April 26, 2011

Path to Interop: Open Fabric for the Cloud

Interop Las Vegas is approaching.  Given the recent buzz around network switching fabrics for the data center, I thought it would be good to engage in some discussion around this subject.

First off I need to address: Just what is a network fabric?  While there is no formal definition, some attributes of a data center switching fabric could be the following:

  • A fabric provides high-speed, low-latency interconnectivity
  • Non-blocking/non-oversubscribed would be another requirement
  • Arguably, fabrics are being equated to Layer 2-type connectivity
  • Multiple active paths with fast failover
  • Mesh connectivity rather than a tree-type topology
  • Simple management, configuration and provisioning would be another desirable attribute

So, given the above set of attributes, what is the best way to build a fabric for the cloud infrastructure?

Well, looking at a lot of the Cloud providers, their cost of goods sold is their infrastructure. So it would follow that many Cloud providers would want to avoid proprietary technologies that would lead to vendor lock-in and lack of pricing leverage. So perhaps one could argue that a non-proprietary, open standards-based approach would be a good way to go.

The second requirement, again going to the cost argument, is that 40GbE pricing compared to 100GbE pricing is more palatable for the Cloud providers. Given that servers are moving to 10GbE, the access layer of the network is going the 10GbE route. That means that the interconnectivity or fabric to tie the access layer switches together will move to 40GbE as the more cost-effective technology. So a high density/high fan out, non-blocking 40GbE solution can provide a good, standards-based approach to providing high-speed, low-latency interconnectivity.

The third requirement would be the ability to provide a mesh of active active paths — for two reasons. One is to avoid taking multiple hops up and down a traditional tree- type architecture. The other is to utilize bandwidth on all available paths. Here again, there are standards-based approaches to multi-path forwarding such as TRILL or SPB. If the access switches connecting into the fabric have dual uplinks, then perhaps even an MLAG-type approach may be feasible.

Lastly, the requirement for simplified management is being met once again through an open standards  approach. OpenFlow holds great promise as an up and coming, breakthrough technology for provisioning, configuration and administration. Already initiatives such as OpenStack are looking at OpenFlow as an open approach to provisioning in large, Cloud-scale data centers. The Open Networking Foundation (ONF), which will drive the OpenFlow effort, is backed by a large cross-section of both consumers and providers of this technology.

The combination of these technologies, i.e. high density, non-blocking standards-based 40GbE interconnectivity, multi-path support, and a standards-based provisioning solution such as OpenFlow should provide a viable, open-standards solution to the demand for Layer 2 switching fabrics for cloud architectures.


About The Contributor:
Extreme Marketing Team

See My Other Posts