Network Management Overload, Rerouting Traffic

The network is down

KNG Infoway, an Internet service provider providing connectivity to both
home and corpo rate users, faced a serious problem last August. Its network was
down for several hours before the problem could be rectified. This downtime took
its toll on KNG. Several dissatisfied corporates switched over to a competitor.

Houston, we have a problem

The National Stock Exchange, which handles transactions worth hundreds of
crores daily, went down for three days due to a connectivity glitch when its
INSAT-2D satellite went down. NSE resumed operations within three days, but not
before its transaction losses swelled to more than Rs 1,000 crore.

Painting the map red

It was September, not just a week after 9-11. A large multi-national company
with a nationwide network in India was one of the hundreds hit by the Code Red
virus. At their net-ops center, engineers watched as dozens of lines turned from
green to red, on the projection of a map of India, as local offices took routers
offline for "cleapup", or the link simply went down anyway. The damage
at the end of the day–many hours of downtime, including at call centers across
the company’s network, and substantial loss of revenue.

CIOs today face the challenge of not only ensuring 24x7 connectivity to end
users, but also addressing the ever-rising thirst for information. Integrating
these with the business goals of the organization is yet another challenge.
Optimum utilization of resources is another issue. While some part of the
network may be overutilized, others may not be used to their full.

It becomes imperative, therefore, for CIOs to choose products and system
integration services that offer sufficient support and stability, as well as
ensure healthy returns on investment. Network management strategy is now driven
by the need to minimize network failure, improve service quality, reduce
operating costs, and enable new business opportunities. IT departments in
enterprises have begun to service other departments as clients, and are hence
accountable for the quality of service rendered. In such cases, huge investments
made during network deployment also have to be justified.

Routing and switching technologies are adapting to market needs, but advances
in network capacity and performance also complicate the tasks of design and
implementation and the need to optimize becomes a critical issue. The advent of
large switched networks adds another layer of complexity to network design and
provisioning for routed, switched and hybrid networks. It also raises some
burning questions for the enterprise. Is it possible to achieve high
performance, optimal dedicated bandwidth, better utilization of available
address space and a larger broadcast domain all at once? How do we incorporate
emerging technologies into existing hardware? What are the administrative
burdens associated with different solutions? How does one prepare for future
growth?

One size doesn’t fit all!

As enterprises seek to connect geographically distant locations, WAN, a ‘one
size fits all’ approach, will not work. The process of network planning
depends on the number of users to be connected and typically involves
documentation of business requirements and development of strategic network
architecture. One needs to prepare capacity plans for the physical network,
including bandwidth allocations for protocol and applications.

A thumb rule followed by many consultants in the field is to ensure that the
investment in network management is about 10% of the total network investment.
This percentage could be higher for smaller networks that require a higher
degree of management.

Customers using network management applications are realizing direct
cost-savings and IT personnel productivity gains–immediately. Network
management is not just the icing on the cake, as it can reduce bottomline
expenses and give network administrators huge efficiency savings.

A tool to suit every enterprise need

There are various kinds of tools available in this area, including simple
SNMP-based products which do a great job of monitoring at a generic level. As
against this, vendor-specific tools have a significant impact on improving
availability and predictability. These tools actually go far beyond generic
monitoring to provide the ability to keep the operating system on the device up
to date (all relevant patches applied). They monitor vendor specific information
like the ability to detect changes to the memory on the device and changes in
configuration, and the like. Another piece that gets to be important is the
ability to audit changes occurring in the network along with a trail to the
person making the changes.

Several organizations use ‘free’ software for some of their network
management needs. This is definitely an option. However, you need to be careful
on the availability of continued support and enhancements on the tools you
choose. Many tools today are ‘open source’ and supported by their users.
These are safer options. Many other organizations develop their own applications
and scripts to manage their networks.

The crucial SLA

A network management platform deployed in the enterprise manages an
infrastructure that consists of multi-vendor network elements. The platform
receives and processes events from network elements in the network. Events from
servers and other critical resources can also be forwarded to a management
platform. Network management platforms such as HP OpenView, Computer Associates’
Unicenter, and Sun Solstice can perform a discovery of network devices. Each
network device is represented by a graphical element on the management platform’s
console. Network devices can be configured to send notifications called SNMP
traps to network management platforms.

Service-level agreements are critical

Ensuring stable vendor service and support are major issues and are
important factors in the CIOs’ decision-making process. Network equipment
suppliers should be made to provide local services for replacement of parts,
global standards for on-site services through qualified support personnel.
Without this, enterprises will need to factor in unplanned downtime, which no
management will accept, as it fails to justify the investment in networked
operations. In cases of extremely large network designs, a pilot implementation
is advisable so that the features and functionality indicated by vendors are
proven.

A service-level agreement is a critical criteria to measure performance. In
the SLA model, the IT department becomes a service provider within the
enterprise, and end-users become the consumers. An effective SLA solution
requires three components:

• Enterprise-wide data collection;

• Measurement and intuitive reporting; and

• Ability to enforce the determined SLAs.

There are many options for the CIO when it comes to the
purchase of network management products. The capabilities of the products range
from monitoring to actual management (which comes with the ability to actually
make changes to the configurations and software on the device, among other
facets). Network management products could be available for free or else, the
cost could run into lakhs of rupees. If the right products are used in an
effective manner, the investment pays for itself very quickly.

Team DQ

Best Practices

Work out your needs and then double your estimates! No matter what you think you’ll
need, you’ll require more...

• Decide on a family of networking
  products and then try to keep it in the family. Although today’s
  network products conform to many protocols that promote
  interoperability, there are still components such as management
  software, that might be proprietary.

• Pay close attention to the physical
  environment in the space you designate for your networking equipment.
  Water and dust are intolerable. Heat is marginally tolerable. You
  should also plan to control physical access to the space to maintain
  security.

• There is no problem in buying
  networking equipment online. In fact, the high degree of
  standardization and interoperability makes selecting equipment this
  way easy.

• Suggest solutions to your value-added
  reseller, but then listen for a reply. Working with your consultant is
  important when you select a family of products or design the
  infrastructure. No matter how experienced you are, the VAR can usually
  teach you something.

• There is a whole art to licensing the
  software used on networks. Software companies offer a variety of
  plans. Buying multiple copies of programs with per-CPU licenses is
  usually the wrong thing to do. You can often negotiate site licenses,
  licenses based on the number of active users, limited-time licenses,
  or other types. Software licenses can get expensive as you buy more
  seats.

• Build a good toolbox of handy devices
  and utilities. Practical things to have in your networking toolbox
  include a cable tester, packet decoding software, a kit to crimp on
  cable connectors, a variety of replacement and patch cables, and
  cable-labeling accessories.

• Create a good plan for data backup
  and stick to it. Traditional data backup plans call for taking a
  snapshot of the system at one time of day, saving that snapshot on
  removable media, and moving that media to a separate physical
  location.

• Create a plan for remote access by
  employees. Organizations of all kinds must accommodate work-at-home
  employees and road warriors. The old approach was to use a dial-up
  access server. The modern approach is to use a virtual private network
  (VPN). In networks with a few dozen simultaneous remote callers, the
  VPN server can be part of a corporate firewall.

• Consider the role of wireless in your
  local network. Wireless can extend or replace cable. Wireless local
  networks require more support, but offer great flexibility.