This syndicated post originally appeared at Network World Zeus Kerravala.
Being an analyst I’m often asked to look ahead and predict what markets will look like in three to five years. Recently, I’ve been asked that if I were to design a next-generation network, which would be fully operational by 2020, what would it look like? The network industry has gone through more change in the past two years than it had gone through in the previous two decades so a network in 2020 will look significantly different than the networks of today.
Predicting the future is obviously challenging but sometimes it’s better to look back to understand what the future will look like. In networking, there are many lessons we can learn from the last 20+ years where networks have evolved to be able to meet ongoing application and business needs. The innovation was obviously necessary but each time something new was bolted on to the network, there was a price to be paid. For example, WiFi being deployed as an overlay means having to manage two separate networks. Below are the top lessons learned from legacy networks over the past two decades.
- Complexity is the enemy. Networks have had so many things bolted on over the years that making even simple changes takes months. We continue to do this too with protocol overlays, NAT to handle addressing schemes, etc.
- Refrain from proprietary solutions. This may have short-term advantages but creates lock in long term.
- Can’t manage or secure what you can’t see. Network visibility is extremely limited to rudimentary tools like Ping or Traceroute. This is why dashboards can be green but things aren’t working.
- Nodal configuration and troubleshooting. Each node needs to learn network information independently. Leads to slow re-convergence times and painful upgrade cycles.
- Active/standby model limiting. Complex, inefficient and a waste of bandwidth.
- Lack of automation limits network agility. Scripting is not automation; it’s just more efficient manual processes, unless there are errors, then it’s less efficient, unless you have scripts to fix scripts. Enough said.
- IT dependency and defensive position causes silos. Apps can’t be deployed without IT and networking intervention.
- No orchestration capabilities. Network operations are a series of manual tasks.
- Multicast deployments difficult and not scalable. Looks good on paper but adds more protocols and ultimately more complexity.
- Wireless LAN as a best effort overlay network. Adds to complexity and never seems to work when it’s needed most. Hard to troubleshoot.
- Access control not broadly implemented. Another feature that looks good on paper but hard to implement.
- WAN connectivity is limiting businesses. WANs are inflexible and at least an order of magnitude slower than LANs.
In retrospect, the networking industry accomplished what it had to. Despite all the complexity and efforts, large organizations and service providers were able to run their businesses over a fragile but stable network. The fragility though is what causes hesitation in making changes and why sticking with the status quo is better than risking the unknown with something new.
So, if by 2020 we can address the above limitations, what could the Network of 2020 look like? Here is the key attributes that they network would have:
- Simplified network architecture
- Mobile centric. Any application, anywhere, any device enabled by unified access and multi-gigabit WiFi
- Enhanced and customized customer experience with location and proximity detection
- Automation and orchestration capabilities that are driven from the application
- Full visibility of infrastructure behavior to predict performance issues before users call
- Active/active model everywhere
- IPv6, not everywhere but at least somewhere
- Meshed Ethernet architecture optimized for IP services
- Cloud services flexibility (Private, Hybrid, Public)
- Network agility to support digital business resiliency
- Big Data analytics to trigger business workflows
- Multi-tenant services embedded
- Control Plane separation/segmentation. Not necessarily SDN, but could be implemented that way
- Software implementation of network services for elasticity
- Analytic approach to layered security model
- Standards and open architecture for greater interoperability
- Secured and automated IoT provisioning
- Hyper-converged support leveraging Ethernet meshed architecture
- Ethernet and broadband WAN become the norm
- On-demand services elasticity
- One virtualized Enterprise (Data Center, Campus, Branch)
Latest posts by Zeus Kerravala (see all)
- Forget Heliocentrism-Embrace the cloud and Zeus-centrism - April 17, 2017
- Avaya’s post-bankruptcy plan should not impact customers, partners - April 14, 2017
- There’s never been a better time for Cisco Services - April 12, 2017