13 Commits (bfbe848b4b9d15d61b2898895f55dc1c9468384c)

Author	SHA1	Message	Date
phani	ed234f9ac5	default: async io for http req/resp. sync based io doesn't seem to work with huge data like 500k-1m I tested. They just timeout with concurrency	5 years ago
phani	0c92419978	fixed syscall based sync http-req/resp	5 years ago
phani	b9d7278c54	fixes 99% latency problem here (but a problem with 1m data transfer test, it is broken.)	5 years ago
phani	9528f65b32	Synchronous I/O for http request/response fixes 99%tile latencies ... * Async I/O is being used for reading and writing http request and response respectively. This causes the worker core to steal new tasks as when one sandbox blocks on "read" or "write". This causes true round-robin on timer interrupts, leading to those high latencies. * Ideally, the request/response should be synchronous I belive because we are expected to have "low"/near-realtime latencies and the read/writes are only performed when a client connects, so ideally not block! Plus, we expect only small data transfers (in KBs) so that should also support synchronous request/response processing.	5 years ago
phani	69b1f94fe8	scalable sandbox creation in each runtime core	5 years ago
phani	3fab99424a	fibonacci test update	5 years ago
phani	aa0017f4b0	http request response using uv	5 years ago
phani	8a3e050282	Support TCP (TODO: http parsing based on module specification) ... * Major: - Moved sandbox data struct into linear memory of a sandbox. - (Should work well with other sandboxing mechanisms). - Removed all mallocs from within uv usage, using a pre-allocated memory within sandbox data-struct for read/write. - TODO: convert prints on stdout to a string and send it out with function response body. - TODO: for each function loaded, what is the request and response http header/body specification. To start, keep headers minimal. body either: json, plain. - TODO: track TIMEOUTS. If we go beyond that, return a HTTP 408?	5 years ago
phani	69b6dcf83c	added http-parser	5 years ago
phani	eb8fcf00c4	Work-stealing deque for global request queue ... * all cores steal from that queue. * other bug fixes * tested with NCORES = 2 and 4.	5 years ago
Phani	670e3ad2b5	revert from standalone	5 years ago
Phani	a231796990	for standalone execution	5 years ago
phani	9b0ec6f7dc	Changed directory structure ... * silverfish is a submodule now * memory modules from silverfish runtime is copied in to the aWsm runtime. * runtime/tests/Makefile compiles different tests and silverfish/code_benches..	5 years ago