Real-World FCoE Best Practices Q&A

At our recent live Webcast “Real-World FCoE Designs and Best Practices,” IT leaders from Thermo Fisher Scientific and Gannett Co. shared their experiences from their FCoE deployments – one single-hop, one multi-hop. It was a candid discussion on the lessons they learned. If you missed the Webcast, it’s now available on demand. We polled the audience to see what stage of FCoE deployment they’re in (see the poll results at the end of this blog). Just over half said they’re still in learning mode. To that end, here are answers to the questions we got during the Webcast. As you will see, many of these questions were directed to our guest end users regarding their experiences. I hope that it will help you in your journey. If you have additional questions, please ask them in the comments section in this blog and we’ll get back to you as soon as possible.

Q. Have any issues come up where the storage team needed to upgrade SAN switch firmware to solve a problem, but the network team objected to upgrading the FCFs?  This assumes a shared firmware release on both network and SAN switch products (i.e. Cisco NX-OS)

A. No we need to work together as a team so as long as it is planned out in advance this has not been an issue.

Q. Is there any overhead at the host CPU level when using FCOE/CNA vs. using FC/HBA? Has anyone done any benchmarking on this?

A. To the host OS it is the CNA that presents a HBA and 10G Ethernet adapter, so to the host OS there is not a difference from what is normally presented for Ethernet and FC adapters.  In a software FCoE implementation there might be, but you should check with the particular implementation from the OS vendors for this information.

Q. Are there any high-level performance considerations when compared to typical FC SAN? Any obvious impact to IO latency as hosts are moved to FCoE compared to FC?

A. There is a performance increase in comparison to 8GB Fibre channel since FCoE using Ethernet and 64/66b encoding vs. 8/10b encoding that native 8GB uses. On dedicated links it could be around 50% increase in performance from 10GB FCoE vs. 8GB FC.

 Q. Have you planned to use of 40G – FCoE in you edge core design?

A. We have purchased the hardware to go 40G if we choose to.

 Q. Was DCB used to isolate the network traffic with FC traffic at the CNA?

A. DCB is a set of technologies that includes DCBX, PFC, ETS that are used with FCoE.

 Q. Was FCoE implemented on existing hosts or just on new ones being added to the SAN?

A. Only on new hosts.

Q. Can you expand on Domain_ID sprawl ?

A. In FC or FCoE fabrics each storage vendor supports only a certain amount of switches per fabric. Each full FC or FCoE switch will consume a Domain ID, so it is important to consider how many switches or domain IDs are allowed in a supported fabric based on the storage vendor’s fabric recommendations. There are technologies such as NPIV and vendor specific technologies that can be helpful to limit domain ID sprawl in your fabrics.

Remember the poll I mentioned during the Webcast? Here are the results. Let us know where you are in your FCoE deployment plans.

Screen Shot 2014-12-19 at 9.15.15 AM

 

 

Leave a Reply

Your email address will not be published. Required fields are marked *