-
Notifications
You must be signed in to change notification settings - Fork 6
/
TODO
64 lines (37 loc) · 3.23 KB
/
TODO
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
TODO:
Sept 23, 2014, Add a properties pour le file
July 3, 2014, Change LIBS to lib (and fix git repos issues, Flavien's request)
July 3, 2014, Change LIBS to lib (and fix git repos issues, Flavien's request)
June 30, 2014, Relocate VMs to other nodes (randomly) when a node crash (turnOff is invoked by the injector)
For the moment, the VMs stay on the ''off node'' in order to let the injector to change their loads if need be.
Please note that the VMS (as the node) are not considered by the scheduler.
June 30, 2014 Add network stress (in/out)
Oct.28, 2012 MB_MIG is not per node but per scheduling/reconfiguration process.
It would be good to know exactly which PMs are really involved in the reconfiguration
process (for both the centralized version and the DVMS one).
This implies to analyze the reconfiguration plan from Entropy2RP
(note: in the two files: i.e.for centralized version and the DVMS one :()
June 20, 2012 There is a bug (or a feature ;)) in node.isOverloaded() from DVMSNode:
this.getCPUUsedByVMs() > DVMSProperties.getCPUOverloadThreshold()
defaultThreShold is a percentage where CPUsedByVMs is not a percentage but
like it is defined in simulation.properties (note that default threshold is
declare in dvms.properties)
June 12, 2012 Add a generic way to enable (or disabled) the traces and to specify the trace file Adrien/Flavien - Done Nov. 12, 2012
=> Resolve log (i.e.logback) issue. Actually, Entropy is leveraging sfl4j:
a wrapper for logger utilities (log4j, logback, jdj14,....).
By default, slf4j tries to bind itself with one (and ONLY one) logging utilities.
1./if there are more than one binding possibility, slf4j does not succeed to bind correctly
and no messages will be logged => let only ONE binding (i.e.either slf4j or logback for instance)
2./ Once you are sure that slf4j is bound to only one logging system, apply the corresponding
configuration process.
For logback, a logback.xml file should be placed in the classpath (in our case, placing such a
file in the correct location was not sufficient. We have to set the logback system properties
by adding -Dlogback.configurationFile=config/logback.xml as one of the VM arguments.
June 12, 2012 Add a counter per VM to know how many times the CPU has changed ?
June 12, 2012 Add a generic way to start different sizes of instances (cf . python scripts) Adrien - Done June 12, 2012
June 6, 2012 the injector should lead to underloaded situation not only normal or underloaded Adrien - Done June 30, 2014
May 25, 2012 Validate the distribution model (it seems there are too many events) Adrien - Validated by Lionel June 12, 2012
May 25, 2012 Trace Simgrid Adrien - Done June 6, 2012
May 25, 2012 Pthread limitation (see if we can know the maximum phtread number according to the OS). Adrien/Jonathan - Switch on Davinci JVM and take advantage of the coroutines Oct 25, 2012
May 4, 2012 Update current configuration after performing a reconfiguration Adrien - Done May 23, 2012
May 3, 2012 Automatically generate platform and deployment files Adrien - Done May 25, 2012, see ./generate.py at the root of the project