You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@cassandra.apache.org by Siddharth Karandikar <si...@gmail.com> on 2014/10/10 10:56:59 UTC

Internode latency requirements

Hello Everyone!

I am going over multiple cassandra documents to understand network
requirements of deploying cassandra cluster. Doacuments mention that
it is recommended to have 1gbps (or more) bandwidth between nodes and
its better to have multiple interfaces. I did not see any specific
latency numbers mentioned there.

Lets say I want to deploy ~200 cassandra nodes and each of them
running at different geographic location. Distance between any 2 nodes
will range from 5ms to 40ms in terms of RTT. I plan to configure nodes
close to each other in one datacenter. And such multiple datacenters
will be connected to each other.

I know this is a very specific requirement. So wanted to know if
cassandra is a right fit there.


Any inputs/suggestions/questions are welcome!


Thanks,
Siddharth

Re: Internode latency requirements

Posted by Siddharth Karandikar <si...@gmail.com>.

Thanks Alex for inputs. As you said, choosing right consistency levels
will be critical here.

On Mon, Oct 13, 2014 at 8:37 PM, Alex Major <al...@gmail.com> wrote:
> Just make sure you understand the effect of consistency levels on your
> performance. You (probably) don't want to be going over the WAN for reads
> etc.
>
> We run across the US/EU AWS regions and don't have any problems with higher
> RTT.
>
> On Mon, Oct 13, 2014 at 2:51 AM, Siddharth Karandikar
> <si...@gmail.com> wrote:
>>
>> Thanks for confirmation.
>>
>> Anyone has any other important points that need to keep in mind for such
>> distributed-distributed deployments?
>>
>> On 10 Oct 2014 23:13, "Robert Coli" <rc...@eventbrite.com> wrote:
>>>
>>> On Fri, Oct 10, 2014 at 1:56 AM, Siddharth Karandikar
>>> <si...@gmail.com> wrote:
>>>>
>>>> Lets say I want to deploy ~200 cassandra nodes and each of them
>>>> running at different geographic location. Distance between any 2 nodes
>>>> will range from 5ms to 40ms in terms of RTT. I plan to configure nodes
>>>> close to each other in one datacenter. And such multiple datacenters
>>>> will be connected to each other.
>>>
>>>
>>> That's fine?
>>>
>>> People run multi-DC over much slower WAN links than that.
>>>
>>> =Rob
>>> http://twitter.com/rcolidba
>
>

Re: Internode latency requirements

Posted by Alex Major <al...@gmail.com>.

Just make sure you understand the effect of consistency levels on your
performance. You (probably) don't want to be going over the WAN for reads
etc.

We run across the US/EU AWS regions and don't have any problems with higher
RTT.

On Mon, Oct 13, 2014 at 2:51 AM, Siddharth Karandikar <
siddharth.karandikar@gmail.com> wrote:

> Thanks for confirmation.
>
> Anyone has any other important points that need to keep in mind for such
> distributed-distributed deployments?
> On 10 Oct 2014 23:13, "Robert Coli" <rc...@eventbrite.com> wrote:
>
>> On Fri, Oct 10, 2014 at 1:56 AM, Siddharth Karandikar <
>> siddharth.karandikar@gmail.com> wrote:
>>
>>> Lets say I want to deploy ~200 cassandra nodes and each of them
>>> running at different geographic location. Distance between any 2 nodes
>>> will range from 5ms to 40ms in terms of RTT. I plan to configure nodes
>>> close to each other in one datacenter. And such multiple datacenters
>>> will be connected to each other.
>>>
>>
>> That's fine?
>>
>> People run multi-DC over much slower WAN links than that.
>>
>> =Rob
>> http://twitter.com/rcolidba
>>
>

Re: Internode latency requirements

Posted by Siddharth Karandikar <si...@gmail.com>.

Thanks for confirmation.

Anyone has any other important points that need to keep in mind for such
distributed-distributed deployments?
On 10 Oct 2014 23:13, "Robert Coli" <rc...@eventbrite.com> wrote:

> On Fri, Oct 10, 2014 at 1:56 AM, Siddharth Karandikar <
> siddharth.karandikar@gmail.com> wrote:
>
>> Lets say I want to deploy ~200 cassandra nodes and each of them
>> running at different geographic location. Distance between any 2 nodes
>> will range from 5ms to 40ms in terms of RTT. I plan to configure nodes
>> close to each other in one datacenter. And such multiple datacenters
>> will be connected to each other.
>>
>
> That's fine?
>
> People run multi-DC over much slower WAN links than that.
>
> =Rob
> http://twitter.com/rcolidba
>

Re: Internode latency requirements

Posted by Robert Coli <rc...@eventbrite.com>.

On Fri, Oct 10, 2014 at 1:56 AM, Siddharth Karandikar <
siddharth.karandikar@gmail.com> wrote:

> Lets say I want to deploy ~200 cassandra nodes and each of them
> running at different geographic location. Distance between any 2 nodes
> will range from 5ms to 40ms in terms of RTT. I plan to configure nodes
> close to each other in one datacenter. And such multiple datacenters
> will be connected to each other.
>

That's fine?

People run multi-DC over much slower WAN links than that.

=Rob
http://twitter.com/rcolidba