Hyperscalers don't actually share any savings with customers. In fact, they spend more because they maintain idle capacity for availability, which customers ultimately pay for. While hyperscalers are more expensive by definition, they provide two unique advantages: )Rapid scaling up and down capability )Cost savings on managed infrastructure, eliminating the need for large sysadmin teams managing thousands of servers Cloud PaaS/IaaS makes sense in specific cases: )Startups with high burn rates needing fast scaling and lacking resources for their own datacenter )Services requiring rapid scaling (e.g., when normal load fits in one VM but rush hours like Black Friday need 100-1000s VMs). This often leads to hybrid scenarios - own datacenter for base load, cloud for peak demand )Cases where supporting particular applications in-house becomes too expensive (especially mail servers or big data solutions like BigQuery) Worth noting some relevant terms: "Cloud repatriation" describes moving data/applications back on-premises after cloud migration. Its good if RIPE engineers get familiar with it and stories before moving to cloud, to not regret later. "Cloud cost overrun" refers to unexpectedly high cloud costs. We saw this on a sister project where an automated crawler running BigQuery queries generated a $50k bill instead of the usual $600/month. Some issues are more cloud platform bugs than user mistakes and just ticking bomb. For example, historically with S3 buckets, you'd pay for queries even with access restrictions in place (this may be fixed now, but it was a known issue). Unfortunately, processes at RIPE appear completely non-transparent, relying on just a few people's competence and trust. Experts in RIPE community IMHO can't properly assess what technical challenges the RIPE team faces or what relevant experience they might share. And thats sad. On Thursday, November 07, 2024 21:47 EET, Randy Bush <randy@psg.com> wrote:
Could you please explain how moving the large scale project to "cloud" can theoretically save the money?
i suspect that the theory is that the hyperscale cloud providers have sufficient savings due to scale (leverage with oems, ability and talent to highly automate, ...) that they can afford to share some of those savings with the customer. color me skeptical.
randy ----- To unsubscribe from this mailing list or change your subscription options, please visit: https://mailman.ripe.net/mailman3/lists/members-discuss.ripe.net/ As we have migrated to Mailman 3, you will need to create an account with the email matching your subscription before you can change your settings. More details at: https://www.ripe.net/membership/mail/mailman-3-migration/