Notes
Lecture 17
----------
Questions?
- OceanStore cache encrypted?
- did anyone actually use pond/oceanstore?
- why later systems used replication vs erasure codes?
- was encryption really everywhere, including replicated fragments?
- search in contents of files?
- issues with MACs?
Do we have anything like OceanStore today?
- untrusted cloud infrastructure for storage?
NOPE
- while you can encrypt remote data, you have to manually do multi-cloud deployments
and deal with everyone's APIs
- well, you could do S3-like things with multiple clouds, keep data encrypted
- but you're working against how they want you to use the systems
Note that OceanStore has no provisions for search or comparison
- kind of goes against its design
- modern work on search, computation on encrypted data, but not practical yet
What distributed computing & storage infrastructure do we have today that is "untrusted"?
- cryptocurrencies, i.e., blockchain-based systems
- these are VERY WASTEFUL, but why?
For the rest of this class, we're going to be talking about trust
- nothing works without some sort of trust
- the trick is deciding what to trust, making that trustworthy,
and to make the system fail gracefully when trust is violated
What is trusted in OceanStore?
- the software!
- encoding, replicating, encrypting, etc
- small bug can lead to data loss, privacy breach, really anything
- and thus, the developers of the software
- note the design is pretty monolithic, not much room for alternative mechanisms
- who is responsible for fixing the bugs? UNCLEAR
Contrast this with Amazon S3
- bugs will happen, but are you trusting the software or the organization?