Key Themes for Intelligent Networks
My 25+ years of experience in analyzing the IT, enterprise network, telecom, UC, and security markets causes me to reflect on some of the hot topics/questions that I believe will impact the intelligent network over the next five years. I firmly believe that it will be an Intelligent Network– one that will require scalability, reliability, security, quality of service, programmability, etc. – not a commodity market to deliver cheap, dumb pipes.
So here the 5 key themes (in no particular order) that I plan to write about:
Software Defined Networks (SDN)
This is the hottest topic in the enterprise network space today – one with lots of start-ups, standards efforts (e.g. ONF and Openstack), and user confusion. Software, not hardware is the key component of next generation network design. Manageability and programmability (i.e. the ability to customize an application to the network) are becoming key requirements, and both are in need of significant improvement. Despite being hard to define, I believe SDN is real and will have a major impact on the vendor landscape over the next five years – more detailed posting(s) on SDN to follow.
Despite the focus on capital equipment costs (CAPEX), OPEX which accounts for 80% + of enterprise and telecom network costs is the key problem/opportunity. Poor network management tools, proprietary management systems, and incomplete standards make the life of the network operator or network manager VERY challenging. Configuring the network is too hard, changing the network takes too much time, and any customization of the network is challenging. Dynamic allocation of network resources to specific applications at specific times is not really possible today. Managing the network requires dedicated, trained, experienced staff (expensive) that are hard to replace when they leave. The migration to larger networks (e.g. 100 millions of mobile users or hyperscale data centers) will continue to make the network management challenges worse – the reverse of Metcalf’s Law.
Network Security or Secure Networks
Security (and the lack there of) is one of the top problems facing IT and C-level folks in any medium to large organization. Given the trends in mobility and BYOD, having a true network perimeter is largely impossible – and securing the network is getting harder and more costly. The network security community is very fragmented – lots of suppliers and lots of different boxes. One customer I spoke to recently has 30 different types/brands of security products in their shop – and a mandate to significantly reduce the number of suppliers. Significant changes in the data center and the data center network (e.g. virtualization, scale, SDN, etc) mean that network security will have to change as well. One of the key questions is who will lead here between network vendors (e.g. Cisco and Juniper), security specialists (e.g. McAfee, Checkpoint, Palo Alto), and IT suppliers (e.g. IBM and HP).
I like how Avaya termed the problem – loosely unified communications. The challenge/opportunity is how to take the dozen or so ways we communicate in our business life and link them together. Easy enough to do with a Cisco, Microsoft, or Avaya IP voice systems plus IM, email, link to a calendar etc – if your organization has upgraded to one these (fairly proprietary) platforms. But what about your mobile phone(s) and texting? And, Twitter, LinkedIn, Facebook, Skype, etc? I use all of these to communicate with my business contacts on a weekly basis. High end video conferencing is great if you have it (expensive) and if you can conference with the people you need to reach – doable inside a company, much harder with 3rd parties, suppliers, and others outside the company. Lack of interoperability will continue to inhibit high end video conferencing adoption – while “free” low end solutions continue to get better (e.g. HD).
I have spent a lot of time looking at this question over the last 5 years. Network, Telecom, and Security equipment platforms continue to the challenged by the need for very high performance (e.g. 10GB, 100GB, 4G) with the traditional needs for scalability, reliability, and long life span (5-7 years) – which does not align with the design goals of the IT industry (e.g. servers). However, the days of customers being willing to pay for specialized hardware platforms is waning (fast). Many security suppliers, with the exception of high end firewalls, have already moved towards commercial off the shelf platforms (COTS) – i.e. servers/Intel. Most enterprise networking and telecom suppliers have not. For example, Ericsson still makes its own proprietary blade server. Cisco and Juniper spend a significant part of their R&D on ASICs. The key question is which platforms (e.g. mobile core) are ripe for conversation and how will SDN accelerate COTS in the traditional (large) Ethernet switch and router businesses.