Lecture 2: RPC and Threads

why use Threads

I/O Concurrency = using mutiple thread to call remote procedure, each thread will call remote i/o and wait for answer

Parallelism

Convience = do something in background or periodically

 

with out Threads

event driven programming = has single thread and single loop (like nodejs) , cant have parallelism , hard to write sequential programming , 

 

Thread chanllenges

threads share same memory = synchronous issue

 

Thread Coordination

channel = just sending data to other thread

sync.cond = give other thread a "kick" to perform its operation

wait group = waiting for every other thread to finish

 

 

crawler with shared memory (= fetch map structure)

done.Done() = decrement by 1 which is what "done.Add" added , we are calling this before ConcurrentMutex because it might fail during gorotuine operation

done.wait() = wait until done get to zero

u string = it uses copy of for_, u:= range urls , because outside u keeps changing and gorotuine might refer to other url which is not our intention

 

crawler with channel

notice worker thread(function) doesn't have any shared memory , doesn't need to worry about lock

ch is channel , worker is sending to that channel while master is receiving from that channel