What an excellent question. Not an easy one to answer off-the-cuff but then again that¿s performance related questions for you¿

To get the ball rolling I have collected some of my thoughts below.

Definitely during the initial (full) sync the primary array will be the one more affected. The transfer map has marked all bits as modified and in need of transfer across to the secondary array. For every extent marked in the transfer bit map and is yet to be replicated the first write to the source LUN extent location will cause a 64KB Copy-On-First-Write (COFW) operation to take place.
At the same token during the initial sync the secondary array will not have a gold copy snapshot of the destination LUN so no COFW penalty should occur there until periodic/delta updates.

With size of the data you are talking about usually the secondary array is brought to the same Data Center as the primary array so that initial replication can occur at high speed before been shipped out to the DR site.

If you can¿t do that then you might have to just sit it out¿ Hopefully you got access to Navisphere Analyzer as it will be of utmost importance to keep an eye on the performance! Some general performance recommendations for the CLARiiON Storage Arrays prior to participation in replication are:

1. Disk utilization < 50% utilized
2. SP utilization < 50% utilized
3. SP utilization balanced between the two SP A and SP B
4. Write cache utilization
¿ Dirty Pages % < High level water mark
¿ Forced flushes on the LUNs max in single figures

As the initial synch can take a very long time sizing the Reserved-LUN-Pool (RLP) for performance (IOPS) as well as capacity (GBs) will be imperative!

There are calculations that can be made to estimate the amount of RLPs both in GB and number of spindles. Having said that they are rather complex and require in-depth-knowledge of not only MV/A but also Incremental SAN Copy (ISC), SnapView as well your own data profile.

To set you up on the right path I would highly recommend you to have a read through at least the following two documents: ¿MirrorView Knowledgebook Flare26¿ as well as ¿CLARiiON Reserved LUN Pool Configuration Considerations¿.

If you still are unsure go with small-ish RLPs (remember they are shared by ALL LUNs independent of size) in dedicated RAID-5 4+1 RG(s) with 25 LUNs and monitor your performance and RLP usage.

I haven¿t heard of a maximum number of MV/A synchronisations you can have on the go at any given time but what you should bare in mind are the array maximums as well keep on eye on the array performance and bandwidth utilization. Some of the restrictions you might encounter on the CLARiiON CX3-80 (post FLARE 24) are:
¿ 8 snaps per LUN
¿ 8 sessions per LUN
¿ 512 RLPs
¿ 16 consistency groups with a max of 16 LUNs
¿ 100 MV/A primary and secondary images

I hope this helps and keep us posted
Carl