Cassandra Usage on smaller projects

[Rui Silva <ru...@telecom.pt>]

Hi all,

first of all, I have read the Cassandra Hardware requirements page on
Cassandra wiki: http://wiki.apache.org/cassandra/CassandraHardware .

I am currently in a simple project that, fetches data from a message
broker. That data can be thought as logging data, about a system user
usage. I need to have a background process to read the data from the
message broker and persist it somewhere. The data I want to persist,
could be stored in a single SQL table, however there is no data to
relate, therefor, I thought I would not need a Relational database
approach. I have been trying Cassandra, and have write the background
process to store all my data, in 3 column families, ordered by timeuuid,
that could be handy for some analysis:).

I have read some articles/case studies and all of them talk about the
need to support thousands of writes per second. Well in my case I would
only need about 10 writes per second in the worst case scenario. Do I
really need to have 16 GB RAM servers running Cassandra?

I have two servers, with 4 cpu (2 cores each cpu). And about 4 GB of RAM
each. I was planing on deploying a cassandra node on each, and an apache
web server, to host a simple web application, that will query my
cassandra cluster and use the data stored to present graphs and tables
(statistical information).

Do you think I may get into trouble with this design choice in the future?

Best Regards,

Rui Silva

Re: Cassandra Usage on smaller projects

[Michael Dürgner <mi...@duergner.de>]

10 writes / second could even be done with every sql/nosql solution, even with plain files.

So I think the storage choosen should be the one optimized for the queries you wanna have.

Sent from my iPhone

On 06.08.2010, at 05:54, Sal Fuentes <fu...@gmail.com> wrote:

> Would you care to elaborate?
> 
> On Thu, Aug 5, 2010 at 8:27 AM, Mark <st...@gmail.com> wrote:  
> MongoDB may be a better choice for this?
> 
> 
> 
> -- 
> Salvador Fuentes Jr.

Re: Cassandra Usage on smaller projects

[Sal Fuentes <fu...@gmail.com>]

Would you care to elaborate?

On Thu, Aug 5, 2010 at 8:27 AM, Mark <st...@gmail.com> wrote:
>
> MongoDB may be a better choice for this?
>



-- 
Salvador Fuentes Jr.

Re: Cassandra Usage on smaller projects

[Mark <st...@gmail.com>]

On 8/5/10 8:01 AM, Rui Silva wrote:
> Hi all,
>
> first of all, I have read the Cassandra Hardware requirements page on
> Cassandra wiki: http://wiki.apache.org/cassandra/CassandraHardware .
>
> I am currently in a simple project that, fetches data from a message
> broker. That data can be thought as logging data, about a system user
> usage. I need to have a background process to read the data from the
> message broker and persist it somewhere. The data I want to persist,
> could be stored in a single SQL table, however there is no data to
> relate, therefor, I thought I would not need a Relational database
> approach. I have been trying Cassandra, and have write the background
> process to store all my data, in 3 column families, ordered by timeuuid,
> that could be handy for some analysis:).
>
> I have read some articles/case studies and all of them talk about the
> need to support thousands of writes per second. Well in my case I would
> only need about 10 writes per second in the worst case scenario. Do I
> really need to have 16 GB RAM servers running Cassandra?
>
> I have two servers, with 4 cpu (2 cores each cpu). And about 4 GB of RAM
> each. I was planing on deploying a cassandra node on each, and an apache
> web server, to host a simple web application, that will query my
> cassandra cluster and use the data stored to present graphs and tables
> (statistical information).
>
> Do you think I may get into trouble with this design choice in the future?
>
> Best Regards,
>
> Rui Silva
>    
MongoDB may be a better choice for this?

Re: Cassandra Usage on smaller projects

[Jonathan Ellis <jb...@gmail.com>]

It sounds like you would be fine doing what you propose.

On Thu, Aug 5, 2010 at 11:01 AM, Rui Silva <ru...@telecom.pt> wrote:
> Hi all,
>
> first of all, I have read the Cassandra Hardware requirements page on
> Cassandra wiki: http://wiki.apache.org/cassandra/CassandraHardware .
>
> I am currently in a simple project that, fetches data from a message
> broker. That data can be thought as logging data, about a system user
> usage. I need to have a background process to read the data from the
> message broker and persist it somewhere. The data I want to persist,
> could be stored in a single SQL table, however there is no data to
> relate, therefor, I thought I would not need a Relational database
> approach. I have been trying Cassandra, and have write the background
> process to store all my data, in 3 column families, ordered by timeuuid,
> that could be handy for some analysis:).
>
> I have read some articles/case studies and all of them talk about the
> need to support thousands of writes per second. Well in my case I would
> only need about 10 writes per second in the worst case scenario. Do I
> really need to have 16 GB RAM servers running Cassandra?
>
> I have two servers, with 4 cpu (2 cores each cpu). And about 4 GB of RAM
> each. I was planing on deploying a cassandra node on each, and an apache
> web server, to host a simple web application, that will query my
> cassandra cluster and use the data stored to present graphs and tables
> (statistical information).
>
> Do you think I may get into trouble with this design choice in the future?
>
> Best Regards,
>
> Rui Silva
>



-- 
Jonathan Ellis
Project Chair, Apache Cassandra
co-founder of Riptano, the source for professional Cassandra support
http://riptano.com