That really sums up the cloudflare experience and this is from someone heavily invested in their workers platform. They have lots of products and keep pumping more but except for DNS, most of them are half assed with weak maintenance/support.
That really sums up the cloudflare experience and this is from someone heavily invested in their workers platform. They have lots of products and keep pumping more but except for DNS, most of them are half assed with weak maintenance/support.
That's not a fair take. I will give Cloudflare a lot of shit for some of their products, but some of their products are 100% best in class. For instance, R2 is just better than S3, and KV is better than AWS/GCP options. The pricing is better, it's multi-region by default and there's less ops overhead.
R2’s writes are much slower than S3/GCS. It’s not a good fit unless your workload does many reads for each write (e.g. assets). Also it is notably not multi-region - it just picks a region automatically based on where you created the bucket from.
KV is super expensive - once you’re operating at non-trivial scale, reading another configuration value per-request in your worker starts to cost thousands per month. KV’s tail latencies are also surprisingly bad (I’ve seen over a minute), even for frequently read keys that should be easily cached.
This is good to know. I haven't used R2, it's been on my radar but I haven't taken the steps to start using it. Partly because my experience with the rest of Cloudflare has been middling to poor. I'd love to save on our S3 bill, which is substantial, but it's going to take significant development to get there and it's an unknown how much it'll actually save. There are too many stories of people getting called by enterprise sales when their usage crosses some line in the sand that only the sales people know.
I agree with R2 but KV is un-realiable. I said DNS but I meant CDN which R2 kind of falls into. Cloudflare is good in moving lots of data but most of their other products are not polished. It doesn't mean that they are not exceptional products. I have deployed a wasm-worker 5 years ago and it is still up and running to this day. I don't think a server would have survived or any other product from any other provider would have guaranteed such backward compatibility.
R2 is very high latency with huge variance, definitely lower quality than S3.
In my experience even backblaze b2 performs (way) better.
Their community forums are full of such reports.
KV is so expensive that it’s barely usable, and like R2, is very slow.
Slightly higher latency. I've seen about 20-30% increase from S3 to R2. But the bill is magnitudes lower.
Agree with the KV point, Upstash is the same. But I just use dragonflydb on a single VM. No point paying for transactions.
Hell, S3 could have 20ms latency and it wouldn't matter since I can't afford it.
Where are you seeing orders of magnitude lower R2 bills? The storage price for S3 is $0.023 per GB, and the price for CF is $0.015 per GB. The operation pricing is even more similar - S3 is $5/million writes and $0.40/million reads, while Cloudflare is $4.50/million writes and $0.36/million reads.
Egress fees are the largest part of many AWS bills, this is by design, that price is deceptive. R2 has no egress fees.
I literally know an engineer that works on the storage layer for R2 and even he wouldn’t agree that it is better than S3
He wouldn’t disclose any details to me but from point of view S3 was best in class