Exclusive discount for LET members, get our S3-compatible object storage at $36/TB charged Yearly

Comments

• sayem314 Member

  May 10

  @TimboJones said:

  @sayem314 said:

  @systemfreaks said:
  Nobody asked you to open a ticket or move this privately.

  Questions about replication, infrastructure, backend setup, and failure domains have already been answered publicly on this forum before, including during the BETA discussions and previous offers. A quick search will show those answers.

  That said, you are welcome to ask your exact technical question again here and I will answer it publicly.

  But please either ask the question directly or stop making unsupported claims about the service.

  You need pay attention first because Timbo was the one saying “open ticket/DM/email” and thats exactly what i replied to. You didnt even understand context properly.

  Actually, you need to read. I said you would have done those things if you actually were interested in the answer and their services. You're all over the place, just stop.

  And since you asking again, my original question was basically:

  what storage stack are you using? ceph/minio/custom?
  what durability model?
  replication factor or erasure coding layout?
  if EC, what config? something like 4+2, 8+3 etc?
  how many simultaneous disk failures or node failures can cluster tolerate without object loss or unavailability?
  is storage distributed across multiple physical servers or mostly RAID-backed single nodes?
  single datacenter only or cross-node/cross-rack redundancy?

  Wait, you were saying "question" before like a single question but you asked several technical questions that very likely need to be answered by someone else, like someone in Support.

  These are very normal questions for object storage customers because durability architecture matters more than raw $/TB pricing.

  Please dont even answer now if this was already answered publicly before. Just link:

  • docs/faq
  • previous LET thread
  • beta discussion
  • forum reply

  where these exact things are explained and i will shut up immediately.

  Because i genuinely checked your site first and did not find this information.

  And also no need get angry if someone repeats technical questions. This is storage infrastructure business, people will repeatedly ask about EC layout, replication, failure domains, durability guarantees, backend stack etc before trusting provider with large datasets.

  Don't engage with this user. You don't want him for a customer.

  Timbo shut your goofy ass up.

  Nobody was even talking to your dumb fuck ass and you jumped in crying like unpaid customer support.

  Actually embarrassing.
• systemfreaks Member, Patron Provider

  May 10 edited May 10

  @whynotlearn said:
  Hey @systemfreaks hope you are having a good day.

  I am sorry if this has been asked before but I wish to ask again if it hasn't just in case but can you please tell me more about your fair use case policy for unmetered bandwidth, I think other people might be interested in hearing this too!

  And good luck with sale! :-]

  Hey !

  Buckets may use up to 1 Gbps continuously, 24/7.

  A Fair Use Policy applies only in exceptional cases, such as hardware failure or temporary network capacity reduction caused by an outage. In such situations, until the issue is resolved, larger buckets may be prioritized for available connection speed over smaller buckets to maintain service stability.

  For public buckets, we also use a CDN, so any speed reduction is expected to be rare and unlikely under normal conditions.

  Thanked by 1whynotlearn
• webontop Member

  May 10

  @systemfreaks said:

  @webontop said:

  @systemfreaks said:

  @webontop said:

  @systemfreaks
  Connection between nodes is at 50 Gbps - Raid is not used since is not recommended with erasure coding
  Being an OVH partner https://partner.ovhcloud.com/en/partner/systemfreaks-ltd/ All servers are based in OVH data-centers current nodes in use HGR-STOR-1 which supports hot swap of disks.

  This is not my game. Just asking because you mentioned this - I have 3 simple questions and need 3 simple answers.

  1. As per your info (EC - you using) - you have NO Offsite Backup - So, If someone runs rm -rf / or ransomware hits, EC will happily "erase" the data across all nodes simultaneously. RIGHT?
  2. As I know - Your nodes are in the same rack at OVH (OVH Racking exactly like that) and a Top-of-Rack switch fails, the whole cluster goes dark. RIGHT?
  3. What is your K+M Config?

  Im asking because you are going for an extremely low price, which creates an extremely high risk. Google OVH destroyed how much data in the last 15 years!

  How come you consider 3x replication across Canada, the US, and Europe "no offsite backup ? "

  Same OVH network is not considered as an off-site backup! And it's RISKY! low cost tomaintain but risky.

  Also mention - While 8+2 EC is efficient (~25% overhead), doing 3x replication on top of that means your total raw-to-usable ratio is massive.

  There is also - Slow write speeds due to trans-continental round-trips. High bandwidth (50 Gbps) does not solve the speed-of-light delay between Paris and Montreal.

  Your durability is entirely dependent on OVH’s backbone and internal DC routing.

  I think it may concern you! and it's more..

  MinIO does not rely on a separate central bucket-index metadata database for object visibility. Object/version metadata is stored with the object itself in xl.meta files on the drives. Therefore, the “bucket index database desync” argument is not an accurate description of MinIO’s architecture. Async replication can still have temporary cross-region lag, but that is replication delay, not metadata database corruption.

  Object storage writes to the local/primary cluster first, then replication happens in the background to the other regions.

  At this point, I believe the conversation has reached its end.

  You are still making assumptions and presenting them as facts about a service you have never used. That is unfair and misleading.

  You are absolutely free not to use our service, but please do not defame it based on incorrect assumptions.

  Look bro Im not against your business - I'm made the inquiry because of

  Connection between nodes is at 50 Gbps - Raid is not used since is not recommended with erasure coding
  Being an OVH partner https://partner.ovhcloud.com/en/partner/systemfreaks-ltd/ All servers are based in OVH data-centers current nodes in use HGR-STOR-1 which supports hot swap of disks. 
  

  This - As the program goes, OVH CA can't b lower than $20.00 PM. As you are a partner (Sorry! Im giving an example) can't be lower than $15-$12PM.. other locations $8 PM. If you are using OVH infrastructure and offering $4 or less means you are burning your own money! You can be not an issue!

  Then I take a tour to some of the companies previously asociated by OVH! Where I found $10,000 in free cloud credits Startup Program by OVH! They just burn out the Free CASH! When they ran out, they tried to increase the pricing, but some people never checked their mail or couldn't move - silently just vanished with their businesses.

  Im not ruining your rep! Just tried to understand how you can offer such a price with OVH infra.

  Some gous tried with multiple accounts. A company I can't name them they actually bought clearance hardware from OVH pays only the internet cost and started the s3 by own and failed too..

  Sorry to bother you! Thanks.
• systemfreaks Member, Patron Provider

  May 10 edited May 10

  @webontop said:

  @systemfreaks said:

  @webontop said:

  @systemfreaks said:

  @webontop said:

  @systemfreaks
  Connection between nodes is at 50 Gbps - Raid is not used since is not recommended with erasure coding
  Being an OVH partner https://partner.ovhcloud.com/en/partner/systemfreaks-ltd/ All servers are based in OVH data-centers current nodes in use HGR-STOR-1 which supports hot swap of disks.

  This is not my game. Just asking because you mentioned this - I have 3 simple questions and need 3 simple answers.

  1. As per your info (EC - you using) - you have NO Offsite Backup - So, If someone runs rm -rf / or ransomware hits, EC will happily "erase" the data across all nodes simultaneously. RIGHT?
  2. As I know - Your nodes are in the same rack at OVH (OVH Racking exactly like that) and a Top-of-Rack switch fails, the whole cluster goes dark. RIGHT?
  3. What is your K+M Config?

  Im asking because you are going for an extremely low price, which creates an extremely high risk. Google OVH destroyed how much data in the last 15 years!

  How come you consider 3x replication across Canada, the US, and Europe "no offsite backup ? "

  Same OVH network is not considered as an off-site backup! And it's RISKY! low cost tomaintain but risky.

  Also mention - While 8+2 EC is efficient (~25% overhead), doing 3x replication on top of that means your total raw-to-usable ratio is massive.

  There is also - Slow write speeds due to trans-continental round-trips. High bandwidth (50 Gbps) does not solve the speed-of-light delay between Paris and Montreal.

  Your durability is entirely dependent on OVH’s backbone and internal DC routing.

  I think it may concern you! and it's more..

  MinIO does not rely on a separate central bucket-index metadata database for object visibility. Object/version metadata is stored with the object itself in xl.meta files on the drives. Therefore, the “bucket index database desync” argument is not an accurate description of MinIO’s architecture. Async replication can still have temporary cross-region lag, but that is replication delay, not metadata database corruption.

  Object storage writes to the local/primary cluster first, then replication happens in the background to the other regions.

  At this point, I believe the conversation has reached its end.

  You are still making assumptions and presenting them as facts about a service you have never used. That is unfair and misleading.

  You are absolutely free not to use our service, but please do not defame it based on incorrect assumptions.

  Look bro Im not against your business - I'm made the inquiry because of

  Connection between nodes is at 50 Gbps - Raid is not used since is not recommended with erasure coding
  Being an OVH partner https://partner.ovhcloud.com/en/partner/systemfreaks-ltd/ All servers are based in OVH data-centers current nodes in use HGR-STOR-1 which supports hot swap of disks. 
  

  This - As the program goes, OVH CA can't b lower than $20.00 PM. As you are a partner (Sorry! Im giving an example) can't be lower than $15-$12PM.. other locations $8 PM. If you are using OVH infrastructure and offering $4 or less means you are burning your own money! You can be not an issue!

  Then I take a tour to some of the companies previously asociated by OVH! Where I found $10,000 in free cloud credits Startup Program by OVH! They just burn out the Free CASH! When they ran out, they tried to increase the pricing, but some people never checked their mail or couldn't move - silently just vanished with their businesses.

  Im not ruining your rep! Just tried to understand how you can offer such a price with OVH infra.

  Some gous tried with multiple accounts. A company I can't name them they actually bought clearance hardware from OVH pays only the internet cost and started the s3 by own and failed too..

  Sorry to bother you! Thanks.

  You are making repeated assumptions about our business and presenting them as facts.

  We are not reselling OVH Public Cloud S3, and we are not running on startup credits. We operate our own storage stack on dedicated OVH infrastructure.

  The service has been running for over a year without the issues you are implying.

  Questions are welcome, but this has crossed into speculation and abuse. If the service does not meet your needs, that is fine, but please stop derailing the offer thread.
• TimboJones Member

  May 10

  @sayem314 said:

  @TimboJones said:

  @sayem314 said:

  @systemfreaks said:
  Nobody asked you to open a ticket or move this privately.

  Questions about replication, infrastructure, backend setup, and failure domains have already been answered publicly on this forum before, including during the BETA discussions and previous offers. A quick search will show those answers.

  That said, you are welcome to ask your exact technical question again here and I will answer it publicly.

  But please either ask the question directly or stop making unsupported claims about the service.

  You need pay attention first because Timbo was the one saying “open ticket/DM/email” and thats exactly what i replied to. You didnt even understand context properly.

  Actually, you need to read. I said you would have done those things if you actually were interested in the answer and their services. You're all over the place, just stop.

  And since you asking again, my original question was basically:

  what storage stack are you using? ceph/minio/custom?
  what durability model?
  replication factor or erasure coding layout?
  if EC, what config? something like 4+2, 8+3 etc?
  how many simultaneous disk failures or node failures can cluster tolerate without object loss or unavailability?
  is storage distributed across multiple physical servers or mostly RAID-backed single nodes?
  single datacenter only or cross-node/cross-rack redundancy?

  Wait, you were saying "question" before like a single question but you asked several technical questions that very likely need to be answered by someone else, like someone in Support.

  These are very normal questions for object storage customers because durability architecture matters more than raw $/TB pricing.

  Please dont even answer now if this was already answered publicly before. Just link:

  • docs/faq
  • previous LET thread
  • beta discussion
  • forum reply

  where these exact things are explained and i will shut up immediately.

  Because i genuinely checked your site first and did not find this information.

  And also no need get angry if someone repeats technical questions. This is storage infrastructure business, people will repeatedly ask about EC layout, replication, failure domains, durability guarantees, backend stack etc before trusting provider with large datasets.

  Don't engage with this user. You don't want him for a customer.

  Timbo shut your goofy ass up.

  Nobody was even talking to your dumb fuck ass and you jumped in crying like unpaid customer support.

  Actually embarrassing.

  So you whine when nobody is talking to you and when told what you did wrong you whine someone is talking to you. You're the only one crying, I'm pointing out what you did wrong and how to do better next time.

  Nobody wants to do business or talk to you.
• jonathann Member

  May 13

  @systemfreaks said: CDN for public buckets

  does it support custom domain name, example: "images.example.com" ?
• systemfreaks Member, Patron Provider

  May 13

  @jonathann said:

  @systemfreaks said: CDN for public buckets

  does it support custom domain name, example: "images.example.com" ?

  Only if you rewrite using cloudflare or your own reverse proxy , but it's something we are working on implementing including multiple buckets sharing the same quota
• yoursunny Member, IPv6 Advocate

  May 13

  Our object is smol.
  Can we buy 0.000001 TB and but pay for 1000 years?
• systemfreaks Member, Patron Provider

  May 13

  @yoursunny said:
  Our object is smol.
  Can we buy 0.000001 TB and but pay for 1000 years?

  Haha, i am sure we can work some special deals up to 5years if cost - transaction fees make sense

  Thanked by 1yoursunny
• jonathann Member

  May 13 edited May 13

  @systemfreaks said: Only if you rewrite using cloudflare or your own reverse proxy , but it's something we are working on implementing including multiple buckets sharing the same quota

  thanks; did you mean rewriting (e.g. with nginx reverse proxy) from something that looks like https://sf-bucket-name.sf-objectstorage.com/something/somefile.jpg as the upstream to https://images.example.com/something/somefile.jpg (images.example.com = my custom domain name) where images are served?

  does the bucket name need to be an exact match to the custom domain name?
• systemfreaks Member, Patron Provider

  May 13 edited May 13

  @jonathann said:

  @systemfreaks said: Only if you rewrite using cloudflare or your own reverse proxy , but it's something we are working on implementing including multiple buckets sharing the same quota

  thanks; did you mean rewriting (e.g. with nginx reverse proxy) from something that looks like https://sf-bucket-name.sf-objectstorage.com/something/somefile.jpg as the upstream to https://images.example.com/something/somefile.jpg (images.example.com = my custom domain name) where images are served?

  does the bucket name need to be an exact match to the custom domain name?

  Do you already have a bucket with us?

  If yes, we can potentially send you a ready-to-use config file.

  Also, I will inform the team that this should be added to our knowledge base.