Q: When is JIT compiled code performance higher than c++? See separate blog
Q: difference between JVM stack vs native stack?
Q: ThreadLocal internal implementation?
Q: data structures with concurrent modification notifications — how is it implemented?
— IPC between processes (language-neutral) —
Q: how is shared memory managed?
Q: messaging uses sockets and has high overhead. What other solutions can maintain FIFO?
%%A: nothing new. The earliest MOM has dealt with this problem long ago. Perhaps multiple files with single producer and single consumer would be ideal. The 2 processes need to operate on both ends of the file. (There could be some kernel support for this.) See https://coderanch.com/t/278842/java/Reading-writing-concurrent-threads-file
Q: what kind of jdk locks have you used?
%%A: readwrite lock, reentrant lock
Q: How would you size your thread pool, based on processor count?
Q: For a market data gateway, when would additional threads help (and when would they be useless or counterproductive)?
%%A: I/O bound, the processors could be 99% idle. More threads would increase the utilization rate. Ideal is simultaneous saturation.
Q: thread cancellation without using Futrues?
Q: default methods in interface — is it a breaking change? http://stackoverflow.com/questions/22618493/does-introducing-a-default-method-to-an-interface-really-preserve-back-compatibi has a concise answer
Q: how is lambda implemented in java 8? See my separate blog post.