Changing a network design is not something you do every day. It can cost time, money and a great deal of effort.  Often network managers shy away from a redesign until forced to do so. 

Changing the network structure means changing your physical and logical equipment and configurations, your management procedures, your management metrics and sometimes your deployment of technical staff.   If it involves new technologies, there may be an associated training requirement.   Finally, it needs to be done without disturbing existing business operations and affecting users.

Not a trivial issue.

Therefore, the answer to the question “How frequently should you change your network design?” is often “As little as possible”.

So why and when would you consider changing your network structure?

As to the why:

1. Adoption of Standards

   

   A network structure should conform to standards.  Frequently, state of the art equipment and network management software will not operate optimally in a non-standard environment.  A flat network is inefficient, not resilient and in essence, unmanageable.

   For example:

   • The network structure should be star-based in a core-distribution-access layer structure, not made of daisy-chained switches;  
   • the number of switches in a stack should be limited, as should the number of devices in a subnet; and
   • no external UTP cabling.  All external cabling must be fibre.

   An older network will need to have its network structure and design altered to meet current standards.  Some equipment, including non-manageable switches will need to be redeployed or discarded.

2. Inefficient Network Structure

   

   If, as is often the case, a network has grown over the years with no fundamental look at its structure and organisation, it is likely to be very inefficient and not able to easily utilise high-bandwidth applications, even if it is standards based.  It is likely to have firewalls and NAT that limit performance, it may also have unmanaged switches that limit the ability to manage and monitor the network.  There are also likely to be holes in network security that pass unnoticed.

3. Introduction of new technologies

   

   Some technologies need specific features in a network structure.  Networks have speeded-up from 100Mb to 1GB as standard.  Some have now adopted 10Gb as standard for high-bandwidth applications.  Routing and switching equipment must support these speeds. Older networks may need cable changes to support high-speed networks, for example to replace UTP with high-speed fibre.

   For example, VoIP and video-conferencing need specific device support:

   • Cabling.  On UTP cabling, VoIP only operates on CAT5e cabling or above. The network structure may need to be revised to replace older cabling with Cat6 or above.
   • VLAN technology.  VoIP needs VLANs.  A network design may need revision to be able to reconfigure switches to support an appropriate VLAN topology; and
   • Removal of NAT. If an organisation is using off-site Video-Conferencing through firewalls and routers, then NAT can cause problems with some technologies like H.323.

4. Running out of IPv4 Addresses

   

   A large corporate network may at some point run out of free IPv4 addresses in their available address block(s).  This is often an unexpected consequence of introducing a WiFi network to the corporate network environment.  There are several options at this point:

   • Reduce the DHCP device lease time, so that addresses are released more frequently.  This can be an effective strategy when coupled with frequent address scavenging;
   • Look at the network design to see if there are any subnets with excess address space and reallocate accordingly; and
   • Move to IPv6.  Not an easy exercise, and some legacy devices and software will not support IPv6 addressing.

5. Change of Business Use

   

   The most frequent cause of a network structure redesign is because the organisation has changed the way it operates.  A new building joins the network, an existing building is repurposed, or an educational institute adds a new campus to the core network.

   Perhaps the existing design cannot easily cope with the changes, so a new design is needed. If the change is large enough, then it will definitely be needed.

6. Business Continuity Planning

   

   Continuous availability of network services lies at the core of most business activity, particularly so for online businesses.  Continuity conscious businesses therefore have a business continuity plan in place to ensure that as far as is possible network services can continue under all circumstances. 

   Introducing backup and hot-standby sites usually requires redesign of the core-distribution level with the introduction of high-speed links, often up to 40Gb. 

   As to the when, continuous, or at least regular review.   The prudent network manager uses network metrics to look for potential trouble spots, areas of network congestion perhaps.  User feedback is another indicator of potential network problems.

   The best tools though are prediction and planning.  Planned changes of usage on the network, new usage and other potential changes need to be factored in and used to predict network performance in the future.  At some point, it will become clear that a network redesign is needed.