measure block-level disk data-delta for replication bandwith calculation

[daikyu]

hi community

i hope this question isnt too stupid, or maybe i am thinking too complicated.

i want to measure the block-level disk data-delta from time A to time B for replication bandwith calculation.

how do you specialists out there, such a calculation. is my approach the correct direction ?

please share some light on this particular topic.

thanks, peter

[frankdenneman]

just to simplify it so i can understand it, you want to compare the version of data over time?

This way shows you the amount of data which is being replicated, am i correct?

Isn't just a size compare enough to estimate the replication data.

Storage array replication are block level replication, so only the data being changed is replicated.

VMFS should not be fragmented, ESX will try to allocate the whole virtual disk as one large chunk of continuous data.

If thats the case and no fragmentation, you only need to find out how much the data inside the vmdk is grown.

I don't know a cli command, but there is a vi perl toolkit script called VI3_diskfree website

maybe this wil help you

[daikyu]

hi Frank_D

yeah, thats what i meant in an aspect. i thought maybe the VMFS knows how much space is free inside each vmdk.

and also the memory is on the storage and gets synced IMO. so just size isnt enough when i think over it again. (i know i can modify the memory location, its more a theoretically question to learn ...)

for example the size maybe doest change on a DB-transaction, but the memory does.

reason: we want to know, how much data gets produced & needs to be replicated actually.

thanks, peter

[frankdenneman]

you're right, the swap file is also inside the directory.

But by playing with the memory reservation setting, you can alter the size of the swap file.

Which might come in handy while you are testing the diskgrowth. modify the memory location is a possibility, but i wouldn't want to change the location.

if something goes wrong, troubleshooting becomes a lot harder when you have to monitor two datastores, paths and replication links.

[happyhammer]

is the storage SAN based ? if so you could snapshot the LUN and take a look at the number of changes within the specified period. I know from an Equallogic point of view it will show the size of the snapshot and i guess most SAN's should have the same functionaility

[daikyu]

hi happyhammer

yeah exact. it is SAN based. SAN snapshot size should be exactly what i need. (blocklevel changes over specific time period)

the point is i am talking from a conceptually point of view. to calculate which WAN bandwith do i need to replicate the delta.

i know its hard to say in advance (depens on the users/systems,and much much more things ....) but anyway i need to find a formula/tool or best practise for this topic.

to be a little bit more concrete: does somebody here have storage replication over WAN with 2Mb Link.

thanks, peter

[happyhammer · Accepted solution]

if you PM me i have a spreadsheet that may be of help

[daikyu]

thats what i love about the community! :smileygrin:

thank you very much, sirs.