If you are comfortable with the Cassandra data model, you understand how a 
QUORUM read and write work at various replication factors and you think that 
the growth of your dataset will somewhat fast as Rob mentioned. Go for it!

Just remember to always benchmark performance and test what happens when your 
node / nodes go down so you understand how your app will break when the 
inevitable happens. The good thing about VPS/cloud providers is its very cheap 
and quick to test your plans which will save you a world of hurt in the future. 

Disclaimer: I work for Instaclustr and we provide managed Cassandra hosting, so 
the below advice is a little biased ;)

Your time is your most valuable thing, even when you don't have a ton of cash. 
Don't spend weeks trying to get something working on a cheap service provider 
that will save you $100 a month. Instead suck it up, pay the extra $100 a month 
for the alternative and move onto things more important things like customer 
acquisition.

Good luck

Ben Bromhead
Instaclustr | www.instaclustr.com | @instaclustr 

On 28/08/2013, at 3:36 AM, Robert Coli <rc...@eventbrite.com> wrote:

> On Tue, Aug 27, 2013 at 3:04 AM, Aklin_81 <asdk...@gmail.com> wrote:
> But so far what I have seen, it's something very different with Cassandra. 
> People usually recommend starting out with atleast a 3 node cluster, (on 
> dedicated servers) with lots & lots of RAM. 4GB or 8GB RAM is what they 
> suggest to start with. So is it that Cassandra requires more hardware 
> resources in comparison to MySQL,  for a website to deliver similar 
> performance, serve similar load/ traffic & same amount of data. I understand 
> about higher storage requirements of Cassandra due to replication but what 
> about other hardware resources ?
> 
> If you have enough data to fit on a single computer, or two computers, and 
> only need one or two copies of it for redundancy/availability... you probably 
> don't have enough data to need something like Cassandra. Vertically scaling 
> RDBMS or in-memory k/v stores to significant size is a tried and true 
> technique. If you may eventually have more data than can fit on a single very 
> powerful computer, Cassandra or something like it may be indicated. It really 
> starts to shine once you have N>=3, because then you can have RF>=3, which 
> allows you to use QUORUM reads and writes for consistency. But because 
> Cassandra is designed for horizontal scaling (and because it runs in the 
> JVM...) it is natural that it will slightly underperform from the perspective 
> of vertical scaling.
> 
> =Rob

Reply via email to