Query on the Load Balancing ?

[Former Member]

Guys

I would like to understand how the FAV_COMPUTE_TIME has been calculated for the system to indentify the FAV_COMPUTE_SERVER

I have enclosed a snapshot of our prd server.Appreciate advise on this subject

Integer-Storage MSGSERVER

REL_SYSTEM 4,640 8,464 0 0 0 0 0 0 0 0

REL_PATCHNO_00000144 2,381 8,465 0 0 0 0 0 0 0 0

sapdp16_P16_16 58,643 7,353 122 5 46 520 14 0 0 21

sapap16b_P16_16 58,867 7,353 539 179 214 201 16 0 *73741824- 56

sapap16a_P16_16 58,736 7,353 889 215 184 20 16 0 *73741824- 63

sapap16c_P16_16 58,804 7,353 650 192 181 156 16 0 *73741824- 49

sapap16d_P16_16 58,758 7,353 664 145 173 150 16 0 *73741824- 94

Short-Text-Storage MSGSERVER

FAV_COMPUTE_TIME FAV_COMPUTE_TIME 58867

FAV_COMPUTE_SERVER sapap16b_P16_16 16:21:07

Logon Favorite Storage MSGSERVER

P16 3,216 46C (sapap16b_P16_16)

P16-TPA 3,216 46C (sapap16b_P16_16)

SPACE 3,216 46C (sapdp16_P16_16)

USED-BY-BW 3,216 46C (sapap16b_P16_16)

USED-BY-EBPRO 3,216 46C (sapap16b_P16_16)

Long-Text-Storage MSGSERVER

Regards

Karthi

[Former Member · Accepted solution]

Hello Karthi,

I would like to understand how the FAV_COMPUTE_TIME has been calculated for the system to indentify the FAV_COMPUTE_SERVER
Short-Text-Storage          MSGSERVER
FAV_COMPUTE_TIME     FAV_COMPUTE_TIME    58867
FAV_COMPUTE_SERVER   sapap16b_P16_16     16:21:07

the value for FAV_COMPUTE_TIME is very easy to calculate: simply count the seconds since midnight, i.e. the value 58867 only represents the time stamp 16:21:07 (16 hours = 57600 seconds, 21 minutes = 1260 seconds. 5760012607=58867).

Very easy, isn't it?

In other words: the information FAV_COMPUTE_TIME and FAV_COMPUTE_SERVER is only a remark at what time an instance has forwarded its load information to the message server. It has nothing to do with the quality value itself. In your example the instance sapap16b_P16_16 has calculated and forwarded its load information (resp.time 539ms, 214 users, quality value 201) at 16:21:07. But this does not mean that this instance is automatically the favourite instance (ok, in your example it is indeed the favourite instance for all logon groups, but that's only by chance. Some minutes later most probably sapap16b_P16_16 is still favourite server for the logon groups but for FAV_COMPUTE_SERVER another instance is listed).

Personally I think it is sufficient to have a look at the load distribution screen in SMLG (Goto - Load distribution (or simply F5)) than to try to guess araound what exactly the figures in the message server status area mean.

But if I understand you correctly you are more interested how this mysterious quality value is calculated, right?

It is calculated in report RSRZLLG0 (or more detailed in RSRZLLG0_ACTUAL) which is started on each instance each five minutes (by sapmssy6) and each fifth logon.

Believe a simple maths is involved

Not, not really simple. The report uses as well response time and the number of logged on users. But it uses also the difference between the current values and the thresholds (if they are used), i.e. if the number of users is very close to the (defined) threshold the quality will be lower than compared to an instance with less users (and same threshold).

Also the report calculates at first each part seperately (users, response time) and uses then an additional weighting so that the response time has more influence than the number of users (see also the already mentioned [note 51789|http://service.sap.com/sap/support/notes/51789]).

Maybe describing the principle as it happened know contains enough information for you, but if you really want to understand the coding in every detail, you should ask an ABAPer to explain it to you

Regards, Michael

[anindya_bose]

You can have a look at

51789 - Poor user distribution in logon distribution --Section 3

26317 - Set up for LOGON group for autom. load balancing

For informtion regarding RFC Load Balancing you might find note 593058 useful

Best Regards

Anindya

Edited by: Anindya Bose on Nov 19, 2009 8:45 AM

[former_member66264]

This really is an expert question ...

First of all you should maybe explain to your collegeus you get this info from SMLG.( goto , system diag, msg srv status)

This data is comming from Report RSALTST4

and will import the data from shared memory

have a look at Func Grp SAL2 , and Package SRZL

if you speak or can read german:

Die Favoritenliste wird zyklisch neu berechnet und im Message-Server

• gespeichert. Die Berechnung findet statt im Hintergrundabap,

• der auf allen Appl.Servern alle 5 Minuten gestartet wird.

• Grundlagen der Berechnung sind Konfigurationsdaten aus der Datenbank

• (siehe unten) und Zustandsdaten aus den Appl.Servern.

• Zustandsdaten eines Appl.Servers betreffen:

• - mittlere Dialog-Antwortzeit in ms

• - Anzahl Transaktionsschritte pro Minute

• - Anzahl angemeldeter Benutzer

Very fast translation:

data builded in message server, run by a btc background abap ( every 5 minutes)

its based on diffent statusses:

response time

trxn per minute

total logged on users

Its using ADM messages ( check some notes on this topic)

I hope this response is usefull ! Llet me know if you want to know more!

[Former Member]

Nico,

Thanks for the information. I am not an abaper .

Believe a simple maths is involved in identifying the fav server.Intrested to know the theory behind it

I understand its based on the weight..again here we have no of users, avg response time and dialog steps ...

Regards

Karthi