The death of spanning Tree - Here comes Intelligent Resilient Framework technology

The death of spanning Tree - Here comes Intelligent Resilient Framework technology

Anybody who has worked with traditional three-tier networks will appreciate what a pain spanning tree can be in large deployments. STP is widely used, and is fairly effective at making a switching network resilient— that is in basic terms, dealing with a sudden failure of a link between core, aggregation or edge switches. In a network that is operating normally, when a link or switch goes down, STP automatically chooses a backup path, and the network reconverges. After a short period, the network is operating normally again. But for modern, high-speed, mission-critical networks, this approach may not be the best solution, especially if you are running complex high performing virtual technology services.

The problem with STP is simple.

The reconvergence time for STP can be several seconds depending on deployment complexity—a lifetime by modern computing standards. When network latency has been pushed to near real time sub millisecond delays, STP is now out-dated.  Say you are using iSCSI or FCoE for your resilient SAN infrastructure, or even utilising VMware DRS, the few second delay experienced with STP reconvergence will cause issues. 

 Something I have experienced personally after D.R testing.

Even though MSTP and RSTP converge more quickly than the original STP protocol, all of these protocols can be painfully difficult to configure properly, especially in large networks. You must manage the switches individually, and need to set up spanning-tree instances on each switch in turn, making sure that the parameters for one switch match those of its neighbor.  This means troubleshooting spanning-tree-related issues is no walk in the park, usually requiring a great deal of time to locate the root cause of the failure.

Lastly, STP is a resource "hungry" protocol and was not designed for performance. STP blocks all parallel paths except the one it has selected as active, even when the network is operating normally STP actually reduces the effective bandwidth. In fact, half  of the available system bandwidth can be squandered using STP.  Working with STP on cross vendor products will also cause issues.

During a visit to the tower of London to see the latest HP networking portfolio, I was greeting with a new technology to which I was very impressed with. The answer to STP; HP IRF or Intelligent Resilient Framework technology.

IRF technology extends network control over multiple active switches. Management of a group of IRF enabled switches is consolidated around a single management IP address, which vastly simplifies network configuration and operations. Unlike VRSP or VIP configurations requiring multiple reserved IP address.

Basically - visualisation of your switching infrastructure, spanned over multiple physical switches.

HP state that you can combine as many as nine HP A-series switches to create an ultra-resilient virtual switching fabric comprising hundreds or even thousands of 1-GbE or 10-GbE switch ports..

The new buzz word for this is “FlexFabric”

As you can see from below; you can now get away from using aggregation layer switches and simplify large scale deployments – lower capital cost while increasing redundancy, yes please!

One IRF member operates as the primary system switch, maintaining the control plane and updating forwarding and routing tables for the other devices. If the primary switch fails, IRF instantly selects a new primary, preventing service interruption and helping to deliver network, application, and business continuity for business-critical applications. Should a network failure occur, IRF can deliver rapid recovery and network reconvergence in under 50 milliseconds.

Within the IRF domain, network control protocols operate as a cohesive whole to streamline processing,  improve performance, and simplify network operation.  Within an IRF domain, the geographic location of switches does not matter. Switches can be extended horizontally, and they continue to function as a single logical unit whether they are installed locally, distributed regionally, or even situated at distant sites. Moreover, employing IRF can enhance disaster recovery by linking installations up to 70 kilometers apart and giving them the same fast failover as if they were sitting side by side within the data centre.

So routing protocols calculate routes based on the single logical domain rather than the multiple switches it represents. Moreover, edge or aggregation switches that are dual homed to IRF-enabled core or data center switches  as a single entity, eliminating the need for slow convergence technologies such as STP.

All in all, the new networking portfolio from HP after the acquisition of 3COM, H3C and Tipping point, is starting to look very promising. Couple this TCO of up to 66% less compared to like for like models from Cisco, I predict a strong and exciting fight coming from the HP camp this year.