how can calculate vflash for each VM

[baber]

Dear all

Hi

i have 2 question about vflash :

1-  what does means reservation and block size

2- I have SSD disk with 960G size now want to know how had to calculate the size for vflash for a vm ?

BR

Babak

[baber]

not configure vflash anyone ?

[IRIX201110141]

Your talking about vFRC right?

- You have to configure SSD/NVMe as Flash resourse for the Host

- You have to configure VM vDisk to take a potion of that resource. The Flash is managed in buckets which can have different sizes (block size). If you use a huge bucket size like 16K but your VM are only writing 4K than you leave 12k empty space in the bucket which canot use elsewhere

Regards

Joerg

[baber]

thanks

but my means is exactly vflash such as attach pic

how do i had to calculate reservation and block size ?

[IRIX201110141]

Thats vFRC.

- To find out the Block size check out if there is a Perfmon in Windows otherwise

Find out WorldgroupID

vscsiStats –l

Start the Collection

vscsiStats -s -w WorldgroupID -i 8195

Fetch the IO Lenght

vscsiStats -p ioLength -c -w WorldgroupID -i 8195

Stop the collection

vscsiStats -x -w WorldgroupID -i 8195

- If you think its to much work... choose 8Kb

- SIze of Cache...start with 10% .  There is a CMD to fetch statistic data

Regards

Joerg

[baber]

i have a few questions :

1- what does 8195 means here ?

2- i ran all of these commands with specific worldgroupID

[root@lab1:~] vscsiStats -s -w 3158159 -i 8195

[root@lab1:~] vscsiStats -p ioLength -c -w 3158159 -i 8195

[root@lab1:~] vscsiStats -x -w 3158159 -i 8195

didn't show any output

how can get suitable number for those parameters ?

Reservation

Block size

BR

Babak

[IRIX201110141]

Ah sorry.... -i needs to machts the SCSI Disk handle.  Take a look to a VM which have more than one vDisk and you'll see different hande IDs

[root@esx-node-97:~] vscsiStats -s -w 72811 -i 16087

vscsiStats: Starting Vscsi stats collection for worldGroup 72811, handleID 16087 (scsi0:0)

Success.

[root@esx-node-97:~] vscsiStats -p ioLength -c -w 72811 -i 16087

Histogram: IO lengths of Write commands,virtual machine worldGroupID,72811,virtual disk handleID,16087 (scsi0:0)

min,4096

max,102400

mean,8003

count,217

Frequency,Histogram Bucket Limit

0,512

0,1024

0,2048

0,4095

143,4096

0,8191

38,8192

22,16383

3,16384

2,32768

7,49152

0,65535

1,65536

0,81920

1,131072

0,262144

0,524288

0,524288

[root@esx-node-97:~] vscsiStats -x -w 72811 -i 16087

vscsiStats: Stopping all Vscsi stats collection for worldGroup 72811, handleID 16087 (scsi0:0)

Success.

In this case the most IOs (143) comes in 4096bytes which means 4K.

You should wait a couple of minutes before fetching results.

Regards

Joerg

[baber]

Thanks

this is output :

[root@lab1:~] vscsiStats -s -w 3158159 -i 8370

vscsiStats: Starting Vscsi stats collection for worldGroup 3158159, handleID 8370 (scsi0:0)

Success.

[root@lab1:~] vscsiStats -p ioLength -c -w 3158159 -i 8370

Histogram: IO lengths of commands,virtual machine worldGroupID,3158159,virtual disk handleID,8370 (scsi0:0)

min,512

max,270336

mean,25419

count,51

Frequency,Histogram Bucket Limit

1,512

1,1024

2,2048

1,4095

20,4096

0,8191

7,8192

3,16383

1,16384

1,32768

1,49152

11,65535

0,65536

1,81920

0,131072

0,262144

1,524288

0,524288

Histogram: IO lengths of Read commands,virtual machine worldGroupID,3158159,virtual disk handleID,8370 (scsi0:0)

min,9223372036854775807

max,-9223372036854775808

mean,0

count,0

Frequency,Histogram Bucket Limit

0,512

0,1024

0,2048

0,4095

0,4096

0,8191

0,8192

0,16383

0,16384

0,32768

0,49152

0,65535

0,65536

0,81920

0,131072

0,262144

0,524288

0,524288

Histogram: IO lengths of Write commands,virtual machine worldGroupID,3158159,virtual disk handleID,8370 (scsi0:0)

min,512

max,270336

mean,25419

count,51

Frequency,Histogram Bucket Limit

1,512

1,1024

2,2048

1,4095

20,4096

0,8191

7,8192

3,16383

1,16384

1,32768

1,49152

11,65535

0,65536

1,81920

0,131072

0,262144

1,524288

0,524288

[root@lab1:~] vscsiStats -x -w 3158159 -i 8370

vscsiStats: Stopping all Vscsi stats collection for worldGroup 3158159, handleID 8370 (scsi0:0)

Success.

now i am so confused

this out put has 3 part

Part 1

Histogram: IO lengths of commands,virtual machine worldGroupID,3158159,virtual disk handleID,8370 (scsi0:0)

min,512

max,270336

mean,25419

count,51

Frequency,Histogram Bucket Limit

1,512

1,1024

2,2048

1,4095

20,4096

0,8191

7,8192

3,16383

1,16384

1,32768

1,49152

11,65535

0,65536

1,81920

0,131072

0,262144

1,524288

0,524288

Part2

Histogram: IO lengths of Read commands,virtual machine worldGroupID,3158159,virtual disk handleID,8370 (scsi0:0)

min,9223372036854775807

max,-9223372036854775808

mean,0

count,0

Frequency,Histogram Bucket Limit

0,512

0,1024

0,2048

0,4095

0,4096

0,8191

0,8192

0,16383

0,16384

0,32768

0,49152

0,65535

0,65536

0,81920

0,131072

0,262144

0,524288

0,524288

Part3

Histogram: IO lengths of Write commands,virtual machine worldGroupID,3158159,virtual disk handleID,8370 (scsi0:0)

min,512

max,270336

mean,25419

count,51

Frequency,Histogram Bucket Limit

1,512

1,1024

2,2048

1,4095

20,4096

0,8191

7,8192

3,16383

1,16384

1,32768

1,49152

11,65535

0,65536

1,81920

0,131072

0,262144

1,524288

0,524288

[root@lab1:~] vscsiStats -x -w 3158159 -i 8370

vscsiStats: Stopping all Vscsi stats collection for worldGroup 3158159, handleID 8370 (scsi0:0)

Success.

now want to know

1- why it show 3 part ?

2- what does mean, max , mean , count in each part ?

3- what does mean Frequency,Histogram Bucket Limit ?

4- why is a different huge between amount of min , max in part 1 and part 2 ?

5- which part is a main part for get decision ?

BR

Babak

[IRIX201110141]

1. Dont know

2. Min, Max and maybe average bucket size

3. How many IOPS have which size aka are placed in which bucket. Placed the numbers into a MS Excel and draw a diagramm if this helps. [1]

4. Every Application, Process within your GuestOS use different IO sizes. But again i dont know why it returns 3 results with different values. The Part2 looks like foobar maybe because lack of IO within your VM so now real values and it returns garbage

5. From 51 IOPS the majority are counted (20) as 4K Blocksize

[1] http://myvirtualcloud.net/vscsistats-excel-macro-for-automated-charts/

Regards

Joerg

[baber]

Can anyone answer to my questions ?

BR

[baber]

who can help me for optimize vflash parameters ?

[IRIX201110141]

Calculate the avg. blocksize of the disk and youre done.

Regards

Joerg

[baber]

how can calculate ?

As i said in previous posts i want to know how can calculate require parameters for vfrc ?

[IRIX201110141]

You have to measure the blocksize your Application is used.

Regards
Joerg

[baber]

Thanks

but how do i had to do measure that ?

with specific tools ?

really i am confused till  now could not get a clear answer or a solution

[IRIX201110141]

The size of your working data set needs to fit into the cache. So specify the size vFRC

The number of buckets depends on the bucketsize and the size of the cache. The bucket size needs as near as possible to match the IOsize of the vDisk. If the bucket is to large you waste portions of your your cache. If the bucket is to small you have increase latency because more buckets are needed to store the IO.

I have show you how to measure the given IO Blocksize of your vDisk.

Why not simply add 10% of your vDisk as vFRC cache and choose 8KB just for start.  After a couple of hours fetch the vFRC stats and see the numbers.

Regards

Joerg

[baber]

thanks

but as i understand there is not any define formula or a method or a tools for calculate the vfrc parameters

for example when i have a databse system that has 2TB disk size and 32G memory now how can understand suitable parameters for that ?

[IRIX201110141]

The size of your working data set needs to fit into the cache. So specify the size vFRC

[baber]

in virtual flash cashe setting there is 2 parameter :

Reservation and block size for that system (database with 2TB disk and 32G memory) how do i had to calculate these parameters?

Do i had to use a tool ?

[IRIX201110141]

Reservation means Cache Size...  how much vFRC do you have? How much are you willing to assign as Cache to that vDisk?  I am pretty sure that only a small part of your 2TB database is active right? This small part should fit into the Cache.

AND NOW FOR THE LAST TIME...... you calculate the Blocksize by SIMPLE MEASURING the current Block size of your vDisk and the best way i already told you.

Regards

Joerg