ronniesahlberg at gmail.com
Wed Nov 19 18:44:08 MST 2014
On Wed, Nov 19, 2014 at 3:49 PM, Richard Sharpe <realrichardsharpe at gmail.com
> Hi folks,
> In Tridge's 2007 paper:
> he claims the following performance scaling:
> NEW (CTDB) approach
> 1 node 42 Mbytes/sec
> 2 nodes 168 MBytes/sec
> 3 nodes 211 MBytes/sec
> 4 nodes 243 MBytes/sec
> This seems counter intuitive. 2 nodes gets four times what one node
> gets and four nodes gets almost six times what 1 node does?
> What is the explanation for that?
The superlinear scaling is likely due to the increase of memory for caching.
This is recall is the uncontended case where you have little cross node
In SoNAS we did demonstrate near linear scaling in many tests up to ~30
nodes or so for the uncontended case.
The contended case is different. When you have a very hot record that is
contended across all nodes (say every single client opening and closing the
same file repeatedly) things do not scale as well.
> Richard Sharpe
More information about the samba-technical