Musings of Migration
Network migration is the process of changing the network infrastructure from its current state to a future one. Whenever the subject arises at industry conferences, everyone stands up and salutes the flag with a proclamation that, “We must have a smooth migration. In particular, we should not ignore the installed base that is not yet being migrated.” Makes sense, the service provider paid good money for the current infrastructure, and to the extent it is still doing what it was intended to do, why shouldn’t it continue providing value for years to come.
So this got me thinking. Can migration really be all that difficult that it needs a shout out. After all, haven’t we been doing this continuously since the inception of the telecommunications industry. Even just looking at the last 20 years we have managed smoothly a massive shift from circuit to packet networks, as well as large increases in fixed and mobile access bandwidth.
The next 20 years promises to be even more dramatic. Based on our view to the horizon and a bit beyond we are heading to some type of virtualized-cloudified-AI controlled network, with practically unlimited bandwidth. The lines will be blurred between networking and computing, and we will enjoy our services globally for a flat fee.
So my thinking about what is past and what is to be, lead me to classify two major categories of migration, and how they deal with legacy infrastructure.
- Forklift: Old infrastructure is placed on the garbage heap and is replaced with a new one. Customers typically get the same services as before but with better performance and features. SP's also get better operating costs. An example is replacing copper-based DSL with FTTH solutions. There is no issue with a legacy installed base because it is no longer exists. Although admittedly as different parts of the network are replaced over time, the old and the new will need to be managed in parallel during the rollout period.
- Overlay: New infrastructure is deployed for new services. All growth is on the new infrastructure while the old infrastructure is maintained for the old services. Examples are successive 2G-3G-4G radio services, or packet services replacing TDM circuit services.
This presents more of a challenge for dealing with the legacy installed base, because it could exist for an indefinite amount of time. Although in this case, that period of time is usually dictated by how lucrative those old services are, so things tend to get balanced out. In addition, means are found in some cases to support the more lucrative old services on the new infrastructure, like using circuit emulation for TDM services.
Another dimension to consider is that we are also migrating to next generation SDN-based management and control systems. This can lead to a decision between two types of “burden” when migrating with an overlay strategy, as shown in the table below. Do you maintain the relative inefficiency of current generation NMS for an extended period of time, or do you invest in creating an adaptation layer to manage the legacy infrastructure in an SDN framework.
In summary, until now the major migration decision has been whether to use a forklift or an overlay approach. We must now add another layer of decision whether to adapt any legacy installed base to next gen network management and control. The good news is that telecommunications industry has always handled migration well, and we should expect to continue doing so in the future.