Scheduling Parallel Jobs Using Migration and Consolidation in the Cloud

Algorithm 2

Deploy(𝑗, BackfillType)—job deploy function.

   input :  𝑗: the job to be deployed;

     BackfillType: the algorithm who calls this function.

   output:    𝐼𝑠𝑆𝑢𝑐𝑐: indicate whether 𝑗 be deployed successfully or not;

(1)  begin

(2) 𝐼𝑠𝑆𝑢𝑐𝑐=𝑓𝑎𝑙𝑠𝑒;

(3)  if  𝑗   is from the background tier then

(4)  if   𝑗  can still run in its original processors then

(5)     swap the CPU priorities of its bg VMs and fg VMs;

(6)     𝐼𝑠𝑆𝑢𝑐𝑐=𝑡𝑟𝑢𝑒;

(7)   else

(8)     /* Job kill is not allowed in REASY but allowed by KEASY */

(9)     𝐼𝑠𝑆𝑢𝑐𝑐=𝑓𝑎𝑙𝑠𝑒;

(10)   if   𝐵𝑎𝑐𝑘𝑓𝑖𝑙𝑙𝑇𝑦𝑝𝑒=='K' then

(11)    kill  𝑗 and Dispatch(FG, 𝐣 ).

(12)    𝐼𝑠𝑆𝑢𝑐𝑐=𝑡𝑟𝑢𝑒;

(13)  else

(14)   Dispatch(FG, 𝐣 ).

(15)   𝐼𝑠𝑆𝑢𝑐𝑐=𝑡𝑟𝑢𝑒;

(16)  return IsSucc;

(17) function Dispatch (𝑓𝑙𝑎𝑔,𝑗)

(18) sort processes of 𝑗 in descending order of their processor utilization;

(19) if  𝑓𝑙𝑎𝑔==𝐹𝐺  then

(20)  𝑝← sorted idle processors in ascending order of their utilization;

(21)  for each process  𝑗_𝑖  in the sorted list do

(22)   place 𝑗_𝑖 to 𝑝_𝑖 and run 𝑗_𝑖 in the foreground VM;

(23) else

(24)  𝑝← sorted idle processors in ascending order of their utilization;

(25)   for each process 𝑗_𝑖  do

(26)   place 𝑗_𝑖 to 𝑝_𝑖 and run 𝑗_𝑖 in the background VM;