Today, enterprise networking is at a very interesting and dynamic state. With the availability of LAN bandwidth speeds of 40GB and 100GB in the data center, virtual switching technologies with all links forwarding traffic, creative new Quality of Service (QoS) configurations, application optimization, intelligent path control and the addition of automated orchestration, the possibilities today are endless.
All of these current options for features and configurations are especially exciting because enterprise networking is the foundation of any solid networking design and construct. With all the features available, it’s easy to see how the most time-consuming aspect or sometimes forgotten part of an end-to-end network can now be constructed to provide return on investment to the business and efficient maintenance and operation for the engineers supporting it. So, in essence, a dynamic and intelligent enterprise network is going to allow converged technologies to work flawlessly, while delivering great application performance all the way to and from the data center.
“But Momma, I wanna go fast!!!”
Those were the words of a young Ricky Bobby in Talladega Nights. The truth is, we want our networks to be fast. With 1GB and 10GB speeds, as well as MultiGig (MGig) technology available at the access layer, and 40GB speeds available for uplinks in multiple port channel bundles of up to eight ports, it is now easier than ever to provide 1GB bandwidth to user desktops or phones and 10GB bandwidth to a WAVE 2 Access Point servicing your mobile clients. This speed availability is also very nice to have as a backbone construct in high-density mobile environments like public venues or stadiums. It allows flexibility and scalability to grow the bandwidth as your needs grow without any major forklifts.
Get all those links rocking man! Don’t be a dinosaur living in the past. For many years Spanning Tree has served its purpose and saved us, but with it comes the loops. The truth is, a properly constructed Layer 2 should not present any loops. We need to get these ports that are blocking traffic to forward traffic. What better way to do that than to go back to the classics: Virtual switching technology and port channels are great to get those expensive uplink ports and optics to forward all traffic on all links, while providing the high availability and redundancy that helps us IT professionals get a good night’s sleep. Stitch it all together with the right fabric — think Virtual Extensible LAN (VXLAN) from Cisco or Shortest Path Bridge (SPB) from Avaya.
Imagine if you got in your car tomorrow and there were no traffic lights, no HOV lane, no stop signs and, for that matter, no signage on where to go. That would be an epic disaster. Collisions and crashes, people and cars fighting for the right of way, and pedestrians being run over. Granny would never leave the house and, in essence, only the most aggressive drivers would survive, intimidating the rest of us into staying home. Sounds like a sad world to live in, doesn’t it? Well, guess what? This is the world your network traffic lives in with no QoS. You have crashes, collisions, drops, jitter, delay, packet trampling and just basic network lawlessness and chaos.
Quality of Service is essential to any and all networks. It helps prioritize traffic through the network, it guarantees service and bandwidth to the applications that need it, and it allows fair and equal access to the network. On today’s networks, speed is not a problem; QoS is more about addressing traffic drops then latency and jitter. QoS, at a minimum, should be implemented at the WAN edge where the bandwidth usually funnels from high to low. In best practice terms, it is best to classify traffic end-to-end on all access ports, trust those markings at the distribution and core layers, and match the number of buckets end-to-end. What does this mean? Don’t get crazy.
Twelve queues may sound super rad, but why? Four to six queues that can be retained end-to-end as they traverse the WAN is much easier to maintain. Remember, you can only queue to your lowest common denominator on the network. This will typically be dictated by the number of queues supported at the access layer or WAN carrier. Also remember you can nest markings. I find that QoS is only as hard as you want to make it. With some common sense and heads-up thinking, you can be the QoS ninja of your organization. QoS prevents network chaos and makes sure the user experience on your network is more like 2001: A Space Odyssey rather than Kurt Russell’s Escape from New York.
The World’s Most Optimized Network
To optimize or not to optimize that is the question? I say yes, using technologies like Cisco’s AppNav to replace the almost archaic Web Cache Communication Protocol (WCCP), along with the availability of virtual appliances and router containers. Optimization, at a minimum, keeps things light and it really helps with classic user traffic like Common Internet File System (CIFS) and the caching of web content. And, overall, the compression you get from transport flow optimization (TFO), data redundancy elimination (DRE) and persistent Lempel-Ziv (LZ) for Transmission Control Protocol (TCP) is very useful in any WAN. Installations that used to take weeks can now be performed in days. Also, complex configurations that required WCCP can now use AppNav as the redirection method for optimizing traffic on the WAN.
Now there are some disclaimers. Not all traffic is a good candidate for optimization. A medical image that needs to be viewed and is in a raw data format is a bad candidate for optimization. An Excel spreadsheet shared by 10 finance folks at different branch sites is great for optimization, as it will only send and receive delta changes. I live by a simple mantra when it comes to optimization on the WAN: I don’t always optimize but when I do, I prefer AppNav.
The ABCs of IPC
Intelligent Path Control (IPC) is probably the hottest topic in enterprise networking right now. I get asked about IPC WANs quite often — how to use and deploy them, what makes them different from SD-WAN. Well, let’s finally demystify Intelligent Path Control — its components, how it works and its dependencies for intelligent decision-making. I’ll start with the components of an IPC WAN.
The first and most important piece to an IPC WAN is decoupling from your service provider or WAN carrier. This can be done using Virtual Route Forwarding (VRF) and Dynamic Multipoint VPN (DMVPN) with Multipoint Generic Routing Encapsulation (GRE) and IPSec. Being able to put the provider in a VRF and decouple their routing from your global routing construct allows you to then create a unified overlay with one unified Internal Gateway Protocol (IGP). This removes overly complex route redistribution and allows you to use lower cost transport options, like Internet circuits, in conjunction with guaranteed services like Multipoint Label Switching (MPLS) — all under the same Interior Gateway Protocol (IGP). Sounds pretty rad right? Well, it gets better.
Once this tunneling construct is created, and assuming that you’ve done your due diligence on the aforementioned QoS markings and classification, you are ready to feed the IPC beast. With protocols like Performance Routing Version 3 (PFRv3) and DMVPN Phase 3 with Next Hop Resolution Protocol (NHRP) shortcut, the path intelligence can begin. DMVPN will not only find the shortest path to a site from a site, but creating a PFR domain will now make use of those dual tunneled transports. PFRv3 is going to be dependent on those QoS markings discussed earlier in this post. These markings will be used to create traffic classes that PFRv3 will use to decide which traffic takes which tunnel and also to look for drops, delay, loss and jitter to move traffic to the tunnel with the best available service path. Now the picture comes together.
The beautiful thing about PFR is it centralizes the traffic classes and their configuration to a master controller located on an appliance or a virtual router. This next-generation WAN will allow you to become independent of service provider route tables, allow usage of disparate transports, provide path selection and load balancing — not too shabby!
Configure Like a Boss
Now comes the glue that holds it all together, the icing on the cake, and what I like to call the coup de grâce: automation and orchestration. That’s right, manage the entire network from one place, avoiding inconsistencies and fat finger events that would have trashed your old network. These new tools promise to shake the very foundation of onerous and much hated change control. Much of this value is driven by the use of Python-driven scripted tools. Python is widely becoming a de facto standard in network operation scripting and replacing TCL. It’s flexible and, most important of all, powerful. With some creative coding, you can create a front end with variable inputs that can perform quick, mass configurations of anything imaginable for Layer 2 and Layer 3 configurations. Tear access control lists (ACLs) down as needed, shut down ports and switches based on usage — the possibilities are endless. Dynamically map your network daily or hourly with scripting — there are no limits. If you don’t have an application whiz in-house, it’s no big deal; just buy an out-of-the-box solution from your hardware vendor like Cisco’s Prime Infrastructure with APIC-EM Integration. This will give you the construct to centrally deploy templates and configurations like a scripting boss.
In today’s state of current enterprise networking and ongoing into the future, it’s going to be about doing more with less. Organizations’ IT systems and technologies are growing at a very fast pace — so fast that it has outpaced the organization’s ability to staff and find talent accordingly to accommodate the growth. This means it is our responsibility as engineers to do more with less. To accomplish this we need better tools and better ROI decisions, and we need to embrace a new era of networking that is not just more efficient but also more productive and innovative. The way I see it, the wave is coming; either you ride or get swallowed by the tide — the choice is yours.
This blog post brought to you by: