Hi again,
I'm wondering what speed impact you'd typically expect to see moving from a
single controller to a distibuted controller.
I'd expect it to be slower, as it's obviously doing a lot more work. However
I've been running some tests using real-world transactions (lots of different
read-writes).
Using a single controller (but set up for distribution), this work takes around
2 mins (slower than without seqoia, but quite fast enough).
However, when I add in the second controller, the same work takes over 20
minutes. The second controller is basically just accepting requests - all
database activity is initiated through the same controller as the first test.
Both boxes are just ticking over on CPU, have plenty of memory & are connected
via switched gigabit ethernet. I'm running Sequoia 2.10.9 on Solaris 10 (x86),
and they're both talking to identical apache derby databases (2 databases per
node, one for backup).
Is this kind of performance difference usual? Or does it indicate some kind of
misconfiguration on my side?
I was under the impression (perhaps wrongly) that writes to the networked
controllers were asynchronous, so wouldn't slow down individual transactions...
(is this right?)
My virtual database config contains:
<Distribution hederaPropertiesFile="/hedera_appia.properties">
<MessageTimeouts/>
</Distribution>
and by appia config is using the following:
<channel name="TCP Token Channel" template="tcp_token" initialized="yes">
<memorymanagement size="40000000" up_threshold="15000000"
down_threshold="7000000" />
<chsession name="hederalayer">
<parameter name="gossip_address">224.1.1.5</parameter>
</chsession>
</channel>
any advice?
many thanks,
Andrew Lawrenson
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
