Validating Performance

[InsiderEdsHead]

Hi,

I installed the Turpentine extension and all seems to be going well. I've been trying to verify / document and performance improvements, but haven't had any success.

I'm using pageSpeed for Chrome (and firefox) as well as varnishstat to see if i'm getting hits. I've tried across 3 different browsers on different computers, but according to PageSpeed my performance is that same or a click lower. Varnishstat is showing a very low hit ratio which makes me wonder if the requests are being cached at all.

I'd appreciate some insight into how to verify this.

Thanks,

Ed

[aheadley]

Most performance testing tools do not support cookies, which are required to see an improvement with Turpentine (by default). Try setting the Caching Options > Cache Cookie Control > Set Initial Cookie option to off/disabled, then re-apply the Varnish Configuration (under System > Varnish Management).

[InsiderEdsHead]

I have seen an improvement using pageSpeed, although it's somewhat misleading as it takes into account multiple factors, not just response times. By merging and compressing css and js files i see a huge improvement on top of what i get from varnish. What i did notice is that in varnishstat i'm seeing a hit ratio of 70% when the server is busy - this is great since i have over 50k products. On average the hit ratio is between 30-40%.

What i am noticing though is that the cpu load has increased significantly. I'm not sure if it's related, but it seems to have to do with apache. I'm wondering if there is something in the config for varnish / apache that would cause this?

Overall great extension and works with ease (which is nice).

Ed

[aheadley]

What i am noticing though is that the cpu load has increased significantly. I'm not sure if it's related, but it seems to have to do with apache. I'm wondering if there is something in the config for varnish / apache that would cause this?

Varnish itself is extremely light on CPU usage so that shouldn't be the cause. If you're using Turpentine version 0.1.x then the ESI block generation could maybe result in a higher number of individual requests to the backend (Apache) but those ESI blocks can be cached and removing the full page generation load from Apache means this seems unlikely. Do you mean CPU load has gone up since installing Turpentine or since changing the Set Initial Cookie option?

[InsiderEdsHead]

I'm on version 0.1.2. I'm not sure if it's since installing Turpentine or with the initial cookie set. I disabled the module today and reverted back to apache because my service provider doesn't support varnish. I'm going to reinstall tomorrow and can play with the options and let you know. CPU load has gone down a bit since disabling, but I'm not sure this was the cause as of yet.

[aheadley]

I'll close this ticket for now, feel free to re-open/reply once you've done your testing.