Open eksoverzero opened 10 years ago
Here is the start of some normal staging operations. A solid mix of cached vs uncached requests:
Just incase it was question, this is my response time to the magicaly server:
64 bytes from 54.72.87.102: icmp_seq=0 ttl=47 time=222.495 ms
64 bytes from 54.72.87.102: icmp_seq=1 ttl=47 time=227.093 ms
64 bytes from 54.72.87.102: icmp_seq=2 ttl=47 time=330.777 ms
64 bytes from 54.72.87.102: icmp_seq=3 ttl=47 time=236.994 ms
64 bytes from 54.72.87.102: icmp_seq=4 ttl=47 time=281.330 ms
64 bytes from 54.72.87.102: icmp_seq=5 ttl=47 time=223.149 ms
64 bytes from 54.72.87.102: icmp_seq=6 ttl=47 time=269.614 ms
64 bytes from 54.72.87.102: icmp_seq=7 ttl=47 time=223.872 ms
64 bytes from 54.72.87.102: icmp_seq=8 ttl=47 time=311.896 ms
Wow, nice work! Any takeaways for you? Nothing here jumps out as me as a glaring problem or something Magickly can do much about, since it is CPU-bound at the ImageMagick level. I will comb through the info above again in a day or two, though.
Resizing is a fairly simple operation – the big performance bottlenecks happen when multiple/complex effects are applied. https://github.com/afeld/magickly/issues/16 will help greatly.
Hey,
Just thought I'd share some test results with you quick.
In case you get bored to death before you get to the bottom of this, somewhat useless piece of information, thanks.
So, I'm here...
... the server is here...
...and the src is here:
Using Siege I started a siege doing a 'resize=640' and then a few minutes later asked someone else to hit a new one, 'resize=1080', and then a few minutes later again, another person using 'resize=720'
I took 2 screenshots and a
top -c
output copy/paste from the server. Once as the second siege got started and then again while all 3 were running.Here are the app and monitoring stats: