This post originally appeared in Human Infrastructure, a magazine about being a human in IT. Human Infrastructure is available with a free membership at Ignition, the Packet Pushers’ professional development site.
We have a lot of network orchestration tools out there. Cisco has NSO, ACI, Viptela, SD-Access, and many others.
VMware has Cloud Foundation, NSX, and VeloCloud. Other commercial options include Juniper Contrail, Apstra AOS, Gluware, and AppviewX.
And in open source, we have Open Daylight and ONOS among many others.
Just five years ago, Cisco’s CLI was king and there was only one way to configure the network. Anyone could spend a few weeks learning some basics, read a bit of documentation on the specific Cisco model you were using, and start earning a substantial salary.
A Possible Explanation
Each market for Enterprise IT products has most often been dominated by a single supplier. For example Cisco in networking, Oracle in databases, Microsoft in desktops and servers, HPE in servers, and so forth.
We don’t see this dominance today. HPE is still larger than Dell in server sales, but effectively it’s a market of equals. Cisco data center dominance has been shaken by Arista’s strong growth. Microsoft grapples with robust Linux competition. Oracle is losing to the cloud.
The emergence of public cloud helped disrupt the status quo. Many companies have slowed their infrastructure spending while they consider how to incorporate ‘cloud’ ideas.
It seems likely that both off- and on-premises computing will be around for some time, but infrastructure operations will be very different. The concept of ‘software defined’ has become normalized, and the major impact is in the automation and orchestration of operations.
Will there be one dominant vendor of ‘software defined’? Will one orchestrator triumph?
As I see it, a vendor needs three things to dominate an IT market:
- Low rate of change so that customers don’t move on. They can stick with what they know.
- The market need must be narrowly focused such that a single solution can cover all needs.
- The product must be broad enough in function and features to satisfy most customers but limited enough for a vendor to create and sustain.
Looking back at Enterprise IT, this proposition was true. IT managers would buy the same equipment form the same suppliers and deploy the same applications. This was called various things: ‘best practice’, ‘trusted supplier’, and ‘skills focus’.
Most assets lasted from five to ten years, with limited improvements or changes over time. Products might get faster, bigger, and a bit better but not really much different.
Public cloud overturns that pattern. Providers have rapidly moved to offer services in addition to infrastructure: identity management, bulk storage, analytics, DNS, telephony, thin clients, load balancing, databases, logging, AI systems and, of course, serverless are just a few.
This approach has been so successful that enterprises invest billions every quarter in cloud.
Diversity Of Solutions?
For the near future, I think that there will be diverse suppliers for a wide range of needs. The market keeps changing as startups bring new ideas to market.
For example, look at formal verification from Veriflow and Forward Networks, intent-based networking from Apstra, and orchestration engines like Gluware.
Of course, these platforms can grow to encompass more features and functions. For example, VMware’s Cloud Foundation is attempting to be an ‘everything cloud platform’. Today, VCF is far too unwieldy and complex for widespread usage.
I’m wondering what other people think. Will a dominant vendor emerge for network orchestration, or will diversity reign?