Network performance using couchbase 2.0
Fri, 01/06/2012 - 07:45
Hi again, I have some questions about network performance using couchbase 2.0:
PROBLEM:
Having a huge database that stores, for example, emails with the following JSON:
{
"type" : "mail",
"from" : "someone",
"to" : "user01",
"message" : "some message text"
}If an specific user want to query for the list of emails sent to him, and those emails are randomly stored over the whole cluster, this search will affect the network performance?
QUESTIONS:
1) Is it solved by using views? or the indexes of the views are divided per node (so all nodes are queried in order to get all the information needed, each time I perform a search)?
2) Can I group all the mails sent to an specific user in a single vBucket? Is this a recommended approach?
3) Do you think this is a "real world" problem? Do you have statistics that can clarify this?
thanks!
Hi,
this is a question about Couchbase Server 2.0 (and not Single Server). Hopefully an Admin can move it to the right forum.
1) Yes, this is solved by using views. The indexes will be divided per node and all nodes will be queried to get the information.
2) You could createa Bucket per user. I would say it would be faster as there will be less data to index/search on every query. Though keep in mind that you can't do cross queries across Buckets.
3) Yes, I think this is a real world problem. We are still working hard on Couchbase Server 2.0, so there aren't much statistics yet.
Cheers,
Volker