RACS: a case for cloud storage diversity
This paper presents RACS, a proxy that transparently spreads the storage load over many storage providers. Stripping user data across multiple providers can allow customers to avoid vendor lock-in, reduce the cost of switching providers, and better tolerate provider outages or failures. RACS incorporates erasure coding rather than replication to tolerate provider failures.
providers charges with bandwidth in/out, requests and hosting the actual data, and compete aggresively on price, thus expensive to switch from one provider to another.
replicate data to multiple providers has high storage and bandwith cost.
RACS assumes a minimal storage interface (put, get, delete, list)
Outages and economic failures are considered and distinguished.
RACS decreases the prohibitive costs of such failures and vendors lock-in.
RACS stripe data accross multiple providers using erasure codes, and retrives chunks from surviving nodes to recover loss data when node failure occurs.
This paper implements RACS and REST interfaces for different storage providers. for multiple proxy, they use ZooKeeper to implement synchronization.
Avoid vendor lock-in
Provide fault torlerance for both storage nodes and proxy (multiple proxy)
Proxy-based transparent interface to client
Saving costs compared with replication
Transient failures are not considered (in future work)
Relatively simple prototype implementation.