Platform concepts are hot in tech these days, and it’s not surprising that there’s an interest in a unified platform for network operations management, both among enterprises and network operators. There are a lot of drivers to the movement, and enterprises and operators both see issues as well.
We actually used to have a fairly broadly accepted “platform” approach to network management. HP OpenView was the baseline for many enterprises, comprising a foundation for 58% of enterprise network management tools according to my 2006 survey. However, HP rebranded it, and then it was sold off with much of HP software, and today it barely gets a mention. Given current interest, and given HPE’s pending acquisition of Juniper, I wonder if they wish they’d kept OpenView!
What likely led to HP’s decision is one of the drivers today, ironically. In the early part of this century, enterprise networking transformed from devices and trunks to VPNs, and so (in the WAN) “network management” really became service management, with a related service management tool becoming WAN management. Obviously we have largely completed the VPN transformation of the WAN, but virtual networking shifts are still underway. SD-WAN is adding another system to the WAN, one that’s functionally an extension of the traditional MPLS VPN, but which is supported by different equipment, often equipment purchased by the enterprise. Many enterprises build MPLS VPNs from multiple suppliers, get SD-WAN from managed service suppliers, and so a service transformation of the WAN doesn’t mean that management is centralized.
A second driver in NMS platformization is the componentization of software and the resulting complexity in deployment and redeployment. Anyone familiar with DevOps and container orchestration knows that most hosting configurations now use virtual networks to link the pieces of applications, and these now contribute to the workflows associated with user experiences. A complete view of a user connection now involves the data center LAN and the configuration of those virtual networks. At a deeper level, the componentization of applications and interconnection of the resources assigned them from the data center pool means that application component state is harder to determine and to distinguish from network state.
Then there’s the cloud, which magnifies all the issues raised by software componentization by adding a resource-as-a-service model that has its own internal network. Not only that, agents of a VPN, via SD-WAN and SASE, extend the enterprise VPN into the cloud, making the cloud network in part an element in the VPN at the same time as it’s an independent and opaque network of its own.
The multiplication we’re seeing in network elements is probably enough in itself to drive interest in network management platforms, but the drivers are also broadening the issue set. The real goal of network management isn’t to assure the network, but to assure what’s on it by addressing issues that are out of scope for IT management. The drivers we’re discussing complicate IT management, in that they introduce concepts like network-connected virtual resources and the notion of a “host” for applications as a web of resources that can span technology, geography, and even ownership. A more universal NMS platform either misses these issues, kicking a major management challenge can down the road, or addresses them and then becomes something beyond NMS. So is NMS obsolete?
Not likely for network operators, of course. For enterprises, it’s possible. At least, what we may be seeing is that both network and IT management need to be subducted into some master management platform. Call it IT or network management, or perhaps application or experience management, but it’s really an accommodation to virtualization. At the network level, virtualization creates a network service to applications/experiences that shares physical infrastructure and hides both other tenants and the properties of the infrastructure. At the hosting level, it creates various kinds of virtual hosts whose realization is drawn from a pool of real resources, and whose creation is surely going to end up creating a virtual network need.
But what does this mean for NMS or management overall? Some enterprises have suggested that we need to divide management into a “real network” piece, a “real hosting resource” piece, and a “virtualized union” piece that presents an application/experience view of the virtualized resources on both the network and IT side. Others don’t like this approach, saying that divided management only increases the chances of errors, not to mention raising costs.
A few vendors, cloud providers, and MSPs have suggested that the acronym “MSP” be redefined to mean “management services provider” and that the management service being provided be one that unifies management of real and virtual, links the combination to applications and experiences, and offers a variety of options ranging from facilitating user control through guaranteeing SLAs as currently is often the case with managed services.
I don’t think there’s a consensus here. Enterprises may be reluctant to cede management platform control to a third party for security reasons. There is already an investment, in tools for management and staff training and procedures, in IT and network management as largely independent elements. Virtual networking today tends to be integrated with one or both of these, and breaking the current relationships, even without effectively promoting it to a senior role, is likely to be difficult.
Difficult, but I think it’s inevitable. Virtualization changes everything, disconnecting IT overall from a fixed real world to a malleable virtual one. We’ve recognized this shift implicitly, and we’ve created technologies and features to address it and to manage it, but we’ve never really addressed the profound core of the challenges it presents. Even a lot of the issues being raised in security demand a virtual-domain solution, difficult when we don’t explicitly recognize the existence of a virtual domain.
What this says about NMS plaformization, I think, is that not only is it not likely to bring about hoped-for simplification of user or operator problems, but also that it may well do more harm than good. Investment in a new network management platform solidifies the separation of management overall, and also the separation of elements of the emerging virtual domain that we need to be facing now, and will need to face even more in the near future.
