Data Migration from EMC Clariion to HP EVA in different data center

[moonchild]

We have a data center in Atlanta that we are decommissioning but we need to migrate all the data to a new data center in NY.

Data center in ATL has Clarrion and DMX storage and Brocade silkworm swicthes. The new data center will have 2 x HP EVA 6000.
Total data that needs to be migrated is around 60TB.

Both data centers are connected thru DS3 (45Mbps)

all 148 hosts are windows based and couple VMs.

any suggestions on how to do it with minimum downtime?

thank you inadvance

[Amar_Joshi · Accepted solution]

I did a simple math on DS3 transfer and seems like you won't be able to transfer more than 486GBytes of data in a day which means for 60TB you will need more than 120 days. So this clearly wipes out any of the replication solution from the table.

If you can pick some business critical servers and want to replicate them over the DS3 I will suggest to use Softek TDMF software which has both Host based and IP based replication (mirroring) features. It's comparatively cheap solution to any of the Continuous Access replications and it's been in field for so long that you can trust also.

My 2 cents worth.

[IBaltay]

Hi,

Here are the possible pilot scenarios:

1. Tape backup based - not preffered

a)full tape backup of one host EMC disks
b)host zoning of one test host in a new locaation with EVA
d)EVA new LUNs creation/presentation to the test host
e)restore from the tapes to EVA vdisk

check if it successfull and measure the efforts and time

2. Metrocluster - preffered (or some other EMC/HP replication product for migration only) (e.g. EMC SRDF), with doing the initial copies from Atlanta to NY per partes not to oversaturate the bandwidth at a time. Finaly all the LUNs (after the series of the initial full copy replication sessions will be journal assynchronously/synchronously replicated. The increments are not as bandwidth demanding as the full copy at the beginning.
If this is done, then you can plan the physical weekend action:
1. stopping the IO to the LUNs at Atlanta
(shutdown the servers
2. stopping the replication
3. move the host to NY
4. zone them in a new fabric with EVA
5. if the EVA LUNs are allready presented
do all application checks

3. LVM (e.g. Veritas) for Windows solution
after the EVA LUNs (VDISKs) configuration
add the EVA disks into the LVM, mirror them
(the initial mirroring can be here delayed to the series of sessions, not to be done in one time)
and before the move reduce EMC LVM logical volumes

[Steven Clementi]

4. Ship the EVA's to Atlanta, connect to the switches there, create luns and present, copy data...test... ship back to NYC

I know.. probably not too feasible... but it is an option.


If you have the availability to connect the 2 SAN's (if not already connected), this probably would become an easier situation.

5. Obtain 2 IP Distance Gateways, configure one in each site and merge 1 fabric. Present storage, copy data... etc.


Steven

[Steven Clementi]

Are you moving hosts too? or just storage?

[moonchild]

Steven,

The hosts will be migrated to a virtual infrastructure (VMWare).

[Cass Witkowski]

Is this data in a database?

If so then:
1) Perform a backup to tape
2) Ship tape to new site (As Hof would say almost infinite bandwidth)
3) Restore tape to disk
4) In the meantime send journal files from old site over net to new site
5) Apply journals at new site to bring new database close to being in sync
6) Shutdown old site
7) Move last journal file and apply to new site
8) Come up on new site.