Information Technology Reference
In-Depth Information
100000
Complete History
Granularity 20 000 ev.
Granularity 50 000 ev.
Granularity 100 000 ev.
10000
1000
100
10
1
0
2000
4000
6000
8000
10000
12000
Size of the original trace (MB)
Fig. 11.
Comparison of history database sizes for different cases
5.3 Query Time
For the different database update strategies of the previous experiment, we mea-
sure the query time, which is shown in Figure 12. The best case is using the
complete history, where all data is accessible through a direct query of the his-
tory tree. For the other cases, although the size of the database is smaller, it
first needs to fetch the states stored in the previous snapshot, and then reread
the trace to extract the current state values up to the query time.
70
60
50
40
30
Complete History
Granularity 20 000 ev.
Granularity 50 000 ev.
Granularity 100 000 ev.
20
10
0
2000
4000
6000
8000
10000
12000
Size of the orignal trace (MB)
Fig. 12.
Comparison of query times for different cases
6Con lu on
In trace analysis, it can be sucient to process the trace once, update the current
state all along, and displaying it live. In
a posteriori
analysis however, one wants
to be able to navigate through the trace back and forth and be able to look at
the system state for any point in the trace.
To solve this problem, we proposed ecient data structures and algorithms
to store and manage the system state values gathered from trace logs. In the
