Five Network Design Considerations
Redundant links can be installed on the LAN and WAN. The question of whether redundant
links are required is generally guided by a combination of key factors;
a. Finance - is there sufficient money to cater for the additional equipment and any
additional operational costs for a second link?
b. Business Requirements - what is the business impact of lost network connectivity in
terms of risk and revenue?
c. Service Level Agreements - what are the existing service level agreements with carriers
and third party providers? Are the restoration times in line with business expectation?
A heavily congested link is almost as bad as a down link. The usual suspects causing
network congestion include;
a. Misconfigured or Chatty applications - some applications attempt to download large quantities
of data over a link in order to update their software or due to misconfiguration errors.
b. Viruses and worms - amongst other things, viruses and worms can launch denial of service
attacks which are designed to overwhelm network services.
c. Oversubscribed link - too many links aggregating into one point can cause congestion
in much the same way that freeways and motorways become congested during peak hour.
One mitigation against heavy congestion is to implement Quality of Service. This service
will allow you to prioritise traffic based on relevance and importance to your business.
3. Hardware limits
Routers in particular are prone to high utilization under heavy loads. This varies, of
course, depending on the router model but they all have their route processing limits and this
limit is usually well below the interface speed. In other words, if you are passing packets
between networks at high speeds the type of hardware should be a major consideration.
Other hardware considerations include the size of the routing table, oversubscription
rates, the number of features required by each device (i.e. QOS, Neflow, access lists, port
security, multicast etc) and the effects of those features on the hardware.
There is monitoring and then there is monitoring. Just knowing whether a device is up
or down is usually not enough, additional protocols that can help to determine the performance
of a router include;
a. SNMP - by far the most popular network management protocol. It can be used to graph
just about any counter including CPU utilization, memory utilization and interface utilization.
b. ICMP - still a useful protocol to determine uptime and round trip latency.
c. Net flow - a fantastic protocol that captures traffic and allows you to keep records
of conversations, volume and protocols across a particular link.
d. IPSLA - this protocol is proprietary to Cisco equipment. It is a very powerful tool
for network performance monitoring.