SSMC 3.7 Service Time Calculation not correct

[Rapadila1]

Dear All,

Just want to check if you are facing the same issue with me.
I'm using SSMC 3.7.
when I to open the VV performance chart seems like the service time total, not properly calculated from read and write. from screenshot below the formula should be read + write = total service time. but the total seems wrong.

 Kindly need your advice on this.
Thanks in advance

[sbhat09]

Hi Rapadila,

I will review this and get back to you.

Regards,

Srinivas Bhat

[sbhat09]

Hi Rapadila,

Can you please share below command outputs to understand the VV details:

showvv (vv_name)

showvv -d (vv_name)

showvv -cpgalloc (vv_name)

Regards,

Srinivas Bhat

[Rapadila1]

Hi Shrinivas, 

thank you for looking on to this issue. 

Below are the result. I think its the SSMC shown incorrectly, It is not only happen on 1 VV, but it is happen on all VV.
I have 3 array connected to the same SSMC and all the result seems not right.

3PARXXXXX cli% showvv UDC1DEV01_3PAR02_07
-Rsvd(MiB)- -(MiB)-
Id Name Prov Compr Dedup Type CopyOf BsId Rd -Detailed_State- Snp Usr VSize
108 UDC1DEV01_3PAR02_07 tpvv No No base --- 108 RW normal 0 1381376 4194304
-------------------------------------------------------------------------------------------------
1 total 0 1381376 4194304
3PARXXXXX cli% showvv -d UDC1DEV01_3PAR02_07
Id Name Rd Mstr Prnt Roch Rwch PPrnt SPrnt PBlkRemain -------------VV_WWN------------- -----CreationTime------ Udid
108 UDC1DEV01_3PAR02_07 RW 2/0/1 --- --- --- --- -- -- 60002AC0000000000000006C00020518 2018-04-05 21:24:20 WIB 108
------------------------------------------------------------------------------------------------------------------------------------
1 total 0
3PARXXXXX cli% showvv -cpgalloc UDC1DEV01_3PAR02_07
Id Name Prov Compr Dedup Type UsrCPG SnpCPG
108 UDC1DEV01_3PAR02_07 tpvv No No base AO2_CPG_SAS10K1.8TB_R6 --
---------------------------------------------------------------------------
1 total

StatVV result
07:56:37 01/06/2021 r/w I/O per second KBytes per sec Svt ms IOSz KB
VVname Cur Avg Max Cur Avg Max Cur Avg Cur Avg Qlen
UDC1DEV01_3PAR02_07 t 12 18 34 51 479 1324 0.786 0.532 4.2 27.0 0
-------------------------------------------------------------------------------
1 t 12 18 51 479 0.786 0.532 4.2 27.0 0

[sbhat09]

Hi Rapadila,

Thanks for the details. I will review this in lab and get back to you.

Regards,

Srinivas Bhat

[sbhat09]

Hi Rapadila,

Sorry for the delayed response on this. I tried to dig down a lot about this. I referred several HPE documents, our lab systems, some of the previously downloaded volume performance reports and also few demonstration videos. Everywhere the 'Total service time' value is always in between the 'read service time' and 'write service time'. But there is no mathematical calculation available on how Total service time is derived with reference to read and write service times. Surely total service time is not the total of read and write like how it is in IOPs and bandwidth reports. Total service time doesn't look like mathematically related to read and write service times.

At every instance, total service time relatively differs from read and write service times. All three look like independent values. What has been constant is, the total service time value is always between read and write service time values.

Regards,

Srinivas Bhat 

[apol]

I think that it IS mathematically related to read and write service times, but it would not be mathematically correct to just add write and read servicetime. You have to take into account the numer of read and write iops as well - e.g.one write io at 100ms and 500 read ios at 2ms does NOt result in 102 or 51ms total service time. Total Service time is probably not the plain arithmetic average, but the weighted arithmetic average. Something like

((1 x 100) + (500 x 2)) / (1 + 500)

I guess - never calculated it on my own...