humanmade / tachyon

Faster than light image resizing service that runs on AWS. Super simple to set up, highly available and very performant.

Home Page:https://engineering.hmn.md/projects/tachyon/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Slow Performance on large crops

EffakT opened this issue · comments

Currently using Tachyon 2.4.0.
Occurs in both Lamba and Docker.

We have a JPG file that is 14176x6132.
Loading this image with "?w=1600&h=600" we get a TTFB of 1s
Loading this image with "?crop=2437px,0px,10095px,5678px&w=1600&h=600". We get a TTFB of around 4s.

As soon as we add the "?crop=" param, the TTFB increases by just over 3s.

Is this expected behaviour? or is there any way to improve performance?

Hmm I've not seen that before. It could depend on how many resources you are giving Lambda though. The TTFB is likely to linearly map to memory limit you provide the Lambda function (as that also scales CPU linearly too).