Need a high-performance backend supporting 200GB of binary blobs for an Erlang app requiring 500 reads and updates per second on commodity hardware. Is Berkeley DB a viable alternative to mnesia/dets?
hey, cassandra seems promissing - scales well with heavy binry workload. u might need extra config but its perf gains could outweigh bdb’s limits for your setup.
An Erlang-native solution like Riak KV may serve as an ideal alternative. In my experience deploying Riak for a project with substantial binary data, its device-level fault tolerance and ease of clustering provided a robust solution without needing high-end hardware. Riak’s design, built in Erlang itself, offers a smoother integration and reliable performance under workloads comparable to the one outlined, especially once tuned to work efficiently over commodity hardware. This approach addresses both high throughput and scalability for data exceeding available memory.