Array Performance and Data Protection
1748195 Members
2634 Online
108759 Solutions
New Discussion юеВ

HPE Nimble performance info GUI vs CLI

 
Alex_Garcia
Occasional Collector

HPE Nimble performance info GUI vs CLI

Hi all

Maybe this is a very basic question but anyways...

We're working on creating some scripts to gather information from several different storage arrays (different models and brands) in order to present a general report for our customer. Customer should be able to match the numbers in the report with the numbers in the different GUIs from each array

 

The information needed is Used space, IOPs, MiB/s

Now, we have some HPE Nimble devices in the environment and they need to be added to the report

We were able to get the used space easily and somehow we already matched IOPs from CLI and GUI using 

stats --diff --show_time

and doing ((Reads+Writes)/10000000)*2.775), actually we're not sure why 2.775, it would be great if somebody could help us to figure it out. Using that formula, we get IOPs numbers really close to reported by GUI

Now, we can't match numbers for MiB/s

We're adding ssdRd+hddRd, and also ssdWr+hddWr, but those values doesn't match GUI Read and Write as you can see in the outputs below

 

stats --diff --show_time --interval 3600 --csv --array XXXXXXXXXX --mb

Original output (CSV)
yyyy-mm-dd,hh:mm:ss,reads,writes,seqRead%,nsMemHit%,nsSsdHit%,netRxMiB,netTxMiB,fcRxMiB,fcTxMiB,ssdRdMiB,ssdWrMiB,hddRdMiB,hddWrMiB
2020-10-07_01:00:00,4146354,2174685,79,6,51,3.89,76,136738,1465189,432116,5451,632560,79537
2020-10-07_02:00:00,8918209,3494764,80,5,63,2.21,3.85,122240,2642445,1788746,8459,373858,34856
2020-10-07_03:00:00,19439492,2550578,70,6,77,2.30,6,122719,4994741,3642579,6987,458407,45164
2020-10-07_04:00:00,21082970,3558193,69,4,79,2.13,3.02,192654,4886533,3528522,27309,526256,79836
2020-10-07_05:00:00,12613098,5837939,50,4,63,2.24,4,312316,2508604,1183863,77874,883688,173509
2020-10-07_06:00:00,16381406,4718974,29,2,89,2.20,4,274059,1996627,1019060,49243,640759,146059
2020-10-07_07:00:00,8496814,4144164,44,5,80,2.27,5,238898,1102107,518502,29933,524664,111850
2020-10-07_08:00:00,5193650,2384727,40,8,68,2.26,4,161199,794088,408313,20973,429480,77535

and these are the MiB/s values from GUI (same date 2020-10-07)

 

R-GUI,   W-GUI,  R+W GUI, Hour
406.71,  37.97,  444.68,  0:00
734.56,  33.83,  768.39,  1:00
1386.96, 33.89,  1420.85, 2:00
1357.39, 53.31,  1410.7,  3:00
696.71,  86.6,   783.31,  4:00
554.49,  75.97,  630.46,  5:00
306.02,  66.23,  372.25,  6:00
220.6,   44.71,  265.31,  7:00
496.3,   37.99,  534.29,  8:00
271.05,  65.26,  336.31,  9:00

Any help will be appreciated

Regards

Alex

 

2 REPLIES 2
Candhade
HPE Pro

Re: HPE Nimble performance info GUI vs CLI

Could you please help us understand from which HPE Nimble Document you got the formula: ((Reads+Writes)/10000000)*2.775)? 


I work for HPE

Accept or Kudo

Alex_Garcia
Occasional Collector

Re: HPE Nimble performance info GUI vs CLI

That's part of the problem.. I don't got it from any document, just with "trial and failure"... Actually it would be nice to know why that number..

Here is part of the code I used (powershell) to get values as close as possible to the ones shown in GUI

stats command to get the information:

$Stats_command_for_performance = ("stats --diff --show_time --interval 14400 --csv --hdr 10000 --from " + $StartDate + " --to " + $FinishDate + " --folder composite --pool default")

Code to "force" those numbers to "match" the ones in GUI

	$IOPS_average_read = $($OBJ_reads.Average * 2.776)
	$IOPS_average_write = $($OBJ_writes.Average * 2.776)
	$THROUGHPUT_average_read = $($OBJ_rdKiB.Average /(13.994*1024*1024) )
	$THROUGHPUT_average_write = $($OBJ_wrKiB.Average /(13.994*1024*1024) )

	$IOPS_total = $(($IOPS_average_read+$IOPS_average_write)/40000000)
	$MIBPS_total = $(($THROUGHPUT_average_read+$THROUGHPUT_average_write))
	
	$Used_Capacity_TB_as_GUI = $([Math]::Round($Used_Capacity_TB,2))
	$IOPS_total_as_GUI = $([Math]::Round($IOPS_total,1))
	$MIBPS_total_as_GUI = $([Math]::Round($MIBPS_total,1))

 

I know this is not "elegant", since I'm forcing the numbers without knowing why or how, but so far values are close to the ones shown in GUI, so we can present them to management 

Result values from script:

Date, ArrayName,UsedSpace,AVG IOPS,AVG THROUGHPUT
10/20/2020,ArrayName1,37.3,3.1,476
10/20/2020,ArrayName2,136.8,4.4,203.9
10/20/2020,ArrayName3,138.6,1.5,0
10/20/2020,ArrayName4,41.1,0.3,5.2

Values from GUI (Infosight)

image_2020-10-20_083517.png

 

 

 

 

As I mention is not 100% accurate, I suspect on the exact periods of time GUI takes (it says averages from last week but doesn't specify the exact date-time). Anyways, those are the numbers our management wants to see in our General report (with values from other brands)

BTW, the third array has a very old firmware version (3.7), so the AVG Throughput doesn't work, anyways, we're just about to upgrade its firmware, so it should be fine after that

Any help to understan those numbers (2.776 and 13.994), or why those formulas seem to work would be appreciated

Regards