Basic knowledge on
Slots
ü
Each node contains -> 24 CPUS = 12 Cores(Physical)
ü
Total nodes (Masters & Slaves) = 108
ü
Number of Nodes Excluding Masters (CLDB & Jt) = 100
ü Total Cores=100
nodes *12 Cores = 1200 Cores
ü
Total number of Cores after enabling Hyper
Threading = 1200*2 = 2400 Cores
ü
Each Node will take 2 cores one for OS and 2nd
one for File System,Warden and TT à
100 nodes will consume 200 cores
ü
Total Number of remaining Cores = 2400 -200 =
2200 Cores
ü
Roughly Each Core will handle 1.5 slots.
ü
2200 Cores * 1.5 = 3300 Slots ( Map Task &
Reduce Tasks)
Example 2:
Total nodes = 13 data nodes
Each data node = 22 Physical ( 44 CPUs, after hyper thread enabled)
Total CPUs = 13*44 = 572 cores ( 1 CPU = 1 Core)
Each data node will consume 2 cores one for OS and another one for Hadoop Service.
13*2 = 26
Remaining CPUs = 572-26=546
Total number of slots = 546*1.5 ( Roughly Each Core will handle 1.5 slots)
Total slots = 819 slots for Mappers and Reducers
Disk Size on each node:
Master Node:
Memory : 94.4 GB
Disk Space =883 GB/3 disks
CPU =32
Network I/O =34.6 KB/Sec
Disk I/O = 144KB/Sec
Slave Node:
Memory =126 GB
Disk Space = 30.9 TB /12 disks (2.5 TB/disk)
CPU=24
Chuck Size:
Example 2:
Total nodes = 13 data nodes
Each data node = 22 Physical ( 44 CPUs, after hyper thread enabled)
Total CPUs = 13*44 = 572 cores ( 1 CPU = 1 Core)
Each data node will consume 2 cores one for OS and another one for Hadoop Service.
13*2 = 26
Remaining CPUs = 572-26=546
Total number of slots = 546*1.5 ( Roughly Each Core will handle 1.5 slots)
Total slots = 819 slots for Mappers and Reducers
Disk Size on each node:
Master Node:
Memory : 94.4 GB
Disk Space =883 GB/3 disks
CPU =32
Network I/O =34.6 KB/Sec
Disk I/O = 144KB/Sec
Slave Node:
Memory =126 GB
Disk Space = 30.9 TB /12 disks (2.5 TB/disk)
CPU=24
Chuck Size:
No comments:
Post a Comment