Import big data to Riak

Rune Skou Larsen rsl at trifork.com
Tue Oct 29 11:50:25 EDT 2013


Den 29-10-2013 16:32, Georgi Ivanov skrev:
> Hi and thank you for the reply. My comment follow:
>
>> Your tests are not close to what you are going to have in production
> My tests are exactly what we will have in production. 1 node or in best case 2
> nodes.
> We don't care about durability here.
If you don't care about durability, why do you want multiple copies of 
each object? I.e. why not just keep n=1?

>   Our request per second will be extremely
> low too(300 per day ?)
> Also we use 2i indexes. 
If you want to be able to find all your indexed objects with 2i, then 
don't increase the n-value with data in a bucket. This will change the 
way 2i looks for your data by skipping partitions, so you won't see all 
the objects until AAE has completed establishing redundancy.

Consider checking out Yokozuna for searching instead of 2i.

Rune,
Trifork




More information about the riak-users mailing list