Distance is not the problem, it's your RPO and RTO. I've seen distances up to 200 miles with a 6ms latency where synchronous replication wasn't a problem at all.

Recover Point Objective = the point to where you can go back after a disaster
Recover Time Objective = the amount of time it takes to go back to the point in time where everything was consistent

If latency on the line is low (which is probably the case, since it's only 15 miles), MV/S shouldn't be a problem at all. However: if your application has peaks in it's I/O pattern and the 1Gb connection cannot handle the MB's, you might consider going to MV/S simply because of bandwidth issues and therefore wait times for each I/O if the primary CX is waiting for the ACK from the secundary CX.

Ooops, yes ofcourse. Switching from MV/S to MV/A means the host has no idea of what's going on in Mirrorview land and bandwidth problems are now no problem anymore.

Rob ..did you mean MV/A ?

Please see the following Whitepaper for MirrorView:

http://powerlink.emc.com/km/live1/en_US/Offering_Technical/White_Paper/H2251_clariion_mirrorview_asynch_dr_wp_ldv.pdf

regards,

glen

So what is the requirement for your hosts ? Does all data have to be on the other side at all times or can you handle a slight delay in your orginisation / applications ?
Remember that you will suffer if using MV/S, but this may vary depending on the latency of the line.

Are you happy with the answers ?

Message was edited by:
RRR

Sorry, for the late reply.

Applications running on the servers attached to CX3-10c are mostly core banking and thus RPO and RTO needs to be zero at all times.

Would using iSCSI come as bottleneck for bandwidth and thus have latency?

regards,
Samir

My guts is telling me iSCSI is far slower than FC, but it will work. Latency is higher, so for each I/O the host will need to wait a bit longer. But it will work !!

you can use iScsi but in a configuration like this where the customer is sensitive to the performance you need to have at least a dedicated link between the two sites and on this link only the iscsi traffic need to be routed .

Also keep in mind there are funtional requirements for latency , since Mirroview /S will fracture the secondary image if it has an i/o outstanding to the secondary image for 10 seconds.

Performance of FC and ISCSI mirrorview are very similar when the latency is very low .
For more info regarding Mirrorview technology over FC and iSCSI please refer to the following white paper :

http://powerlink.emc.com/km/live1/en_US/Offering_Technical/White_Paper/H2417_clariion_mirrorview_asynch_dr_wp_ldv.pdf?mtcs=ZXZlbnRUeXBlPUttQ2xpY2tTZWFyY2hSZXN1bHRzRXZlbnQsZG9jdW1lbnRJZD0wOTAxNDA2NjgwMzcwOWNjLGRhdGFTb3VyY2U9RENUTV9lbl9VU18w
Thanks

Mirrorview/S will work over iSCSI, but response times due to latency over the link will probably result in high response times on the source LUN.

glen

If you truly cannot tolerate losing even a second of work, then synchronous is your only option. I have been running Mirrorview/S for almost 4 years between two CX700s which I then upgraded to a CX3-80 and CX3-40f. I have a dedicated 300Mbps circuit that plugs directly into the MDS and I run FCIP across it. I didn't want there to be any chance of the bandwidth contention with it and other applications. Driving distance between sites is about 30 miles, but my latency is very acceptable at around ~2ms. I don't have any experience with Mirrorview over iSCSI, but if you decide to go with FCIP I can give you some ideas of other things to consider.