Mark, Jonathan (Integic) wrote:
Then run a single zope zeo client per processor core. ZODB makes data
partitioning (database mounts) simple as well.
ZRS is only required for replication, not scaling, for which ZEO
My interpretation of this statement is as follows. You can store
information in separate partitions of the ZODB, and ZEO will route
requests to the appropriate partition.
Is that true only for Zope 3, or is it true for Zope 2.1x as well?
Using multiple storages (which is a better term than "partitions") is a
ZODB feature. It's therefore available in all recent Zope 2 and 3 versions.
Would Zope applications need to be rewritten to support multiple
partitions? Or is it all transparent to the developer?
It's completely transparent. You attach an object from one storage to an
object from the other and voila, you have "mounted" that object from one
storage to the other one (e.g. if that object is a folder/container, it
works much like a Unix mount point).
Has anyone actually ever tried that on a massive scale? What are the
largest and busiest functional partitioned ZODB instances in active use?
Some people like putting their catalog into a different storage, see 
for some discussion of that.
http://worldcookery.com -- Professional Zope documentation and training
Zope3-users mailing list