tawfiqul-islam / RM_DeepRL

Resource Management with DeepRL using TF Agents
12 stars 5 forks source link

Executor Placement of Job #4

Open VishnuVIVE opened 1 year ago

VishnuVIVE commented 1 year ago

print.pdf Dear Sir, I have one doubt in the program flow . When i have implemented the program and run for only one job with the following configuration why the job is not placing ail the executor of Jobs . Example: I have taken Virtual machine as is it(a s program has input) as Vm type =3 an each having price as 1,2,3 their CPU and memory capacities are vm_1-> 4,12:4,12:4,12 Vm->_2 8,24:8,24,8,24: VM_3-> 12,36:12,36:12,36 respectively. No. of Jobs input as 1 with the configuration arrival time= 57, id=1,job type=3, CPU=3, mem= 4 , No. of Ex= 6 and duration =100. while running the code it is placing first executor so remaining placement will be 5 which is correct. but when i am printing the Possible placement it is printing values Possible placement as=1,-1,0,5,10,15,24,33,42. after that successful placement again printing current executor placed=1 . but Reaming placement is still = 5. i think that should decrease to 4 as two executor where placed ..... and so on it is not decreasing like 3,2,1 instead it coming as reaming placement as 5,4,5,4 and some time 5,5,. I have run the reinforcement algorithm for only 1 iteration with only one Job and with beta value =0.25. Kindly can You please elaborate how executor placement is happening. Please find attachment Program run