https://discord.cloudflare.com logo
Join Discord
Powered by
# r2
  • k
    Would of thought they'd push those customers towards Cloudflare Stream
  • k
    I think that charges per minute on the videos?
  • i
    Yeah, there's a difference though, Stream uses HLS but R2 won't, it'll just stream the whole file at top-res.
  • i
    Dealbreaker if you're a streaming company.
  • i
    But for archiving, R2 would be great.
  • k
    I registered for R2 to have a play around but since I put a pretty small amount of storage since it was just "ooo shiny" I'll probably get that in a few months
  • b
    I certainly wish for a way to allow public access without having to rely on workers, and only use workers if you want finer control.
  • b
    I feel like vast majority of the use case will just be controlled upload + public access, and getting charged for request is just a slightly different way of charging for bandwidth.
  • e
    There will be multiple ways to access R2, but don’t forget that S3 has both request and egress charges. And S3 behind Cloudfront has 2 requests charges + egress fees.
  • а
    Please, consider Wasabi and B2 as competitors when setting up prices. Amazon S3 is much bigger then just object storage.
  • k
    similarly r2 provides more than what both wasabi and b2 provide, if you want multi region on either of those, pretty sure r2 will be cheaper
  • i
    Keep in mind that the pricing quoted in the blog post is not final, and the people who set pricing at Cloudflare will definitely take all of that into account 🙂
  • а
    We use wasabi with multi region (two buckets in AMS and NV2 with hand made mirroring) $12 pet Tb (6 + 6).
  • а
    ~100 Tb in total now
  • g
    Hand made mirroring sounds interesting! We use Wasabi also. Did you do this with the API?
  • а
    We just upload to two buckets in different regions.
  • g
    How do you automate it?
  • k
    that's very interesting
  • а
    We have local minio installation. customers upload to minio. Then cron job upload from minio to Wasabie's two buckets. When we get read request, we first check in minio, then Wasabi bucket number one, then Wasabi bucket number two.
  • i
    Hmm... interesting. The only problem I could see with that is that if an upload got cut off in the middle of Bucket 1 upload, then any read requests would be served the corrupted file since Bucket 1 is checked first.
  • а
    No. Read requests first check minio. There is always the freshest version. We remove file from minio only after successful upload to two Wasabi buckets.
  • i
    Do you do checksum validations on the objects? If an upload fails but looks like it succeeds (network troubles etc) then Minio will delete the file but the bucket won't have a correct version.
  • a
    Wasabi has a minimum storage retention policy 30-90 days (pay as you go and contracts), so final cost can be extremely expensive depending on use case.
  • а
    We do not go so far. We assume if Wasabi returns code 200 OK on PUT request it is OK.
  • а
    You r right. For our use case it is OK. "Deleted storage" costs us only about 2% of the bill. We store images for a 2-3 months minimum.
  • i
    Images? Have you looked into Cloudflare Images?
  • а
    Yes I did. CF Images is about 25x more expensive for our use case.
  • m
  • m
    Where is the catch? This is too good to be true
  • i
    There's a minimum duration (a number of days billed, even if you delete immediately)... also, this should be moved to #812577823599755274 if it's no longer about R2.
1...678...1050Latest