Don't overkill it with technology, you could use Riak with a simple 2i
index (integer index YYYYMMDD for the message date so you can search day
by day backward), and for the message sequence or identifier you could
either user ANY SQL database sequence or a UUID generator.
HTH,
Guido.
On 22/10/12 10:04, Rapsey wrote:
On Mon, Oct 22, 2012 at 10:29 AM, Shashwat Srivastava
<dark...@gmail.com <mailto:dark...@gmail.com>> wrote:
Now, each bucket would have conversation between two users or of a
room of a site. The conversation rate for (some) rooms is very
high, some 20,000 - 30,000 messages per hour. We
have observed that users usually don't access conversations past
one week. So, if a bucket has conversation of 3 years, then mostly
users would access the recent conversation upto a week or month.
Can riak handle this easily? Also, would riak use RAM wisely in
this scenario? Would it only keep keys and indexes, corresponding
to recent messages per bucket, in RAM?
Leveldb backend should.
Finally, what is the best approach for creating keys in a bucket?
Earlier, I was planning to use timestamp (in milliseconds). But in
a room there can be multiple messages at the same time. As I
understand I cannot have a unique incremental message id per
bucket (as riak has write capability in all nodes in a cluster so
consistency is not guareented). Please correct me if I am wrong.
One other way could be to let riak generate key and I use
timestamp as a secondary index. But this seems to be a bad design.
Also, what would be the best way to achieve pagination for this
use case?
You could use redis for incremental id's.
Sergej
_______________________________________________
riak-users mailing list
riak-users@lists.basho.com
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
_______________________________________________
riak-users mailing list
riak-users@lists.basho.com
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
