file level or block level

Question

Hi all
I am doing a intial fs_copy for 30 Gb file system which has 10 Gb of data on it.
So i just took a checkpoint of the filesystem and did the intial fs_copy .
It took 10 mins to complete the intial fs_copy .So, here the intial fs_copy depends on the amount of data in the 30 Gb file system which is 10 Gb not on the overall file system size which is 30 Gb . Is that right? .
Thanks,

Rainer_EMC · Accepted Answer

no, that's it - unless you can do general network optimization or fix like jumbo frames, flow control, duplex

you can run multiple fs_copy in parallel (for different file systems)

or you can use tape silvering to transfer the baseline using an NDMP backup

dynamox · Answer

fs_copy is block level copy so it will copy every single block regardless if there is data or not.

Peter_EMC · Answer

fs_copy is block level copying, but it is smart enough to skip empty blocks

dynamox · Answer

that's interesting, how does it know ? Isn't not smart to know if blocks changed right, otherwise we would not have to use snapshots to perform incremental copies ?

Rainer_EMC · Answer

I believe the way it works is that it uses the Smartsnap feature of checkpoints

So it wont copy block that have never been used (written to) before

Its not the same as not copying "empty" blocks or unused blocks - but for a new file system it works out similar so that only the metadata blocks get copied.

vimad · Answer

Thanks for the info. So, the throughput for the intial fs_copy depends on how much data is there in the file system. I performed couple of tests like

30 gb file system with 2 gb data took 2mins for the intial fs_copy
30 gb file system with 4 gb data took 4 mins.
30 gb file system with 11 gb data took 10 mins for the intial fs_copy.

So, here the throuhput is almost the same. Is there anyway we can increase the throughput for the intial fs_copy. I modified the RCP tcpwindowsize parameter on the data mover which increased the throughput or there any other parameters that we can tune on celerra for better throughput.

Thanks,

Celerra

file level or block level

Was this post helpful?