Case 1: In this case, due to a networking issue, a
slave task (CopyDatabase) tries several times to connect
to a database, which makes a response to the master
(SchedulerServer) and is largely delayed. From the log
sequence of the master, our algorithm finds that the
transition time from state #424 to state #428 is much
larger than expected (see Table 6). According to the
learned model, the average time interval is 12.32s,
while the time interval in this case is 42.53s. Therefore,
our algorithm detects it as an anomaly of transition time
low performance.
Table 6. Case 1: Low performance transition of SILK
Time Stamp
State ID
State Meaning
2008-09-09
18:44:52.749
424
Job task is started.
2008-09-09
18:45:35.280
428
A worker progress event is
received.
Dostları ilə paylaş: |