From Applications to ASICs in 3D: Changes, Part Three

3D: Digitization and Disruption in the Data Center

This week at Partner Summit, Cisco made major headlines with Data Center announcements in 3 key areas: Cloud, Hyperconvergence and Switching. One of the headlines reads “Partners Applaud Cisco’s Launch Of ‘Incredible’ New Line Of Nexus Switches, Software”. This blog, the 3^rd in a 3 part series, will discuss what makes these switches so compelling.

Let’s recap a couple key points from part 1 of this blog series. Digitization is having a huge impact on business.  Applications play a major role in this impact because they represent the key point of interface between business and customer.  Applications are essentially the reason Data Centers exist. Traditional applications are being deconstructed into microservices on containers that rapidly appear, disappear, and are moved around private and public clouds. These disruptive shifts have implications on both how and what IT architects, deploys, operates as well as troubleshoots. For more on how, see these links regarding announcements that address how we are automating/managing private and hybrid clouds with CliQr, Nexus Fabric Manager and DCNM 10.0. The rest of this blog will discuss the what, i.e. the switching infrastructure.

The trends above lead to several challenges that switches need to address. This 3 minute video from Brad Casemore of IDC succinctly addresses these issues:

A more detailed white paper complements the video and highlights how Cisco’s innovation creates differentiated value for customers addressing these challenges. It’s also an excellent summary of key issues in the industry today…Highly recommended reading. Excerpts from the paper (in italics) are included below.

One of the underlying principles is that the switching ASIC is the single largest factor influencing switch cost, performance, functional capability, and power consumption. Furthermore, the new Cisco ASICs effectively address challenges resulting from the trends above, which, as we discussed in part 2 of this series, include scale, telemetry, capacity, and intelligence. So let’s address some of the questions raised in that post.

Scale

The … use of containers and microservices will have repercussions for network scale. Cisco is responding by leveraging the expanded transistor capacity of its next-generation switch ASIC for increased route and end host scale.  The paper goes on to show how the new switches offer double, triple, quadruple, and in some cases more, scale than that of competitive switches.

Telemetry

As servers moved from bare metal to virtual machines, it became more challenging to know the location of a given server or to be fully on top of what it was doing. These challenges are only exacerbated with containers, as there are more servers moving faster. As a result, knowing what is happening with the infrastructure and the apps that run on it means there is a need for more information and to get at that information faster. Full flow visibility for every packet has not been available on any datacenter switch for the past decade because the costs to provide the required bandwidth and table scalability were prohibitive. With its new ASIC technology, Cisco is able to provide full flow information and increase the amount of flow telemetry almost fivefold at reduced cost.

Capacity

Containers and microservices also will drive a need for greater bandwidth and less oversubscription. Cisco is responding with bandwidth per rack unit (RU) that is more cost effective than that provided by merchant silicon–based switches. Cisco is offering 25G at the price of 10G, and 100G and the price of 40G. Joe Onisick addressed this quite concisely when he tweeted:

“Want future proof? How about buy a 25/100G switch from me today, for the price my competition sells 10/40G for over the next 2 years.”

Intelligence

As the trends above yield more types of traffic, e.g. distributed storage, more active/mobile endpoints, etc., it becomes more important to have the intelligence to effectively handle and deliver this traffic. To do this the new Cisco ASIC technology also delivers several enhanced queuing and traffic management features. It is beyond the space we have here and, more pointedly, beyond my mental capacity, to elucidate all the mechanics of how this happens, but if you have questions, leave a comment below and I can dig up specific papers/references with additional detail from people whose propellers spin so fast they are used in wind tunnels to test drag coefficients.

Summary

Digitization is driving massive change in business, with applications, and in the Data Center. This change must be reflected in both how IT operates (think software and automation) and what IT operates (think hardware and infrastructure). Cisco announced major news in both. The new switching innovations address these changes and provide value to Cisco’s customers in a way that is unique in the industry.

For more details on this week’s announcements, see the press release and Soni’s Jiandani’s blog.

And finally, this bears repeating: Read the paper.  It’s really good!

Photo credit: flickr.com