We are currently on Cloudinary but are looking at alternatives. We've recently enhanced our e-store with upgrades on functionality in order to stay competitive in the e-comm space.
The current key focus is to increase sales volume by driving more traffic and sales conversions on our e-store, But the skyrocketing cost of the Cloudinary monthly subscription is holding us back. We needed to take the moment to review the scalability of our current setup.
We are even currently afraid to work on advertising campaigns online because we are afraid of that:
- The e-store takes too long to load
- Skyrocketing costs on Cloudinary when more visitors are on our pages.
Summary Overview (on what we want to achieve):
- Keep the cost sustainable
- Scalability - manage and control costs
- Specifications that cannot be compromised - high res images, acceptable time to load, etc.
What is the baseline and how can we measure them accurately? (A new user in incognito and not fetching from local browser caches, etc)
Findings:
- Assets: 2500 images on average per month
- Image Size: 2500x3500 (orginal) which is compressed into 900x1260
- Bandwidth: May-Oct traffic numbers (100k Total). 20k per month
- Location: 80% of the traffic is local traffic
- Does the internal traffic impact on the CDN usage? Our internal team has been accessing and updating the live site a lot.
- Measurements - scientific way to measure the differences between the solutions offered TTFB, FCP measurement
I would highly recommend using your cloud provider CDN whether this AWS, GCP or AZure on top of your object store SO, GCS or Blobstore. It will greatly reduce the cost while maintaining speed and performance. For image compression https://github.com/imazen/imageflow is amazing plus it's written in rust which really helps with safety and speed.
Hello, We have been at similar situation like yours. We deal with serving around 1TB of images for one of our client, and it skyrockets the price way too quickly if we are using any cloud provider services. If and only if you have time or devops people to spare into research I'd recommend setting up your own CDN service in one/multi cloud provider. In my experience going with custom built solution is cheaper. We are using this https://github.com/ilhaan/kubeCDN, it's a CDN cluster setup on k8s cluster.
For image compressing, we have our inhouse library to compress images without losing much resolution making it almost impossible for regular human to detect changes. (Github link)[https://github.com/Comet-App/CImage].
Testing and benchmarking are always messy, cause in our case we can't scale it with existing tools so we simulate acutual 10k req on client's platform. For individual scores, we prefer (GTMetrix)[https://gtmetrix.com/] and (loader.io)[https://loader.io/]. We are also working on custom multiregion supported webpage/api benchmarking platform, will share with if you're interested.
