Heavy Writes to Couchbase
Wed, 11/28/2012 - 23:58
Again something related to infrastructure. Iam creating a proto of a data store that is expected to have write from a Data Access Layer (Conversions and Enrichments to data). The store will also have loads of read from clients that access these documents that are getting saved. So basically we were planning to have the writes done from the data access layer split between the couchbase nodes that we have using a distributed locking system like Apache Zoo Keeper or cages.
Do we have to do this or will multiple threads writing to the same node itself cause any issues ?
The very short answer is:
- You do not have to worry about it, Couchbase will deal with it for you.
Let me explain, when you are building your application, this application will use a client SDK that is responsible of sharding the data across the cluster ( see http://www.couchbase.com/docs/couchbase-devguide-2.0/brief-info-on-vBuck... ). So the cluster itself will deal with the writes.
If your cluster is well sized, you need to be sure that your application can deal with the load itself.
Take a look to this benchmark, you will see that writing into Couchbase cluster is extremely fast:
http://www.slideshare.net/renatko/couchbase-performance-benchmarking
Once again, your application should not "know" about the nodes, your application only deal with the cluster itself.
HTH
Tug
Tug
@tgrall