Hi Liam,

Thanks a lot for your response and it really helps for my assessment.

And also I am mentioning some additional information about my requirement,
in case if you have additional information to provide kindly let me know.

No of messages(including all topics messages) per day = 20000
Maximum size of each message = 1KB
Total size of messages per day = 20000*1KB = 20MB
Message retention = 5 days
Number of topics = 4 (4 partitions are enough)

And thinking to use below configuration.

RAM to run Kafka cluster = 16GB (For all nodes including Kafka (2 nodes)
and Zookeeper (2 nodes))
SD space= 20GB

Thanks and regards,
Naveen


On Thu, Feb 20, 2020 at 9:37 AM Liam Clarke <liam.cla...@adscale.co.nz>
wrote:

> Hi Naveen,  a very broad question, but to use Kafka as a backbone of your
> infrastructure, the brokers need to be on machines with enough disk to
> store the expected data, and with good network interface capacity - we use
> 10Gbps NICs.
>
> Sizing disks is a case of knowing how long you want retain data for, how
> many replicas you want to have, and how much data you expect, and how
> you'll partition it.
>
> For example, if your topic will have 1TB of data produced onto it in a day,
> and you want to have 3 replicas, retaining data for 14 days,  you'll need 3
> * 14 * 1TB disk space across your cluster to fit it.
>
> But you also need to ensure that one partition can fit on disk on a broker.
> So if the above topic has 20 partitions, you'll need to make sure that each
> broker has a disk that can fit 50GB * 14. Plus you need some additional
> space for Kafka metadata and OS overhead etc.
>
> Hope that helps,
>
> Liam Clarke
>
> On Thu, 20 Feb. 2020, 7:02 pm Naveen Kumar M, <eai.naveenkuma...@gmail.com
> >
> wrote:
>
> > Hello Kafka Community,
> >
> > Kindly help me on this.
> >
> > Thanks and regards,
> > Naveen
> >
> > On Wed, Feb 19, 2020, 1:41 PM Naveen Kumar M <
> eai.naveenkuma...@gmail.com>
> > wrote:
> >
> > > Hello Team,
> > >
> > > Kindly help me to understand hardware assessment details to setup Kafka
> > as
> > > messaging broker to de-couple applications/systems.
> > >
> > > Thanks and regards,
> > > Naveen
> > >
> >
>

Reply via email to