like 6k/sec and was really only close to the needed "create
new file" and not rename, move, or delete. You could try to use
NNThroughputBenchmark to atleast update that stat on modern hardware. I
don't know if it also includes anything for the other ops.

paragraph 9: I'd recommend rephrasing the bit about HDSF federation being
insufficient for Accumulo. Accumulo could have found other approaches to
leverage Federation as is, but would have been awkward to use on anything
other than a per-table level. We can showcase MVS as a way to leverage
federation easily within Accumulo rather than as an alternative approach.

Honestly, the only way I expect MVS to be feasibly deployed is with
federation used to name several accumulo data directories in a single
namespace. Federation handles sharing datanodes across all the namenodes
you configure. Unless you want to give up things like short circuit reads,
we'll either need more infra to associate tservers with particular volumes
or we'll need to run multiple datanode processes per host. It'll be messy.

paragraph 17: Is there any chance that this example configuration could be
presented as a set of docker or vagrant nodes that are configured to run an
example cluster with multiple hdfs instances and Accumulo with MVS enabled?
Something people can poke at will be way more popular than XML configs as
they are.

If you don't have more time to polish, I understand. In that case the blog
post will read better if the scripts are in a github repo and linked from
the article (possibly with an embedded preview if github does that sort of
thing.)

My apologies if it's formatted differently on the actual preview. My
request for a roller account hasn't gotten anywhere yet.


On Fri, Jun 20, 2014 at 10:21 PM, <dlmarion@comcast.net> wrote:

> Good idea Sean. Here you go: https://paste.apache.org/p/f9C8
>
> ----- Original Message -----
>
> From: "Sean Busbey" <busbey@cloudera.com>
> To: "dev@accumulo apache. org" <dev@accumulo.apache.org>
> Sent: Friday, June 20, 2014 11:08:31 PM
> Subject: Re: [VOTE][BLOG] New Blog Entry - "Scaling Accumulo with
> Multi-Volume Support"
>
> You could post it to the ASF pastebin and then put a link here.
>
> https://paste.apache.org/
>
>
> On Fri, Jun 20, 2014 at 9:53 PM, Billie Rinaldi <billie.rinaldi@gmail.com>
> wrote:
>
> > Yes, I think they strip attachments.
> > On Jun 20, 2014 9:35 PM, <dlmarion@comcast.net> wrote:
> >
> > > I tried from two different accounts, must be the Apache mail servers?
> > >
> > >
> > > ----- Original Message -----
> > >
> > > From: dlmarion@comcast.net
> > > To: dev@accumulo.apache.org
> > > Sent: Friday, June 20, 2014 9:30:05 PM
> > > Subject: Re: [VOTE][BLOG] New Blog Entry - "Scaling Accumulo with
> > > Multi-Volume Support"
> > >
> > > Sean said the attachment didn't make it, let's try this again.
> > >
> > > ----- Original Message -----
> > >
> > > From: dlmarion@comcast.net
> > > To: dev@accumulo.apache.org
> > > Sent: Friday, June 20, 2014 9:12:30 PM
> > > Subject: Re: [VOTE][BLOG] New Blog Entry - "Scaling Accumulo with
> > > Multi-Volume Support"
> > >
> > >
> > > For those without a blog account, see attached.
> > >
> > > ----- Original Message -----
> > >
> > > From: dlmarion@comcast.net
> > > To: dev@accumulo.apache.org
> > > Sent: Friday, June 20, 2014 9:04:04 PM
> > > Subject: [VOTE][BLOG] New Blog Entry - "Scaling Accumulo with
> > Multi-Volume
> > > Support"
> > >
> > > All,
> > >
> > > Eric and I authored a new blog post on how the benefits of multi-volume
> > > support in 1.6.0. I will be accepting feedback on typo's and grammar
> > > errors. The entry is located at:
> > >
> > >
> > >
> >
> https://blogs.apache.org/roller-ui/authoring/preview/accumulo/?previewEntry=scaling_accumulo_with_multi_volume
> > >
> > > This vote will be open for feedback for 3 days (24 June 2014 0100 GMT)
> > and
> > > then I will promote to the blog site. Thanks,
> > >
> > > Dave
> > >
> > >
> > >
> > >
> > >
> >
>
>
>
> --
> Sean
>
>


-- 
Sean

--001a1133453cc428e804fc8c17c6--