2024 network plans dogged by uncertainty, diverging strategies

It’s barely fall of 2023, but it’s already clear that CIOs aren’t particularly positive about their network plans for 2024. Of 83 I have input from, in fact, 59 say they expect “significant issues” in their network planning for next year, and 71 say that they’ll be “under more pressure” in 2024 than they were this year. Sure, CIOs have a high-pressure job, but their expectations for 2024 are worse than for any year in the past 20 years, other than during Covid. Nobody is saying it’s a “the sky is falling” crisis like the proverbial Chicken Little, but some might be hunching their shoulders just a little.

It seems that in 2023, all the certainties CIOs had identified in their network planning up to now are being called into question. That isn’t limited to networking, either. In fact, 82 of 83 said their cloud spending is under review, and 78 said that their data center and software plans are also in flux. In fact, CIOs said their network pressures are due more to new issues relating to the cloud, the data center, and software overall than to any network-specific challenges. Given all of this, it’s probably not surprising that CIOs say they spend less time on pure networking topics than at any time in the last 20 years.

The combination of the Internet and the cloud has changed computing, and the top mission for enterprise networking is connecting customers, prospects, partners, and workers to applications. Virtual networking has taken over the WAN (MPLS VPNs) and the data center (Kubernetes container orchestration includes virtual-network plugins), and obviously it’s a fixture in the cloud. These virtual networks are what users and applications see, and network engineers really don’t have much to do with their technology at all. These same changes have magnified security risks, which is what created the notion of a separate security group.

I’d argue that all of this is tied to a single trend: virtualization. The goal of virtualization is to create something that looks like a traditional resource but is really a service. Virtualization is a form of abstraction; you represent all those routers and trunks with a service that looks like routers and trunks but is managed by the provider. You represent a data center network as a real LAN, but it’s actually an overlay on a LAN – an overlay you can change and move without impacting gear. It disconnects users and applications from reality, in a sense, and gives more control to the things and people who use the network than to the technology of the network. That’s pulling the traditional network apart as each mission pulls virtualization in its own direction.

Network vendors are responding. Cisco, which has a line of servers and data center software, just bought security vendor Splunk to beef up its security, software, and AI credentials and support those separate missions directly. Juniper’s Apstra product virtualizes the data center network to support configuring it based on goals or “intents,” and it now supports Terraform-linked network changes and added measures for data center and application security. Other network vendors are also doing more for security and AI, and so are suppliers of software and software development tools. There’s no shortage of strategies here.

And that pulls networks apart even more. Every network operator and vendor will stand up and pledge standards support, but every vendor will at the same time design their products and strategies to pull through their whole portfolio. Add different strokes for all those vendor folks to different mission drivers, and you understand why 79 of those 83 CIOs say they don’t really have a “single network architecture model” in place, and 48 say they’re actually moving away from a standard approach. Virtualization can make the unreal look real, so why not allow multiple personalized “unreals”? Look into the virtual-network mirror and you see…yourself.

Nowhere is this more visible than in the management space. A couple decades ago, companies had a “network operations center” and a “single pane of glass” to show network status was the goal. Only 14 of 83 enterprises said they really had a NOC today, and when asked about a single pane of glass, one CIO quipped “I have five single panes of glass!” CIOs say that the current craze in “observability” is a response to the fact that it’s become very difficult to determine what the cause of an outage is. Calls to customer support almost never report the actual problem; the caller will say “I can’t use my network” when in fact they’re conversing with customer care over it! The actual problem could be a configuration problem in the network, but it could also be a security policy error, a failed server in the data center, a cloud outage, a VPN BGP configuration…you get the picture.

OK, you may be thinking, but none of this is new. Why is 2024 shaping up to be so troublesome? The answer is the combination of vendor initiatives to try to increase sales and changes in the drivers and missions that impact network requirements. That starts with the cloud.

“Three years ago, I was supposed to be planning to move everything to the cloud,” one CIO said. “This month the executive committee said that cloud costs were too high and we had to think about ‘repatriating” some applications, meaning bringing them back into the data center. Everything old is new again.” The company had been planning a SASE deployment, and now some of the applications won’t be in the SASE cloud domain. They must either pull them back to the MPLS VPN or use SD-WAN. But at the same time, the executive committee said they needed to make “product search easier and improve purchase rates,” and in some cases this targeted applications that were also being considered for repatriation.

How do CIOs think they’ll address these converging pressures and diverging strategies (other than duck pieces of falling sky)? The answer CIOs gave most often (from 39 of 83 CIOs) was “a unified virtual network strategy.” First, this would address the fact that virtual networks make up the cloud, the data center, and their own VPNs. Second, it would let them change infrastructure and even computing and user distribution with little more than addressing and parameterization changes, reducing opex and capex. It could even let them hold on to current gear a bit longer.

The downside? Only 7 of the 39 CIOs who liked this approach have a specific plan to move on it, or a vendor or vendors identified. This is despite the fact that the majority vendor in 70 of the 83 enterprises has a specific virtual-network strategy, and at least one of the incumbent data center software platform vendors does too. Worse, the remainder of the 83 – those who didn’t mention virtual networking – couldn’t muster any more than 11 backers of any other strategy. In short, there is no consensus on how CIOs should face 2024 network challenges.

Gosh, maybe Chicken Little was right.