Ed Henry

Author Archives: Ed Henry

Three Areas of Impact on the Networking Industry

The state of the networking industry is undergoing a transformation. Both hardware and software have given way to the realization that applications and infrastructure are less about the individual functions of the sub-components of the overall system and more about the workflows that are being expressed against the technology as a whole.

Workflows, however, are also sub-components of the overall business logic that is being described, and ultimately expressed against the infrastructure. Micro-services and/or services oriented architectures are fundamentally changing many facets of technology, both in business practice and in features and functionality of the technology. They’re enabling a more agile and flexible technological landscape, and are helping to define administrative boundaries with respect to overall operations of the infrastructure.

Impact on Infrastructure

When looking at technology overall there is an obvious pattern emerging in all realms of the hardware that is enabling the industry. Programmability: the desire to express logic against infrastructure much in the same way that we’re able to express logic in the context of an application. “Infrastructure as code” is a phrase I’m sure we’ve all heard over the last couple of years, and this is something that is definitely being attained with the ecosystem of Continue reading

I (don’t) like Big Buffers.

Recently Arista released a white paper surrounding the idea that having deeper buffers running within the network can help to alleviate the incast congestion patterns that can present when a large number of many-to-one connections are happening within a network. Also known as the TCP incast problem. They pointedly targeted Hadoop clusters, as the incast problem can rear its ugly head when utilizing the Hadoop Cluster for  MapReduce functions. The study used an example of 20 servers hanging off of a single ToR switch that has 40Gbps of uplink capacity within a Leaf/Spine network, presenting a 5:1 oversubscription ratio. This type of oversubscription was just seen in the recent release of the Facebook network that is used within their data centers. So its safe to assume that these types of oversubscription ratios are seen in the wild. I know I’ve run my fair share of oversubscribed networks in the past.

Treating the Symptom

This particular study actually prods at what is the achilles heel of the traditional leaf/spine network design. All nodes being within 3 switch hops, (ToR <-> Spine <-> ToR), does provide a predictable pathing within the minds of the network operators of today, but I posit that Continue reading