java - Fast and scalable IO passthrough -


original

i have batch job passes through contents downloads various urls s3 storage. using blocking io , have reached point job io bound blocking of time because of io. in order speed whole process thinking using non-blocking io.

unfortunately wasn't able find utility code passing through content set of channels set of channels. since read writing correct non-blocking code not easy, prefer use existing utility/framework write code myself.

the transfermanager seems possible option higher throughput when using aws sdk, offers usage of streams , seems use io threads in background. apparently there no out of box option non-blocking uploads s3.

what recommend? right can imagine following solutions.

  1. stay blocking io , use own io thread pool
  2. use non-blocking io download files local filesystem, upload transfermanager
  3. use non-blocking io pass through

option 1 not scale , 2 work while, keep ios on ebs low i'd rather use 3.

to implement option 3 guess have implement lot myself final question is, whether think it's worth , if so, tools use make work.

edit 1

clarifying io bound meant job waiting io. here can see bandwidth not saturated, if possible. monitoring view

if job i/o bound you've finished. you're bound speed of network, not of code. using nio won't make faster.

clarifying io bound meant job waiting io.

yes, that's 'i/o bound' means. nothing has changed.

here can see bandwidth not saturated, if possible.

you try using larger buffers, said seems me you've finished.


Comments

Popular posts from this blog

angularjs - ADAL JS Angular- WebAPI add a new role claim to the token -

php - CakePHP HttpSockets send array of paramms -

node.js - Using Node without global install -