How Canva saves Amazon S3 costs

Seems crazy to run such large infrastructure on a major cloud.<p>Tens of millions wasted.<p>Canva is likely trapped in S3 never to exit. The cost of getting their data out makes it impossible.<p>S3…. the Hotel California of the cloud. You can check in any time you like but you can never leave.<p>S3 is 9 cents per GB egress fees.<p>Cloudflare R2 charges zero egress fees.<p>At list price of 9 cents per GB it would cost Canva about USD$20 million to export their 230 petabytes. Let me know if my calculation is wrong.

A lot of people in this thread should start consultancies that transition companies off the cloud.<p>Apparently, it would be easy for them to save tens of millions, in turn making millions for the consultant.

I&#x27;m really surprised how much anti cloud opinion is being voiced here (but maybe I should have known better, after all these years on HN).<p>Even if you don&#x27;t use any other cloud tech in your stack, my advice is, use S3 (or equivalent)! That level of durability, availability, and scalability, is not trivial to pull off, and is best made somebody else&#x27;s problem. Even when you&#x27;re as big as Canva.

Something which is often overlooked is that transitioning objects in S3 to another storage class using lifecycle policies is a one-way operation. If you later notice that the chosen storage class doesn&#x27;t fit or your use case changes, you can only go &quot;forward&quot; using lifecycle policies and not &quot;back&quot; to e.g. &quot;S3 Standard&quot; [1]. Going &quot;back&quot; is therefore a quite costly and cumbersome operation, as you have to apply the change of the storage class to each object individually.<p>[1]: <a href="https:&#x2F;&#x2F;docs.aws.amazon.com&#x2F;AmazonS3&#x2F;latest&#x2F;userguide&#x2F;lifecycle-transition-general-considerations.html" rel="nofollow noreferrer">https:&#x2F;&#x2F;docs.aws.amazon.com&#x2F;AmazonS3&#x2F;latest&#x2F;userguide&#x2F;lifecy...</a>

Canva is a case where AWS made sense for the founders. The founders anyway made billions and who cares if you give a few hundreds millions to Jeff and have to hire expensive AWS consultants if you can avoid hiring a few ops people that actually know how to operate a server &#x2F;s<p>More seriously, I think the sweet spot is to buy 2x o 3x what you expect to need but do it off the cloud.<p>At Canva&#x27;s scale you&#x27;ll spend hundreds of thousands, save literal millions and have way more headroom and stability than with crappy auctioned instances on AWS.

Articles like this are a great example of how you can be a hero for fixing a problem, but get no recognition for preventing one.

Surely they negotiated a private pricing agreement given the volume of data&#x2F;usage...one would hope.<p>Having been a part of these negotiations for services other than S3, DTO discounts can be generous if you have a good forecast (3-5 years) on your data egress.

Offtopic — is it some cardinal sin to include a link to the home page from a documentation? I dont know what canva does, if I stumble on a documentation page first, should I not be allowed to see the product&#x27;s landing page at some point? This is so prevalent, it has to be a conscious choice.

Real question, how to scale a business to petabytes of data without cloud? How is it done in practice?

If you&#x27;re at a point of spending $1.6M to transition objects from one class to another and spending millions per month in storage costs -- you need to have a real conversation about if storing your data in a vendor-locked cloud is the right path forward. S3 isn&#x27;t the only option, MinIO + dense storage is one viable option if your spend is high enough to justify running MinIO. Backblaze is another.<p>List price of Backblaze:<p>$0.005&#x2F;GB - Base Storage<p>$0.01&#x2F;GB - Egress (Use a data partner to bring this down to $0)<p>$0 - PUT API Requests<p>$0.004&#x2F;10,000 - GET API Requests<p>No delete penalties a&#x2F;k&#x2F;a minimum storage<p>List price of Glacier Instant Retrieval:<p>$0.004&#x2F;GB - Base Storage<p>$0.09&#x2F;GB - Egress<p>$0.02&#x2F;1,000 - PUT API Requests<p>$0.01&#x2F;1,000 - GET API Requests<p>Billed for 90 days of storage

There is no way they need 230 PB saved. That&#x27;s javascript-tier design in practice. If you can put almost all of it into glacial storage then that&#x27;s because you don&#x27;t actually need to store it.

&gt; Canva now has users across 190 countries, who design in over 100 languages thanks to some amazing localization work done by our Internationalisation team.<p>How may of those countries are profitable? My sense is Internationlization is a relatively fixed cost, but how does customer support in Burma, for example, not trump any realizable revenues there?

This is an interesting post...though I&#x27;m surprised to see this kind of optimization occur so late stage&#x2F;as large as Canva.<p>There&#x27;s quite a few anti-cloud opinions in the comments but for anyone who is hosted on S3 and would like this same level of automated analysis for cost savings on S3 (and other AWS services), we basically profile for this out-of-the-box on Vantage and it&#x27;s free to get an optimization check to see savings: <a href="https:&#x2F;&#x2F;www.vantage.sh&#x2F;" rel="nofollow noreferrer">https:&#x2F;&#x2F;www.vantage.sh&#x2F;</a><p>Not meant to be as shameless of a plug, but seems very relevant to the topic at hand for onlookers.

Yeah no, R2 is a way better choice than being vendor locked like this when burning millions

&gt; 100 million monthly active users<p>Oh, my. That&#x27;s a lot. What does Figma? 300?