The Five “Laws” of IP Planning
The Internet Protocol (specifically IPv4) is now being used for virtually all business and personal digital communications, from data centres to telephones to industrial control systems. More recently, the Internet protocols have been adopted for 4G/LTE mobile systems and are a critical enabler for the Internet of Things. By the end of the decade, IP addresses will be embedded in many billions of devices (some say 50 Billion!).
Planning and managing the deployment and evolution of IP networks, and IP network addressing in particular, has never been more challenging and can even be business critical. Multiple levels of planning are required - everything from the address structure in the protocol itself to the use of DHCP for assigning addresses to specific devices, virtual machines, containers and even the different applications in a company.
Overall network plans should cover a variety of topics, including:
Services to be provided – identification of the types of traffic that will be carried over the network (this is called the “network direction”);
Virtualization – division of network resources into subnets and logical networks that match the business and its traffic requirements (i.e., a software-based divide and conquer approach);
IP Address Management (IPAM) – logical allocation of addresses in an efficient manner, including combined use of IPv4, IPv6 and private addresses;
Access, resource and traffic security – protection of the physical and logical networks from external threats, unintended access, policy violations, and other attacks;
Quality of service and performance settings - optimization of performance and explicitly managed QoS for sensitive applications and for “pay as you go” environments; and
Management, disaster recovery and monitoring – designing for continuous service and operations management, especially for networks with self-service provisioning and external compliance controls.
Let’s take a closer look at the third item – IP Address Management. As a starting point, we suggest the following five “laws” for IP address planning:
First, commit to creating a plan and get it done
Growth and change in IP networks often happens on an ad hoc basis, with new subnets assigned as use outgrows capacity or when new services (such as an on-site private cloud) require entirely new subnets. Many organizations do not have an explicit addressing plan and do not seek to use best practices, which results in a less than meaningful addressing scheme. This leads to:
An inability to easily differentiate security policies and quality of service on a per application basis;
Applications that span multiple subnets, creating additional processing overhead from network ‘tromboning’; and
Unnecessary overhead for managing the network and the services it delivers.
Any major project, such as a data centre migration or adoption of a cloud service, can be an opportunity to re-visit the planning assumptions and identify any new traffic types and patterns.
Next, establish addressing strategies
An initial step in IP address planning is to establish the overall structure and assignment approach.
For example, a “divide and conquer” strategy is one of the simplest approaches - take whatever address range you are working with and divide it in two (e.g., user-facing and network management), then further divide these into organization-appropriate sections (e.g., services, sites, monitoring, backup, etc.). This could form a multi-level hierarchy if it makes sense for the organization.
It is important to build in sufficient flexibility to allow for business growth and to accommodate changes in the network nodes and links (i.e., the overall topology). The eventual migration to IPv6 should also be factored into the plan and roadmap.
Contact LightMesh if you need help getting started with a practical approach to IP address planning.
Third, document your network directions
Each subnet of your network should have a defined “network direction” which specifies its intended use. Network directions would include User-Facing (a.k.a. Front-end, or Services), Management, Backup, KVM/iLO/Remote control, vMotion, and others. These should be closely aligned with the address division process described above.
Each network direction may have differing quality and performance requirements but mostly it’s important to identify them for implementing programmatic access to the network design and topology. You will help with network automation and orchestration by including the network directions in your plan!
Once you have a plan in place, don’t hide it away or ignore it. Get feedback from everyone.
Network plans should be available to anyone with a need to know, especially network engineers, solution designers, network support personnel and, when appropriate, external partners.
The latest updates to the IP plan should be readily accessible and easily understood by those involved in any aspect of network operation and maintenance.
A network architecture that includes a IP address plan will yield greater development and operational efficiency, but only if done in a logical and expandable way.
Finally, keep track of all changes
It almost goes without saying that changes to the plan will occur, probably faster than you would like, and that not tracking these changes will eventually make your plan worthless. Changes in technology, upgrades to equipment, datacenter migrations and consolidations, server virtualization, even mergers and acquisitions will all affect your plan.
A good change management process, rigorously applied, should definitely be treated as mandatory.
It is important to note that keeping track of addresses can be a major challenge by itself. Support tools, such as the LightMesh CMDB and the LightMesh IPAM application, would be essential for tracking any diverse and rapidly changing network configuration.
An IP planning maturity model
Despite the importance of network operations and maintenance, many organizations do not have well-defined network planning or design processes. Explicit network plans may either be non-existent or, at best, be reactively developed and poorly documented.
A simple maturity model for network planning could be used as a baseline. For example:
Non-existent – no formal planning is being done;
Reactive on demand – network plans are documented only when needed for new projects or for troubleshooting, but very little proactive planning is performed;
Repeatable procedures – specific planning tasks, including IP address allocation, are somewhat formalized and generally repeatable, but they may not be automated or managed;
Life cycle IP planning process – a complete address planning, monitoring and documentation process is available, is being used consistently, and is producing measureable results; and
Continuous improvement – tools and systems are in place to support and automate the planning process; IP address management is integrated into an overall IT Service Management process.
For an example of how an IP address planning tool could help your organization become more mature, give the LightMesh Subnet Builder a test run.