Most Lan problems result from a failure to plan properly or simple
human error, according to a survey by networking software supplier
Chevin.
Its main finding was that downtime or ineffectiveness in Lans is
most commonly caused by problems resulting from routers being
misconfigured at installation.
Koby Amedume, vice-president of marketing at Chevin, said, "This
happens often when there is an increase in network demand. To
accommodate more users, people throw hardware at the problem.
Installing a new router without reconfiguring others simply shifts
the problem elsewhere in the network."
Next in the list were three problems that caused unwanted traffic
on the Lan: Ethernet network interface cards which had failed but
had not been detected; broadcast storms caused either by legacy
applications on servers which had not been removed or by faulty
hardware; and unwanted protocols being broadcast, such as a print
server on an IP network sending out Unix or Mac transmissions
inappropriately.
Other common problems are caused by poorly located switches causing
bottlenecks when too many devices are allocated to run through
them, and poor maintenance of servers. A good example of this, said
Amedume, is a mail server being run at near capacity almost
permanently, causing a degradation in performance.
Additional traffic generated by undetected, unauthorised or unused
devices and network monitoring can also add to the overhead.
Amedume pointed out that RMon (Remote Monitoring) and SNMP (Simple
Network Management Protocol) technologies use agents that reside on
devices to collect information for use by network management tools.
These may be useful but can create transmissions that will slow
down the Lan.
Chevin recommends the use of the HSRMon (High Speed Remote
Monitoring) standard because it is less packet-hungry in its method
of interrogating and reporting on network status.
Amedume identified three ways of avoiding such problems. "Firstly,
an efficient network is the result of working intelligently - not
from throwing hardware at a problem," he said.
"Secondly, when you buy hardware, buy quality equipment, as many
network problems result from undetected faulty hardware. And,
thirdly, baseline your network - make the network visible and
assess your needs and likely growth trends before you go on a
buying spree."
Steve Broadhead, director of network testing company NSS, echoed
Amedume's advice. "Know exactly what is on your network, monitor it
constantly to spot problems before they arise, provide for
redundancy in the Lan to reroute traffic if necessary, and make
sure
your network management solution is not actually a bandwidth-hungry
'mismanagement' solution," he said.
"Problems - usually resulting in some kind of costly downtime -
arise regardless of what preventative methods you put in place.
Therefore, you need to have tools available that can track down any
problem on the network as quickly as possible."
Chevin's survey, published last week, was carried out among its
customers in the retail, manufacturing, transport, finance, local
government and construction sectors.
Free SNMP Reporter
Chevin is giving away a cut-down
version of its SNMP Reporter, part of its flagship Tevista
management suite.
The trial version is limited to gathering information through four
ports but Chevin said this will still give useful information,
provide performance statistics and show exactly what is running on
the Lan.
The SNMP Reporter module is available for free download from
www.chevin.com/.