Blackout! In October 2002, UUNET experienced a major
service outage practically shutting down Internet traffic on several parts of
WorldCom’s vast IP backbone network. The outage crippled many businesses that
depended on WorldCom as their sole backbone provider. The outage also hit
service providers and enterprises with multiple transport contracts who found
their IP traffic severely restricted. Experts diagnosed that the problem was
caused by software upgrades WorldCom performed on its border routers - the
routers at all peering points and hosting locations. They contended that
WorldCom had bought some bad software from a vendor, and when the software went
live - thousands of routers went out of operation.
|
The Slammer Punch: At the time of writing, the Slammer worm was causing havoc
in networks across the world. The worm shut down ATMs of the Bank of America and
caused problems resulting in delays at Continental Airlines in the US. Unlike
most worms and viruses, the "SQL Slammer" did not rely on e-mail as a
carrier–instead it spread directly through networks….
Network management means different things to different people. At the most
basic, it is the management of a company’s local area and wide area networks–including
switches, hubs, routers, VSATs and the software that goes with them.
A broader definition involves managing everything on the network using
separate tools for each network element. For instance, CiscoWorks to manage
Cisco routers and switches and Insight Manager to manage HP/Compaq servers and
desktops. It includes...
n  Server
management
n  Desktop
management
n  Mail
management
n  Security
management
n  Back
up management
n  LAN
management
n  WAN
management
n  Storage
management
n  Inventory
management
n  Software
management: distribution, inventory and licensing
n  Application
management
n  Call
management
The different services from network management service providers are: n  Onsite |
In its broadest sense, network management is the management of everything on
a corporate network through an integrated enterprise management platform...a
level at which NM essentially becomes enterprise management.
Network management solutions can, therefore, range from point solutions from
companies like Concord Communications, RiverSoft to network/enterprise
management platforms like HP OpenView, CA’s Unicenter, IBM’s Tivoli and Sun
Solstice.
The fundamentals: Monitoring, upgrades and data
Whatever the scope or definition, the basis of all network management is the
monitoring of the network, configuration changes and upgrades when required and
possible and finally, providing of data for any design changes that may be
necessary for improved performance.
Network Integrators |
n Wipro Infotech |
n Datacraft India |
n HCL Comnet |
n CMC |
n IBM Global |
n HCL Insys |
n Network Solutions |
n ARM |
n Tata Infotech |
n Ramco Systems |
Monitoring is done in many ways: Through replies to specific queries
through SNMP
(Simple Network Management Protocol) — which is a language used by all
devices to talk to each other; setting up of alerts or thresholds inside a
device so that an alert is sent out through SNMP whenever that threshold is
crossed; or through setting up devices to report specific statistics
periodically.
A key thing to remember however is that network management is not network
design or implementation. It is not even about actual network maintenance. Any
network management software can only manage a network (monitor and make some
upgrades) that has already been designed and implemented. It can provide data
that can help improve that design — but the new design itself is generally not
in the purview of an NMS. Nor are things like a hub conking out. An NMS can tell
the administrator that a hub is not working, can help rectify it to the extent
that it is an issue with rebooting, configuration or a new OS upgrade. But if
the problem with the hub is physical — it is the realm of network maintenance,
not management.
In fact the International Organization for Standardization (ISO) model,
popularly called the FCAPS model defines five functional areas of network
management. They are:
n Fault
Management that involves the detection, isolation, notification, and correction
of faults encountered in the network.
n Configuration
Management involves the configuration aspects of network devices such as
configuration file management, inventory management, and software management.
n Accounting
Management looks at usage information of network resources.
n Performance
Management deals with monitoring and measurement of various aspects of
performance so that overall performance can be sustained.
n Security
Management aims at providing access to network devices and resources to
authorized individuals and entities.
|
Different needs, different concerns
The criticality of network management is no longer in dispute. What often
becomes the issue is identifying which component of network management is key
and critical to any organization. Historically, various aspects of network
management have been over-riding issues. In the mid-90s LAN management was the
big concern when networks operated over 10 Mbps shared LAN segments. When LANs
became switched LANs and got upgraded to fast Ethernet 100 Mbps LANs, they were
no longer the key bottleneck and therefore ceased to be so crucial from a NM
point of view.
Then with the software and enterprise networking booms, WAN management and
monitoring of VSAT links became crucial. For one, ERP software like SAP, Baan,
PeopleSoft, JD Edwards etc usually ran on a central server. For another, WAN
links cost the moon. So WAN management and keeping that link un-choked became
crucial. Now of course other options like leased lines are available.
NM Vendors |
n Alcatel |
n Compaq |
n Compuware |
n Concord |
n Ericsson |
n HCL Comnet |
n IBM Tivoli |
n HP |
n Lucent |
n Microsoft |
n NetScout |
n Network Associates |
n Seagate |
n Sun |
Either way, bandwidth management has been and will continue to be a key issue
in network management. From verticals like banking which deal with complex wide
area networks that have to constantly deal with disparate customer profiles,
heavy loads and time issues to the now booming call center industry that is
dependent on the bandwidth of its WAN links and its utilization.
In addition, different industries will have specific requirements that are
critical to their business needs. Storage management could be key in a software
development house that uses diverse platforms for instance. Other things like
the profile of applications will also determine the critical management areas.
Some like Telnet will be very time sensitive while others like CAD/CAM might be
resource hungry.
The moral of which story is — one size doesn’t fit all. Different
enterprises in different verticals and different stages of evolution will have
different needs.
NM Solutions |
n CA’s Unicenter TNG |
n CiscoWorks |
n Concord’s eHealth |
n Enterasys Netsight |
n HP’s OpenView |
n IBM Tivoli |
n Lucent’s VitalNet |
n Nortel’s Preside |
n Novell’s ZenWorks |
Finding the right NM solution
It’s a bit like building a house. One would ideally like it to look
elegant, take care of all possible contingencies, plan for growth, build it once
and hope it lasts forever. That ideal house however would probably look like a
palace, take years to build and cost a bomb. In the real world however where
most enterprises have both time and cost constraints here are some key pointers
to finding the right NMS for you and making it work
n  Design
Carefully: You cannot manage a bad design. A network management solution —
whether driven by point tools or an enterprise suite — is only as affective as
the network itself. So pay attention at the design stage. This might sound like
a no-brainer but it is amazing the number of things that can be overlooked at
this stage. Here are two examples:
  l  A
few years ago a hub kept crashing at an MNC milk factory in India. The network
administrator looked at the design, found nothing wrong and could not figure out
why the problem was recurring. Till he went down to the factory. Sure enough,
the hub was placed 12 feet above ground level and should have been safe from
normal physical damage. Except, he found out that everyday at the end of the
shift, employees sprayed water all over the walls and the roof of the factory to
wash off the milk. In the process, they also liberally sprayed the router with
water.
  l  When
a healthcare company in India decided to set up a new office a while ago, it
chose to use different contractors for what it thought were disparate works.
First came the building contractors who put up the structure. Then came the
electrical contractors who dug up the whole place to put in the electrical
wiring and left. After those holes were fixed came the network contractors who
re-dug the whole place to put in network connections. The place looked like a
bomb raid site and a job that should have been finished in a few months took
over a year.
n  Standardize:
Managing different systems running different software and a network with a
combination of different switches and hubs can be a nightmare. For one,
integration issues have to be dealt with. For another, not all network
management solutions work with all vendor hardware equally. Some solutions may
be great with Sun systems but quite below par on say IBM Mainframes. Finally,
rollout of upgrades could be a trip to hell. A simple Office XP upgrade
requiring minimum hardware and software specs for instance could end up being a
major exercise if you have vendor desktops thrown in with assembled systems and
all variations of operating systems. At the very least, have large bunches of
similar systems.
n  If
it ain’t broken, don't fix it: A standard management mantra that
also works with networks. Identify pain points and strongly consider whether you
want an entire Enterprise Management platform or just a few point tools to fix
two or three issues. For instance, in a medium sized enterprise, the systems and
anti virus software might be working just fine. But there may be issues with the
LAN/WAN traffic. An EM platform can be both time consuming and expensive by
itself. Add implementation costs and you might end up buying a spade to kill an
ant. If the problem areas are limited, it’s possibly a good idea to look for
point solutions. They can handle specific problems far better, are cheaper and
scalable.
n  Plan
the deployment: If you decide to go with an enterprise suite plan the
deployment to address the pain areas first. Network management issues will
simply pile up otherwise. If the big pain areas were storage and security
management, deal with those before you start looking at bandwidth management
issues.
TV Mahalingam in
BangaloreWith inputs from Sarita Rani
Network Management Best Practices
n  Be
proactive and cost effective: If you can see in advance where things are most
likely to go wrong, and put measures in place before anything does, savings in
network and lost productivity costs is likely to be high.
n  Look
before you leap: Research into network management alternatives before pursuing
solutions. Spend time in researching and identifying several key goals that you
want to achieve with its network management. These goals will point the way to
specific hardware and software solutions that are furnished by vendors.
n  Use the
magnifying lens: Carefully evaluate both the vendor and the product. Invite the
vendor for a product demo at your site. If the vendor cannot successfully deploy
his product in a reasonable timeframe, its time to say —"next"
n  Take care
of the pennies: Always look for a compelling return on investment (ROI) from the
solution that you choose. When you combine cost effectiveness with effectiveness
of network management, you have a solid combination.
n  Keep an
eye out for the past: Many network management tools that are commercially
available in the marketplace today are real-time only. They do not offer a
historical analysis. If you incorporate tools in your network management
practice that gives you both present and past performance, it allows you to
perform analyses, and to be proactive with network problems.
n  End to
end solutions: Consider investing in ‘end-to-end’ diagnosis tools that are
commercially available, which will enable enterprise CIOs to look into the
entire chain of network and application components when you trouble-shoot a
problem.
n  Cross-function:
When multiple network and applications components are involved in an
enterprise-wide environment, it is often necessary for you to work together
effectively with individuals from other departments. Explore setting up
cross-functional teams with cooperative goals.
n  Watchout:
Always stay in touch with vendors and security solutions companies to know more
about security threats like viruses and worms and the latest trends in network
security.
FCAPS Best Practices
Fault management
n Enable
Traps on a consistent basis on network devices.
n It is
important to check and update the configuration files of traps to ensure the
proper decoding of traps.
n Conduct
a periodic review of configured traps, which will ensure effective fault
detection in the network.
n Consider
an event management system that is capable of correlating different network
events.
Capability management
n Have
naming conventions for network devices, starting from device names to individual
interface as part of the configuration standard. A well-defined naming
convention provides administrators with the ability to provide accurate
information while troubleshooting.
n When
adding new configuration commands on existing network devices needs, verify
commands for integrity before implementation.
n Changes
to software in network devices should be tracked to assist in analysis when
another software maintenance is required.
Accounting management
n Measure
the utilization of all important network resources. Analysis of the data
provides insight into current usage patterns.
n A
usage-based accounting and billing system is a prerequisite of any SLA. It helps
in defining obligations under an SLA.
Performance management
n Document
the network management business objectives.
n Create
detailed service level objectives.
n Spend
a lot of time on drafting the performance criteria’s in the SLA. An effective
SLA solution requires three components to be implemented:
l comprehensive,
enterprise-wide data collection
l measurement
and intuitive reporting
n Collect
a list of the variables for the baseline, including details like polling
interval, network management overhead incurred, possible trigger thresholds.
n Have
a periodic meeting to review the analysis of the baseline & trends.
Security management
n Create
a platform-specific minimum configuration standard for all routers and switches
that follow industry best practices for security and performance
n Enable
security features that ensure that only authorized management stations are
allowed to perform changes on network devices