I almost fell out of my chair a week ago Friday when HP posted a link to an overview of the Cisco Fabric Extender for HP BladeSystem. If it hadn’t been for tweets by Cisco, HP’s 180-degree reversal would have gone unnoticed in a time when mudslinging has become the networking industry’s de facto message, nowhere more apparent than in Cisco’s live video by Rob Lloyd, “Debunking the Myth of the ‘Good Enough’ Network and HP’s two-year shock-and-awe campaign against Cisco and its architecture with such posts as:
- “The Real Story about Cisco’s ‘One Giant Switch’ view of the Datacenter.
- “HP BladeSystem c7000 with ProLiant BL460c G6 Servers vs Cisco UCS 5100 with B200 Servers Network Bandwidth Scalability Comparison,” a Tolly Report commissioned by HP.
- “HP Networking’s Vision,” an August 2010 podcast with Nick Lippis.
The thawing of the cold war between HP and Cisco breaks down the walls that limited choice and, hopefully, ignites more innovation between the companies. HP customers can now connect and configure their HP BladeSystem c-Class infrastructure with Cisco’s Unified Fabric. Quietly, HP has been offeringCisco’s legacy switching technology in its c-Class BladeSystem enclosure but halted integration with Cisco’s next generation data center products after HP acquired 3Com and Cisco released UCS . Similar to Juniper’s QFabric architecture that extends virtual line cards to server edge, the relationship between a fabric extender and a Nexus 5000 is similar to the relationship between a traditional 6500 line card and its supervisor engine, only now the fabric extender can be connected to its master switch - Cisco Nexus 5000 — with remote fiber connections. This allows I&O to effectively decouple the line cards of the modular switch and spread them throughout the data center, all without losing the management model of a single “End of Row” switch. The master switch and all its remotely connected fabric extenders are managed as one Cisco Nexus switch.
As Rich Fichera points out in his blog, customers are the winner in this coopetition. Even though the walls are coming down, this doesn’t signal the end of HP networking, but customers should be wary of HP’s networking commitment (see Prediction: HP Cuts Loose Their Networking Hardware And Transforms Into A True Networking Alternative ).Customers have been telling me they’re putting pressure on HP’s Enterprise Servers and Storage (ESS) of ESSN (Networking) to open the door to choice. Infrastructure and operation teams are telling Forrester that they want the option to leverage newest data center alternatives to HP’s networking infrastructure so that they can:
- Simplify operations. The top priority within I&O over the past three years has been to increase efficiency. Most networking vendors have responded by offering their own way of eliminating the link down state in spanning tree. A few others have taken increased utilization a step further and offer the capability to manage a large pool of switches under one logical grouping like Juniper’s QFabric. With Cisco’s FEX technology, customers can manage up to 24 modules under one logical grouping, which wasn’t possible under HP’s Virtual Connect technology and their IRF technology in their switch line.
- Standardise processes. HP’s Virtual Connect (VC)technology was designed by server teams to simplifying server operations; for example, virtual connect technology allows organizations to move or replace the server without reconfiguring MAC/WWNs, but VC has fallen short in the network controls. As with SANs, there’s no need for VM or server administrators to manage network policies in the hypervisor while network administrators do it on physical switches. Consequently, I&O personnel want solutions providers to move that network demarcation line from the physical port to the virtual one on the hypervisor; network administrators’ roles are evolving so they take ownership of monitoring and controlling traffic in the virtual world. Cisco’s technology provides consistent control from the physical to virtual world under one management domain.
- Scale capabilities. Forrester Research surveys indicate that few enterprise infrastructures have matured their infrastructure to the point where they have created an abstracted set of resources (storage, software, compute, etc), assessed in a standard way and billed by use. This means a majority of data centres are in the process of evolving with virtualized hardware and legacy systems. Customers can tie them together with Cisco’s 1000v, 2000 series, and HP’s new Fabric Extender.
- Implement it now. At this point, I&O managers are telling us that IBM, Dell, and HP’s converged solution sound good on paper but are too restrictive at this point since they are in the infancy stage. In addition customers are saying they aren't bothered by the proprietary nature of the data centre network, because they can’t wait for VEPA, TRILL, or OpenFlow functionality—mostly waved around by the convergence vendors--and feel all data centre networking solutions have various levels of proprietary functionality: QFabric, VCS, IRF, and FabricPath. Consequently, customers are choosing to stay the course and purchasing networking solutions from traditional networking vendors: Arista, Brocade, Cisco, Juniper, etc... at least for right now.
What does this mean? More choice. More flexibility. The old days of mainframe computing are over, and there is no sign of customers going back to that model no matter how much marketing spin is put behind converged infrastructures. Besides HP, we’ll see IBM and Dell embrace Cisco’s FEX.
And you might ask, “What about their networking acquisitions?” The real value behind them will be the development of networking expertise to create better orchestration solutions.
What approach are you taking when it comes to your infrastructure? Are you buying server, storage, and networking from one vendor, or do you prefer to keep networking as a separate purchase? Why?
Posted by Andre Kindness