Moving to a single voice and data network can save
money, but it means putting all your eggs in one basket - so
checking reliability of service is crucial
With voice and data combined on one network, and the prospect of
lower IT and communications costs and greater business
efficiencies, it is easy to see why converged internet protocol
(IP) networks seem such an attractive proposition.
But moving to a converged network also has a downside. If you
get hit with a network slowdown, unplanned downtime, or a major
virus or worm attack, for instance, all the applications on the
converged network could be affected. By putting all your eggs in
one basket, you multiply the risks.
Will Cappelli, research vice-president at analyst firm Gartner,
says, "There is a significant subjective risk in migrating to a
converged network simply because we do not yet understand the
impact of large levels of converged traffic on our corporate and
public networks."
IP-based networks were designed for particular types of data
traffic, and are now carrying other media such as voice and video,
which behave in a different way to the data traffic created by
traditional business applications.
So it is essential that IT directors are aware of quality of
service issues and have a way to measure the quality of their
converged networks.
Traditional voice networks have become very robust. Data
networks, on the other hand, frequently face problems and require
additional layers of technology to ensure they remain
resilient.
In addition, organisations are increasingly running critical
business applications over these data networks, relying on, for
example, SAP, or Oracle or a commerce website. In essence, the
whole company relies on the network.
The main problem with a converged voice/data network is latency
- the time it takes for packets of data to arrive at their
destination. Half a second may not matter to someone keying
something onto a screen. But if you are using voice, that
application soon becomes unusable.
Mark Blowers, senior analyst at Butler Group, recommends that IT
directors establish some type of governing team for the network,
comprising a person or group of people responsible for measuring
quality of service, and putting in place policies and network
service level agreements.
These policies must apply to internal IT staff as well as
external service providers, and include clear lines of
communication and responsibility.
"It is important to plan this up front, at the assessment phase.
A lot of companies do not know what is on the network," he
says.
Currently, the norm is for several IT units currently share the
responsibility of measuring and monitoring service quality on
converged networks.
These groups tend to be network-based IT staff,
applications-based IT staff, and classical operations teams, who
have a larger overview of the IT system.
"These groups will have to work together for the first time in
history," says Cappelli. More often than not they have their own
tools, languages and perspectives.
"Your network guys call voice over internet protocol (VoIP) an
application, and your application guys say it is not an
application. In some ways both are right, because it lives in the
same infrastructure. But both are going to have to learn to cope
with this. And network guys have to learn how to manage multiple
applications over the network," says Cappelli.
One way to cross the divide is for IT staff to recognise that
business applications and network applications are not separate,
and that the IT department must manage the users' access to
resources.
To facilitate this, organisations will need an end-to-end team
that spans the whole IT system, says Cappelli.
In situations where the organisation has outsourced its network
to a service provider, or where multiple suppliers are responsible
for the communications, e-mail, and various business applications,
service providers will have to work together.
However, there is no sign that this is happening yet, on any
significant scale. Service providers like BT are moving into areas
where they manage the converged network and offer to monitor
certain applications, but few service providers will offer an
IP-based service portfolio that ranges from voice to SAP, for
example.
So for now it mainly falls to the IT director to ensure that his
or her team can measure service quality on the converged
network.
Fortunately, there are many tools available to do this. These
include network-level data preferment tools from the likes of Cisco
or Alcatel, which integrate with IP-based public branch exchanges
(PBXs).
Networking equipment now tends to come with a
technology/standard called quality of service (QoS), which helps to
maintain service levels at the packet level.
Tools that give you visibility over the network are also
available to enable the administrator to see how a particular
router or server is performing, how the network is functioning
overall, and where the bottlenecks are.
Tools from the likes of Network Appliance, Netscaler, Packeteer
and NetIQ, meanwhile, can help identify network bottlenecks, and
even warn of problems in advance, so you can be more proactive in
the provision of bandwidth.
However, the biggest change over the past five years has been a
shift away from tools that examine the events that take place in
the infrastructure, or deep within the application code.
The administrator or application manager has traditionally used
these tools to try to infer what the user is experiencing, but a
new breed of tools attempts to capture directly the user's
experience of factors such as network availability and response
times.
These tools are available from major network management
suppliers like Mercury Interactive, Compuware and IBM, but also
from many smaller, innovative firms. There are about 100 available
in all, either embedded into larger management portfolios, or as
expensive standalone products.
"These have emerged as being really vital to manage overall
quality," says Cappelli. But he adds, "They have their own
drawbacks, because they focus on the high-level end and symptoms,
and it can be difficult to read back to the root causes."
To plug this gap, another level of diagnostic tools has already
grown, which attempt to work backwards from the symptom to the
cause of the performance degradation.
In order to allocate and maintain bandwidth, IT directors or
network managers can take steps to reconfigure or add routers, or
purchase more bandwidth capacity from their service providers.
Many IT departments go down one of these routes, arguing that
adding more hardware is cheaper than investing in technology to
make network allocation more intelligent. However, they then face
the twin problems of manageability and mounting manageability
costs.
In addition, network service providers allow organisations to
increase their bandwidth capacity, but make it hard to reduce it
again if it is unused.
For now then, the network and application-centred tools seem the
best way to maintain service quality on converged networks.
There are innovative new approaches to tackle network
monitoring, such as the technologies developed by emerging
companies such as Bristol Technology and OpTier. These allow users
to track each and every data transaction as it makes its way
through an infrastructure.
A modern application will interact with many different software
components, and it is very difficult to see what is really slowing
down a particular transaction. The new breed of technology will
track each IT transaction, monitor its quality and dynamically
allocate more resources to the important business transactions, in
real-time, according to set policies.
Gartner sees such technologies as the way forward for measuring
service levels on a converged network.
Network quality of service - don't believe the
hype
Users should not believe the supplier hype about the network
optimisation technology quality of service (QoS), says Forrester
Research senior analyst Robert Whiteley.
QoS is designed to differentiate traffic by class of service,
which the user can define - for SAP, VoIP or e-mail traffic, for
example.
Its advocates promise QoS is standard, easy to set up, and part
of a static configuration, says Whiteley.
However, according to Whiteley, the truth is that QoS is not
necessarily standard across various suppliers' equipment, is hard
to get right and requires consistent tuning. "For multi-service
networks, QoS may be a necessary evil," says Whiteley.
"To combat the problem, limit it to only where it is needed,
such as in the wide-area network, and use optimisation appliances
like those from Juniper Networks, Orbital Data and Riverbed
Technology to avoid unnecessary QoS requirements."
Too much granularity in class of service can make the network
slow down or become hard to manage, as potentially thousands of
network devices have to be uploaded with QoS profiles, says
Whiteley.
Case study: VOIP rules the roost on lawyers
network
One firm that has opted for a managed service to monitor its
network traffic is Dickinson Wright, a US law firm.
It encountered problems with its VoIP phone system upgrade in
2004, experiencing frequent outages of long-distance and voicemail
services, which cut vital lines of communication between lawyers
and their clients.
Resolving the outages became a time-consuming, manual process for
the IT department.
The law firm implemented Compuwares Vantage tool to manage
application performance of the network from an end-user
perspective. The tool also allows the company to prioritise its IT
voice and data traffic.
Michael Kolb, CIO at Dickinson Wright, says the IT department
has set VoIP calls as the highest priority, with internet traffic
the lowest. Documents are a priority, because it can become
critical for the lawyers to send these in a timely manner.
The majority of the data replication is done out of hours.
Replication is the biggest network load. We replicate 200 databases
on a daily basis all out of hours, says Kolb.
In addition, the firm uses Cisco IP-based tools to passively
monitor the network to discover which users are accessing the
network too heavily. This links into a Microsoft active directory
to locate the bandwidth hogs by their IP address.
We can be on the phone to the user in two minutes, to tell them
to kill their stream, says Kolb.