8200 4ports, 2 for RC and 2 for Data

[Nikolay_K]

Hi Team,

as FC cards bit expensive, do you think just embedded ports would be enough to link two all flash 8200 - 1 pair for RC, another for Data? ISL 8 gbit, FC fabric 8 gbit but planned to be upgraded to 16 in several months. Planned load - vmware stretched cluster. Current replication bandwidth ~ 100MB\sec but SSD so i cannot use RCIP - due to delays and need to use peer persistance and sync copy.

of cource if one FC switch\fabric fails - we will have array connected by only 1 port to vmware hosts. So much bad?

[MammaGutt]

You will lose persistent port
If you lose one node, you lose one fabric.
If you lose one fabric, you'll lose the performance of one node.
If you lose a single SFP or get into issue with a single port on FC switch or 3PAR it will impact an entire fabric. It will limit troubleshooting as you have multiple components as your only path in a fabric.

The FC HBA is an expensive 4port FC HBA, but in your entire storage environment it is a very small part of the total cost. Not really where I would start if I were to save money.

An to be honest, you are probably better off with a single fabric with 2 connections, but it isn't a good option either.

[kolin]

Thank you, looks like no way )

Rcip not an option?

[Richard Siemers]

It depends...

RCIP is supported, but there are limitations on latency between the arrays... I am concerned that you only have a 100mbit link, so alot depends on your workload... Its counter-productive to have nice all-flash arrays with sync-replication over a slow line. Your disk IO latency will be impacted by your 100mbit link latency. If it's lab/test gear, try it. For real business use, plan some upgrades or shift to using async replication for a DR strategy.