search

junneyang / zumastor

Automatically exported from code.google.com/p/zumastor
0 stars 1 forks source link

Replication not wire speed :-) #143

Open GoogleCodeExporter opened 9 years ago

GoogleCodeExporter commented 9 years ago 
We should measure our replication performance on 
fast LANs, and fix as many bottlenecks as needed
until our replication performance is respectable.

We expect one of the bottlenecks to be the fact that
we only support reading one block from
the snapshot store at a time when sending out
snapshots to remote systems.

Original issue reported on code.google.com by daniel.r...@gmail.com on 17 May 2008 at 2:12

GoogleCodeExporter commented 9 years ago 
For the record:
This issue was derived from a conversation with Daniel
around May 15th; my notes say 
"Allow flushing more than one deferred write at once.
Allow reading more than one block at a time when doing replication."
I've asked him to describe it a bit more, including what he
thinks the performance impact of not reading multiple blocks at a time is.

Original comment by daniel.r...@gmail.com on 23 Jun 2008 at 10:28