ANALYSIS FOR ALL NODES

 

Alarm triggering condition: The standard deviation of the forecast has to be smaller than the residual (F-y) in 70% of the points in the last 2,5 hours.

 

Click on the name to get the excel file.

 

Niit

 

 

 

No alerts!!


UTDALLAS

 

 

 

NO ALARMS!!! PERFECT!!


UIUC

 

 

 

NO ALARMS!!!!!


NSLAB

 

 

 

NO ALARMS!!!


TRIUMF

 

 

 

2 ALARMS!!!  It’s one of them false??? Maybe not.


SWITCH

 

 

NO ALARMS

 

 

Stanford

 

 

 

NO ALARMS!!!

 

 


SOX

 

 

 

 

Total Failure!!!!!!! But it’s obvious since the first week give the tendency. Not even a very little alpha and a high beta solve this. This node is very particular. The values of dbcap finally tend to 368.3 and stay there for several weeks without any change. If we start our analysis from the 5th week we won’t get any alarm

 


INDIANA

 

 

 

ONE ALARM!!!!! PERFECT!!!

 

 

CALTECH

 

 

ONE ALARM IN THE RIGHT PLACE!!!!!

 

RIKEN

 

 

 

I am not completely sure if this should be taken as 1 or many ALARMS. Let’s say that is only 1. Because I am always assuming that if the bad condition repeats itself during consecutive days, it should be part of the same error.

 

PD

 

 

 

NO ALARMS!!!

 

 

NASA

 

 

The same thing than SOX, and it fails pretty badly.

 

MIB

 

 

 

NO ALARMS!!!!!!!!!!!!!!!!!!!!!!!!

 

 

 

MCS… This one needs a little more attention

 

 

            I am pretty sure that this is a good alarm. It seems like it is from my point of view.

 

 

Now THIS??? It is clearly a change in the normal weekly behavior, but is it really an Alarm? I think it is. Maybe not….

 

 

And this!!!! We can see 1 clear step change at the beginning, and then many alarms which are all part of a big failure… And at the end.. ok let’s call it a FALSE alarm.

 


MAN

 

 

 

NO ALARMS!!!

 

 

LSA… closer look here

 

 

 

 

We get this… I think it’s OK

 

 

 

 

I don’t know why, but we don’t get the big one… 1 MISSED

 

 

3 more alarms… the second one is OK, I think the third one too. The first one, I don’t know.

 

 

 

 

BINP

 

 

 

4 alerts…

 

And later..

 

 

We get a double alert at the end.

 Let’s look at this closer: are these good alerts?

 

If we go to the smallest one. On 23/8/2004:

 

This bad performance last more than 12 hours, so maybe it is a good alert.

 

 

CCSORNL

 

 

 

NO ALARMS!!!!

 

 

 

CERN

 

 

 

NO ALARMS!!!!

 

 

CESNET

 

 

The same problem than SOX…

 

DESY

 

 

 

NO ALARMS”!!!!!

 

 

 

DL

 

 

 

NO ALARMS HERE….

BUT LATER ON…

 

 

ONE MISSED…

 

 

RICE

 

 

 

 

NO ALARMS!!!!!!

 

 

FZK

 

 

 

2 Alarms… 1 false… the other, seems to be false too…

 

 

 

In2pr

 

 

 

NO ALARMS!!!

 

 

JPAPAN

 

 

 

NO ALARMS!!! PERFECT

 

NIKHEF

 

 

 

NO ALARMS!!! Good…

 

 

LBL

 

 

NO ALARMS!!!!

 

SDSC

 

 

We get many diurnal alerts that are impossible for HW to predict. These diurnal effects don’t happen in previous weeks.

 

Comments

 

We get almost no false alarms, except when the first weeks show really better performance in average than the weeks that follow or when the standard deviation is very small (constant dbcap many days). SO, when the data find a stable value, the standard deviation is so little, that any residual can cause an alarm.

 

Getting the real minimum MSE instead of a estimative value of this minimization could solve all these errors.

 

We have to remember that the Holt-Winters algorithm assumes that there will be a seasonal trend, and in the case of Cesnet or SOX, there is not a seasonal trend since the values of the dbcap  tend to be constant on the long term and very different compared with the first couple of weeks.

 

We generally don’t miss the important step changes.

 

Without counting the particular SOX situation we get:

 

32 alarms in 3 months on 29 nodes

15 good

6 doubtful…

3 missed

2 false…                                                                                                                           

6 Diurnal on SDSC