I've been researching the various S3 storage classes on AWS, and aside from the standard tier, there are several options designed for slower data retrieval. I'm curious about the underlying technologies used for these different classes. For instance, it's clear that Glacier uses magnetic tapes as the slowest option, which makes sense given its cost structure—cheap to store but expensive to retrieve. But what about other classes like Infrequent Access? How are they able to offer cheaper storage than the standard class? I assume standard storage uses HDDs while metadata might use NVME/SSDs. What storage technology could be slower than HDDs but faster than magnetic tapes?
2 Answers
Honestly, Amazon doesn't share specific details about their storage technologies, so a lot of this is just speculation. I think the Infrequent Access tier may use standard storage in one zone with replication to Glacier in other zones. Since it offers lower availability compared to the standard tier, it makes sense that retrieval is slower but priced in between standard and Glacier. I actually suspect Glacier could utilize HDDs, optimized to extend their lifespan. Check out some discussions on platforms like Hacker News for more insights!
The typical response from an AWS employee would be something like 'why does it matter?' But in reality, the specifics are proprietary and kept confidential to maintain a competitive edge. From what I've learned, there's actually no magnetic tape used at least in the eu-west-2 region—only HDD and flash storage. So, while we may not ever fully know what's under the hood, it seems more like a mix of these technologies rather than outdated methods like tape.
That's a wild thought! Do tech companies go to such lengths to outmaneuver each other? Seems like a cutthroat environment!
But how can they maintain such low prices if they're using HDD and flash? Is the pricing still slightly profitable, or is it mainly the standard tier that makes the real money?