Data Migration

[Vic S. Kelan]

Hi All,
I am planning to migrate data on our hpux dev servers from an EMC symetric to a clariion.

Total data is 785GB, two hba's connected to the server, one to the EMC clariion via 2gb brocade switch and the other also via the switch to the symetric.

I plan to use (cd /old_directory; fbackup –i . –vf - ) | (cd /new_directory; frecover –Xrf -) for the data migration.

my question:
1)how do i calculate/estimate how long it will take to migrate the data.
2)what will be the impact if I run the above command to multiple directories at once, is this any good or do I wait for each directory to be done before starting on the next?

thanks for your ever helpful feedback!

[harry d brown jr]

You have several items that you have to get answers for before a calculation can be "close":
(1) single hba's to each device
(2) speed of retrieving data from the symmetrix (how fast can you get data off this machine?)
(3) speed of writing data to the clariion (how fast will it allow you to write data?)
(4) speed of your servers backplane (what kind of server is it, # of CPU's, amt of memory, ...?)
(5) why use fbackup/frecover ?

If you didn't have bottlenecks, then you could say that you are going to get 80% of the speed of the 2GB HBA (which is about 1.6 gigabit's per sec), 10 bits in a byte, thus 160 megabytes per sec. 785 GB / 160MB = 4906 secs ~= 1 1/2 hours. But we don't live in a perfect world, so there is Walgreen's (sorry I digressed into a commercial). I'm thinking , depending on your hardware and such, that you are looking at 2hrs to 8hrs.

live free or die
harry d brown jr

[Vic S. Kelan]

Thanks for the quick response and the ball park figure, I really appreciate that.

1) On each server we have two hbaâ s one will be connected to the clarion FC switch, the second to the symmetrix fc switch both 2gb switches
2) The only infor I have on that is that buffer to/from media 58-96MB/s, SP to?from buffer 200MB/s
3) Same as above?
4) # CPUs 4, memory, 15579136 Kbytes, OS, HPUX 11.11, rp8400 partitioned.
5) Well I have used it in the past for this type of migration, and its in a doc at ITRC KBRC00011131. I have files larger than 2gb and this will work fine. We donâ t have replication software between the two EMC disk frames to achieve it from the storage side.
Do you have any other suggestions apart from using fbackup/frecover?
thanks!

[Geoff Wild]

I did a simliar migration in the past - old symetrix to emc dmx...I fould vxdump/vxrestore faster then fbackup....

Your biggest factor will be how the clariion is configured - is it parity raid (slowwwwwwwwwwww) or raid10 (much faster)?

On the DMX'es - I was able to do over a1 TB in 8 hours.

Rgds...Geoff

[DCE]

You could put EMC and clarion disks in the same VG and mirror/unmirror them. effectively no down time.

[Vic S. Kelan]

Thanks for the feedback.
DCE, u know, I just realized the reason I wanted to stick with fbackup/frecover was because I just assumed this servers like my others had a max_PV that was already at max and would prevent using mirror/sync.

Thanks for th eye opener there! I have a max_PV of 16, in a particular VG for example, I have 15 8gb disks. I want to consolidate and create a single RAID5 PV of 120GB. So I am still under max_PV.

(-:

[Patrick Wallek]

You may be under the MAX PV limit, but you must keep in mind the MAX PE PER PV limit. Unless you changed it from the default when you created the VG, then you will most likely run into a limity there as well.

The PE SIZE times the MAX PE PER PV determines the largest disk you can use in a VG.

For example:

PE SIZE = 16
MAX PE PER PV = 5000

Largest disk = 16 x 5000 = 80,000 MB

MAX PE PER PV can not be changed without rebuilding the VG.