دانلود رایگان پاورپوینت شناسایی ماشین مجازی غیر فعال در FermiCloud (انگلیسی)

Idle virtual machine detection in FermiCloud
FermiCloud & FermiGrid
Problem Statement
Project overview
Identify an Idle VM
Computing the Indexes
Index in Action: CPU Idle Prime Number Test (T = 5s)
Index in Action: Network Activity Ping Test (T = 10s)
Indexes & Idleness Rules
24 Hours VM Test: CPU Idle
24 Hours VM Test: Keyboard / pty Idle Time
24 Hours VM Test iowait
24 Hours VM Test Memory Paging in/out
Idleness rule
Idleness rule
Final test results
Conclusions

Final test results

During the final test, the detector identified idle VMs with 90% accuracy. This is only a first encouraging result. Deeper analysis and improvements of the idle detector are needed. In particular: The final test results come from the anaysis of a very little pool of VMs (30). The majority of them were always idle, only a few of them were actually used during the test. Idle thresholds must be improved, processing data coming from VMs providing different services (servers, worker nodes, etc.). New ways to use the recorded indexes values may be investigated (use of different averages, recording of the last 60 minute values and processing of these data, etc.).