cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
903
Views
0
Helpful
4
Replies

Connectivity for SAN replication

bberry
Level 1
Level 1

I am a newbie in the SAN world so hope am asking the right questions.

We are looking into replacing our aging SAN. As part of the replacement, managemnt is thinking of placing a secondary SAN at one of our remote locations for backup & DR purposes.

My question revolves around providing connectivity between the two SANS. I have been asked to look into a dedicated point-to-point connection for the SAN replication. Not counting the actual SAN hardware, what other hardware would I need? Am I bacically building an independent network for this? Why this route instead of over our corporate WAN? Any good recommended reading for learning about the SAN world?

Brent

4 Replies 4

Michael Brown
Cisco Employee
Cisco Employee

For remote replication, there are 2 methods of providing SAN connectivity. If you have the need for high bandwidth, DWDM or CWDM are high bandwidth solutions. These technologies use fibre channel end to end.

You might be more interested in FCIP (fiber channel over IP). Depending on the bandwidth needed to keep within the allotted time for the replication, FCIP offers all kinds of enhancements. Compression and encryption are options for FCIP. There is also write acceleration as an option, if its compatible with your replication application.

Here is a link to some FCIP scenarios.

http://www.cisco.com/en/US/tech/tk571/tk751/tsd_technology_support_protocol_home.html

Here is a link on DWDM for SAN extension.

http://www.cisco.com/en/US/prod/collateral/optical/ps5725/ps2011/prod_white_paper0900aecd80181093.html

Hope this helps,

Mike

plowden
Cisco Employee
Cisco Employee

Hi, Brent,

I agree with Mike. I originally posted this before I saw his reply.

Here are a couple of good Cisco-specific document to start with:

http://www.cisco.com/en/US/solutions/ns340/ns517/ns224/ns378/net_design_guidance0900aecd800ed146.pdf

http://www.cisco.com/go/storage

Your options are either native FC over some sort of fiber transport (dark fiber, CWDM, DWDM, SONET) or FCIP over a WAN link. Native FC is usually for synchronous replication for distances up to 200 km for large bandwidth requirements. FCIP is usually for asynchronous replication at longer distances with less bandwidth. The transport largely dictates the hardware. FCIP is generally cheaper.

If you go the FCIP route, a dedicated link is preferable to make sure replication traffic has predictable latency and bandwidth, but you can share an existing link if you have to. Basic QOS is available in this case. Also, FCIP on Cisco devices has special capabilities to help you get the most out of your bandwidth (TCP optimizations, compression, write acceleration, etc.).

Regards,

Phil Lowden

A connectivity solution one of the SAN vendors has proposed is using MLPPP to create a dedicated point to point instead of going over our corporate WAN. I understand that this uses single individual T-1s and the MLPPP protocol bonds them together. Do you know what is needed from a hardware standpoint? Do I use standard 1-port / 2-port T-1 cards (WIC-1DSU-T1-V2) or do I need something different? Other pros / cons with this solution compared to over the corporate WAN?

Brent

Brent,

Most of the replication applications out there like PPCR and SRDF require more bandwidth than a couple of T1s. The low end for FCIP is usually about 10megabits/second. What type of corporate WAN connection do you have, and what can you dedicate to this replication effort? The bandwidth usually comes in to play when you start looking at replication job windows. IE: I need the replication to complete in 2 hours and I have x number of MegaBytes to transfer. The connection from the MDS IP cards are Gigabit Ethernet. You will need to configure traffic shaping on the MDS to limit the max bandwidth that it will send on the GE port. If you end up with MLPPP across multiple T1s, I would assume they would load balance on a per packet basis. If that is the case, any deviation in latency between the links in the bundle would lead to 'TCP out of order' packets. This might cause some additional delay with the FC traffic.

Hope this helps,

Mike

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: