I have a small doubt with respect to the way resources are allocated in Mesos cluster manager w.r.t SPARK. Can the applications ask the resources specifying the data locality or is that Mesos shows all the resources that is available to a framework above it and it is the scheduler of the framework that has to chose as to which resource it wants to run on???
I got this doubt when i was trying to compare the potential pros/cons of MESOS Vs YARN scheduler. In YARN, applications when they send "Resource Request", it is like a vector <#containers, rack/node locality, configuration(cpu,memory...), priority>. Yarn tries to allocate containers in the local rack/node to the maximum extent possible.
Does Mesos do the same or is it different and if so, whats the change.???
Thanks in advance for your reply guys !!
Sai Prasanna. AN
II M.Tech (CS), SSSIHL
Entire water in the ocean can never sink a ship, Unless it gets inside. All the pressures of life can never hurt you, Unless you let them in.