The move from 1GE to 10GE servers started some years ago as server virtualization took hold and just about every IT outfit saw the value in server consolidation & server virtualization that brought significant cost reduction as well as operational efficiencies. The shift from 1GE to 10GE on the server side forced the shift deeper in the infrastructure from 10GE to 40GE as a result.  In data centers, links between leaf-and-spine switches went from 10GE to 40GE. Service providers on the other hand, particularly telcos, moved to 10GE/100GE rather than 10GE/40GE infrastructure for a very reasonable and business driven desire to pack as much data as possible on their long distance (often single mode) fiber links.

With the release of new platforms that are based on Broadcom Tomahawk silicon, we will start to see the next shift in the server access speeds from 10GE to 25GE. It makes economic sense to follow this transition because 25GE provides 2.5X the performance of 10GE while being fairly close to 10GE on the cost per port basis. As the servers move from 10GE to 25GE, just as what happened in the previous cycle, the links that are deeper in the network will move from 40GE to 100GE. One way to think about that is you want to always select the uplinks that are 4X your server access speed to build networks with fairly low oversubscription while providing some level of cabling simplicity. In the previous generation, customers built networks with 10GE server access and 40GE (4X) uplinks. In the next generation, we anticipate that 25GE servers will result in the rest of infrastructure being 100GE (4X of 25GE) in the data center.

This brings me to why I am writing this blog. How does one monitor & secure a data infrastructure that is fundamentally evolving to next generation of server access speed every 18-24 months? By leveraging the same merchant silicon ecosystem to build the visibility fabrics that customers are using to build the production data centers instead of using complex, proprietary, expensive technologies that simply can’t keep pace and scale with this server infrastructure upgrade cycle. 

Big Switch Networks has revolutionized how visibility fabrics are built by combing the power of open networking hardware leveraging merchant silicon technology and software defined networking. By doing that, customers can not only keep pace with their data center infrastructure from one generation to the next, but can also have visibility fabrics that keep pace with that change in the data center or the service provider mobile networks. Whether customers are monitoring 100GE data center infrastructure using TAPs and SPAN ports and bringing the traffic to out of band security and monitoring tools or by placing security tools inline at the DMZ for having production traffic be inspected actively as it comes & goes out of the data centers, Big Switch has solutions.

 Big Monitoring Fabric currently supports Broadcom Trident2 based switches from Edgecore networks and Dell.

In order to provide high scale 100GE monitoring, Big Switch has added the support for  Big Monitoring Fabric software on Broadcom Tomahawk-based switches from Edgecore and Dell. By doing so, monitoring or visibility fabrics can scale relative to the production networks without suffering from complexity, lack of scale and proprietary features. The Dell Z9100-ON and Edgecore 7712 switches are powered by Broadcom Tomahawk silicon and pack 3.2 Tbps of capacity supporting 32X100GE wire-speed ports in 1 rack unit. By placing these switches in a scale-out configuration, customers have the flexibility to build as large or as small of 100GE inline or out of band visibility fabrics as needed. The additional bonus is customers have the choice to deploy the same switches as data center fabrics just by running different software.

Welcome to the next generation of high speed monitoring!


Salman Zahid 

Systems Engineering