Embracing Digital Cohesion by Dealing with the “Performance” Barrier
Nov 22, 2016
What’s the next digital frontier? It’s Digital Cohesion, a future in which applications we use often self-assemble to deliver compelling mega services that enhance our lives. Today, if I need a ride to the airport I might order a ride on Uber. Or if I want dinner delivered to my home I might order a meal through doordash. In doing so, I am using applications manually. In the future, these services will act autonomously, and deploy based on my situation, need, and location, as well as my history and feedback from other services or devices I’ve used. I will no longer need to call such services, as they will automatically appear and serve when needed.
Fueling the Digital Cohesion revolution is a blend of ubiquitous connectivity, sensors and microelectronics that allow almost anything to connect to the Internet. This means heavy traffic. Wherever there’s heavy traffic, there are traffic jams. Why do traffic jams occur? Usually because there’s heavy traffic on the same road that was meant for lesser traffic. What’s the solution? Build more roads or highways or reduce traffic. In the Internet world reducing traffic is never an option. By the end of 2016, global IP traffic will reach 1.1 ZB per year, or 88.7 EB per month, and by 2020 global IP traffic will reach 2.3 ZB per year, or 194 EB per month1. Data traffic is only going to increase and you better have high performance networking (a.k.a., roads/highways) to handle that growing traffic.
In the Digital Cohesion world, all IoT devices will seamlessly connect with each other and this means one thing for sure – large data sets for your applications. Huge computing power will be needed to work with all that data – to analyze and to run machine learning algorithms so that the system can have better AI.
Today, applications run on end-user level devices that have compute and storage limitations. Therefore, to deliver the next-gen autonomous experience, compute and storage will inevitably need to move to the cloud. In making that happen the underlying network equipment and architecture must be able to provide high performance and availability. Essentially, the network must be able to provide the flexibility to support numerous virtualization environments, as well as the shorter deployment times that tomorrow’s services will demand.
If companies like Uber, Facebook, Airbnb, Amazon and Google have taught us anything, it’s that customer experience comes before anything else. An application that opens up slowly is dead to the user. 40% of people abandon a website that takes more than 3 seconds to load2. Slow is the new down. That’s the hard truth and what that means for service providers and enterprises is that they must make sure they have enough performance in their data centers to support what’s out there today and what’s coming next. Simply put, the network between the end-user and the cloud, as well as the network between the clouds, should never become a bottleneck to the required performance.
Data growth means increase in network strain
Clearly, Digital Cohesion will generate an unprecedented amount of data traffic and as a result the workload for business critical applications will also increase. To support that increase, networks will need to scale. Today, because of virtualization many of those applications are running in a virtual machine (VM) on a shared hardware. The overall industry average of VM density today is about 10 VMs per server3. With containers, which are inherently lightweight the density per server will even be greater than what VMs provide today.
Increase in the application density per server will increase the number of IP and MAC addresses in your network. A critical element of network scalability is the size of forwarding table in network routers and switches that store those addresses. Essentially, the size of forwarding table impacts the number of physical servers in your network as well as the number of VMs per server. When forwarding table reaches full capacity scaling problems occur. For instance, if number of MAC addresses in your network is larger than forwarding table some MAC entries will be lost. Any subsequent packets delivered to those MAC destinations will cause flooding and severely degrade your network performance. Simply put, forwarding table size directly impacts application performance and mobility. Every company works hard to earn business. The only thing that will get you repeat business is high quality customer experience. Your application performance is dependent on high performance scalable network. And any degradation in the application performance directly affects your customer experience.
There’s no doubt that high performance networking will become a key differentiator for enterprises and service providers in delivering the ultimate customer experience. At Juniper Networks, delivering high performance networking has always been our goal. To meet your scaling needs, we developed custom silicon in-house for our network routers and switches that gives you unparalleled capacity and performance to run your applications. Our EX / QFX / MX / PTX offer forwarding table sizes that is way ahead of what our competition offers. Moreover, we switch traffic like no other and are fully equipped to help you truly embrace the new opportunities put forth in the era of Digital Cohesion.
For all your networks: campus, data center, network edge and core we have a solution. In campus, we offer the most programmable enterprise core switch – EX9200 series that runs on our Juniper One silicon offering scale of 1 million MAC address and 256,000 IPv4/IPv6 routes. In data center, our core network switch - QFX10K powered by Q5 chipset offers scale of 2 million entries in FIB and 2 million IPv4/IPv6 host routes with 1 million MAC addresses. In network edge, MX2K with its leading Trio chipset offers up to 10 million routes in FIB and 50 million IPv4/ IPV6 routes in RIB. Finally, the sheer power of ExpressPlus silicon in PTX5K gives you the unprecedented capacity that makes sure your core needs are fully covered.