Mailing-List: contact user-help@cassandra.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@cassandra.apache.org
Received-SPF: pass (nike.apache.org: domain of ajazam@gmail.com designates
 209.85.215.45 as permitted sender)
MIME-Version: 1.0
In-Reply-To: 
 <CAEDUwd1j2uevQ6J9YrYvNuqmonYCEWbSw9RPTMn+_Nqw6dRfhw@mail.gmail.com>
References: 
 <CAPqEvGHif6CX_ksNPsE7O5bsg81=dp-e004DTrLtNmoqNwXBHg@mail.gmail.com>
	<CAEDUwd1j2uevQ6J9YrYvNuqmonYCEWbSw9RPTMn+_Nqw6dRfhw@mail.gmail.com>
Date: Thu, 22 May 2014 21:32:23 +0100
Message-ID: 
 <CAPqEvGE9wMsifgs4YMXu1YkFLnMzieOCQYBJ4z1GXXaBv9f8DA@mail.gmail.com>
Subject: Re: autoscaling cassandra cluster
From: Jabbar Azam <ajazam@gmail.com>
To: user@cassandra.apache.org
Content-Type: multipart/alternative; boundary=047d7bb70aa4b0ce3b04fa0302a4

--047d7bb70aa4b0ce3b04fa0302a4
Content-Type: text/plain; charset=UTF-8

Netflix uses Scryer
http://techblog.netflix.com/2013/11/scryer-netflixs-predictive-auto-scaling.htmlfor
predictive and reactive autoscaling but they only refer to EC2
instances. They don't mention anything about cassandra scaling or adding
and removing nodes.

I've just looked at the priam wiki and it also doesn't mention scaling. It
also mentions that vnodes aren't fully supported. That's no use for me as
I'm using 2.x. The other issue, rather feature of cassandra, is that adding
a new node increases the load on the system so this surge would need to be
taken into account.

I think I'll leave this problem for more intelligent people than me and
concentrate on the application logic, which can scale by adding or removing
application and front end servers.

Thanks for all your comments.

Thanks

Jabbar Azam


On 22 May 2014 19:55, Robert Coli <rcoli@eventbrite.com> wrote:

> On Wed, May 21, 2014 at 4:35 AM, Jabbar Azam <ajazam@gmail.com> wrote:
>
>> Has anybody got a cassandra cluster which autoscales depending on load or
>> times of the day?
>>
>
> Netflix probably does, managed with Priam.
>
> In general I personally do not consider Cassandra's mechanisms for joining
> and parting nodes to currently work well enough to consider designing a
> production system which would do so as part of regular operation.
>
> =Rob
>
>

--047d7bb70aa4b0ce3b04fa0302a4
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Netflix uses Scryer=C2=A0<a href=3D"http://techblog.netfli=
x.com/2013/11/scryer-netflixs-predictive-auto-scaling.html">http://techblog=
.netflix.com/2013/11/scryer-netflixs-predictive-auto-scaling.html</a> for p=
redictive and reactive autoscaling but they only refer to EC2 instances. Th=
ey don&#39;t mention anything about cassandra scaling or adding and removin=
g nodes.<div>
<br></div><div>I&#39;ve just looked at the priam wiki and it also doesn&#39=
;t mention scaling. It also mentions that vnodes aren&#39;t fully supported=
. That&#39;s no use for me as I&#39;m using 2.x. The other issue, rather fe=
ature of cassandra, is that adding a new node increases the load on the sys=
tem so this surge would need to be taken into account.</div>
<div><br></div><div>I think I&#39;ll leave this problem for more intelligen=
t people than me and concentrate on the application logic, which can scale =
by adding or removing application and front end servers.</div><div><br>
</div><div>Thanks for all your comments.</div></div><div class=3D"gmail_ext=
ra"><br clear=3D"all"><div><div dir=3D"ltr">Thanks<br><br>Jabbar Azam<br></=
div></div>
<br><br><div class=3D"gmail_quote">On 22 May 2014 19:55, Robert Coli <span =
dir=3D"ltr">&lt;<a href=3D"mailto:rcoli@eventbrite.com" target=3D"_blank">r=
coli@eventbrite.com</a>&gt;</span> wrote:<br><blockquote class=3D"gmail_quo=
te" style=3D"margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"=
>
<div dir=3D"ltr"><div class=3D"gmail_extra"><div class=3D"gmail_quote"><div=
 class=3D"">On Wed, May 21, 2014 at 4:35 AM, Jabbar Azam <span dir=3D"ltr">=
&lt;<a href=3D"mailto:ajazam@gmail.com" target=3D"_blank">ajazam@gmail.com<=
/a>&gt;</span> wrote:<br>

<blockquote class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border-left:1p=
x #ccc solid;padding-left:1ex"><div dir=3D"ltr"><div>Has anybody got a cass=
andra cluster which autoscales depending on load or times of the day?</div>=
</div>

</blockquote><div><br></div></div><div>Netflix probably does, managed with =
Priam.</div><div><br></div><div>In general I personally do not consider Cas=
sandra&#39;s mechanisms for joining and parting nodes to currently work wel=
l enough to consider designing a production system which would do so as par=
t of regular operation.</div>

<div><br></div><div>=3DRob</div><div>=C2=A0</div></div></div></div>
</blockquote></div><br></div>

--047d7bb70aa4b0ce3b04fa0302a4--