Today the Imagine Virtually Anything Alliance (IVA) of Cisco, NetApp, and VMware announced the industry’s first certified end-to-end Fibre Channel over Ethernet (FCoE) solution for VMware cloud deployments and virtual infrastructures.
This certification is a significant milestone in the evolution and maturation of FCoE and converged 10Gb Ethernet networks. The adoption of this architecture reduces the total number of storage networking devices and the network cabling required within a data center drivingmeasurablecost savings in both CapEx and OpEx.
Cisco, NetApp, and VMware are currently the ONLY networking, storage, and virtualization vendors providing a joint certified end-to-end FCoE solution that includes the CNA (FCoE initiator), network (Cisco Nexus 5000 series) and storage target (NetApp FAS Unified Arrays) with VMware (vSphere).
The validation of the End-To-End FCoE design was completed as a part of the VMware Hardware Certification Program and is currently listed on the VMware Compatibility Guide.
Support for deployments build on architecture is provided jointly by Cisco, NetApp, and VMware.
The Flexibility In Today’s Announcement
We have so many 10GbE & FCoE deployments with Cisco & VMware it was hard for me to fathom that the VMware certification was just recently completed. I should point out, that the NetApp & Cisco support has been in place for quite some time. Recalling the existing install base has compelled me to expand upon today’s announcement with particular emphasis on items relevant to FCoE architectures and support statements.
In an ideal world all IVA deployments would comprise of net-new compute, network, and storage devices; however, most companies don’t have the resources available to coordinate the amount of change required to implement one oflarge or even modest scale. It is in these types of engagements where the flexibility of our architecture truly shines.
One of the most desirable features of Cisco, NetApp, & VMware IVA deployments is the unprecedented level of flexibility provided in the joint architecture. Many are unaware that we currently support End-To-End FCoE deployments for both the UCS B-Series & UCS C-Series Unified Computing Platforms along with more traditional server and blade server platforms. In all of these cases vSphere and a Cisco Nexus 1000V switch are required. In a similar fashion we also support the use of traditional storage arrays when they are virtualized with NetApp vSeries.
Though this flexibility we are able to accelerate customers’ adoption of cloud computing, by allowing them to build a scalable architecture which adapts to their business demands such as requests for additional services, data center asset depreciation schedules, etc.
More Details Around End-To-End FCoE Connectivity
The certification announced today applies to deployments based on Cisco USC C-Series unified compute servers and traditional server/blade servers where the FCoE connectivity is based on a model of direct connections from the hosts to Nexus 5k switches.
As you may be aware, today the UCS B-Series is challenged with end-to-end FCoE connectivity due to a lack of support for FCoE initiators northbound from the USC 6100 Series Fabric Interconnect to the Nexus 5k switch. This limitation is slated to be addressed in the not-to-distant future; however, it has been one which has required customers to have to consider dual-fabric topologies.
To clarify a dual-fabricrefersto twodiscretenetworks comprised from a Nexus 10GbE Ethernet fabric and a MDS Fibre Channel fabric. Note, I should add that FC-only storage arrays also force the use of dual-fabric deployments. The continued use of thesetraditionalarrays after support for FCoE initiators is added to the UCS 6100 will still require the existence of dual-fabric topologies (or one tovirtualizethem — ahem… vSeries hello?)
Engineers from Cisco, NetApp, and VMware have eliminated the need to deploy a dual-fabric topology with a fully supported configuration that we refer to as an End-&-End FCoE design.In this design we leverage Fibre Channel uplinks between the Fabric Interconnect and the Nexus in order to eliminate the need for FC switches.
Again, this issue is temporary and will be resolved; however, don’t let it force you to deploy dual-fabrics or hold off on adopting converged 10Gb Ethernet. The End-&-End FCoE configuration has been tested & validated by Cisco & NetApp and you can find the support statement in the NetApp Interoperability Matrix (IMT) on NOW.
I hope I haven’t confused anyone by expanding the conversation of end-to-end support for FCoE by raising the issue around the UCS B-Series and how we have addressed it with an end-&-end config. the purpose for including the B-Series is our joint adoption of B-Series in IVA solutions is significant and we can deliver supported, single fabric architectures today. The future of cloud computing is converged 10Gb Ethernet.
The Perspective From Our Partners
I want to share with you these quotes from four of our key IVA solution partners regarding today’s announcement and IVA based solutions.
(note we have over 70 IVA partners today)
Brendon Jarvis, General Manager of Strategic Growth, Empired Limited
“Empired is excited by the collaboration of Cisco, NetApp and VMware with the industry’s first end-to-end FCoE solution for virtualized environments. These vendors form an integral part of our business today. With these key partnerships, Empired is able to provide a clear vision to our customers and offer solutions that accelerate business agility and enable a truly dynamic infrastructure platform. Empired is committed to this relationship and we are pleased to be collaborating with Cisco, NetApp and VMware to help deliver the value and simplicity of an FCoE solution to our dynamic data centre customers across Australia and Asia Pacific.”
Rob Christ, founder and director, Databasement
“Vital to our solution is a unique service-oriented infrastructure based on the Cisco, NetApp, and VMware offering, that includes all server, storage, and networking hardware and software to facilitate sharing, reuse, and dynamic resource allocation. The announcement that VMware has validated Cisco and NetApp end-to-end FCoE in virtual environments will help us consolidate our data center infrastructure so we can achieve greater efficiency, performance, and overall cost savings.”
Bob Olwig, vice president of Corporate Business Development, World Wide Technology
“VMware’s certification of Cisco networking with the NetApp Storage FCoE solution is significant for our customers who are demanding an end to the complexity and headaches associated with a myriad of cables, multiple interface cards, and switches. World Wide Technology is laser-focused in helping customers integrate disparate technologies in complex physical and virtualized environments. We are very pleased to be workin
g with Cisco, NetApp, and VMware, which have delivered the industry’s first FCoE solution that squarely addresses data center customer requirements.”
Mark Hilz, president and chief operating officer, INX
“A validated end-to-end FCoE solution from virtualized server through the network to the storage arrays will help our customers drive a level of consolidation, virtualization, and automation that up until now was not possible. The collaboration among Cisco, NetApp, and VMware continues to yield technology advancements that accelerate business agility and delivers on the promise of enabling a dynamic virtualized data center.”
As more enterprises move towards a flexible cloud architecture; the Imagine Virtually Anything Alliance of Cisco, NetApp, and VMware will continue to collaborate with our solution provider partners, systems integrators, and service providers to help customers accelerate their journey to the cloud. A cloud architecture built on converged 10GbE for FCoE and NFS seems to be ideally suited for Cisco and NetApp to provide the physical foundation.
Dejan Ilic says
First I have to congratulate you on being the first company to get the certification.
I hate to rain on your parade, but I have to comment this.
For a while during autumn I was really excited about FCoE and hoped to be able to drop the aging interconnect infrastructure (read : Fibrechannel) to our Networking group and at the same time go for 10Gb/s on ethernet side.
Then I found out that our Netapp (realy an IBM N-series), for some unknown reason only supports FCoE protocol on the card that uses Ethernet for transport!
Having only one slot to work with the decision was simple, we exchanged last weekend our 4x1G/s card with a standard 10G/s ethernet card and we will use ISCSI when high speed block access is needed. Annoyingly the card bought in 2010 doesn’t do TOE offloading on our Netapp, again for unknown reasons looking at a fairly capable ethernet card in itself (manufacturer : chelsio).
This situation feels really strange, especially on the day Netapp announces support for Intel x520 cards capable of standard (TOE accelerated) IP ethernet, ISCSI and FCoE, giving the impression that you can do “triple play” with the same card on you Netapp.
Please correct me if I’m wrong regarding the FCoE cards on Netapp storage.
Vaughn Stewart says
@Dejan – Thanks for the kudos and the questions.
You should be able have complete CNA functionality (FCoE Target & Ethernet I/O) with Data ONTAP 8.0.1 scheduled for release this fall. Now the N-Series release may be a few weeks behind the official NetApp release (seems to happen on occasion).
As for the Intel card, I’ve always been a huge fan of Intel’s Ethernet chipsets, so I have high expectations of the x520. The ‘triple play’ as you put it, is exactly what a CNA is designed to deliver, all protocols on one wire.
Dejan Ilic says
Nice to hear, even if it comes too late for us. Given the (regular) lateness of IBM releases it means at earlies during next spring then.
It might be a question to your Ethernetguy but you didn’t comment the TOE remark.
I notice that we used about 50% less CPU on the Nseries 6040 (Netapp 3140) with our FC-card compared to our previous 4x1Gb/s card on similar thruput. Now we burn all our available CPU in the controller at about 3,0-3,5Gb/s @ about 6000NFSops so we can’t realy use the available 10Gb/s in the current config.
How much is due to difference in frame size on Ethernet/NFS vs FC and how much is due to offloading on FC I don’t know.
Given that the FCoE card can behave both as “FC” and “Ethernet” I feel it to be important to use available HW offload because the change of CPU would require controller change and that is out of the question. The way you use the card enables or limits its usefullness to us as current owners of Netapp systems.
There seems to be a lot of reasons to upgrade to v8.0.1 when it is released..
Can’t seem to find it in the IMT. Can you provide the quick ref number to get to it directly?