There is one word that crops up at every network
andtraffic managementevent I speak at.
That word is "visibility" or, all too often, the lack of
it.
For example, at seminars I have presented at, it has been
revealed that a typical enterprise
wide area network (Wan) has between 55 and 70 applications
running across it concurrently, when the network manager assumed
there were closer to 15 or 20.
At one such event, David Foster, head of the communications and
networking group at
Cern,
said visibility was a real issue.
"Today's distributed applications require management systems to
have an end-to-end view. Without visibility across network domains
involved in the end-to-end path, it becomes difficult and expensive
to optimise resources, perform capacity planning and isolate
faults," he said.
Cern is no ordinary company. It is the world's largest
laboratory dedicated to particle physics research in pursuit of the
understanding of the nature of matter and the forces of nature -
heavy stuff that needs to be carefully controlled.
If Cern's network has visibility issues, what does that say for
more ordinary networks? Regardless of the prestige and size of the
firm, the problems are the same. If you cannot see what is going
on, how do you control it?
It is one thing to know exactly which devices are connected to a
network - and often the IT department does not even know that for
sure - but it is another thing altogether to know what traffic is
running across the Wan or internet out of your office and which
users are using which services and applications.
So, what kinds of traffic are proving to be invisible to network
managers? The answer is various forms of chat, peer-to-peer traffic
and instant messaging. Often these are applications that users have
installed on their laptops and brought into work and continued to
use.
A whole host of unknown protocols, applications and services are
creating performance issues and making a mockery of the
quality-of-service parameters the network manager has put into
place. All this because of traffic - and lots of it - that network
managers are largely unaware of.
If this sounds like a major problem emerging, that is because it
is. It is problem enough on a Lan, where there is bandwidth
aplenty, but it is even more of a problem on a Wan, and especially
on internet connections, where bandwidth is expensive and in short
supply.
It therefore needs to be optimised to the nth degree, where n is
as close to infinity as possible. The problem with this splendidly
obvious theory is how to put it into practice. There are simply so
many elements and areas to cover that it can quickly become
overwhelming. Traffic loads, traffic types, prioritisation issues,
user behaviour control, Wan monitoring - that visibility thing -
and management, proactive security, general reporting, trend
analysis and so on.
Traffic shaping: a new approach
So what is the answer? How do we improve our visibility? One
answer lies in the form of the EP range of traffic-management and
optimisation devices I have been looking at from
DBam Systems, a start-up
based in the North of England - Wakefield, my home town to be
precise, and now known as the traffic management capital of the
universe.
During the course of testing the DBam product, we concluded that
it is not a case of a single element being sufficient to control a
wide area network and its users. It is the combination of elements,
which is precisely what DBam offers, that is key. These elements
are as follows:
● Being able to see everything that is going over the network -
from layer two to layer seven.
● Being able to optimise what needs optimising and ensure that
bandwidth is available on an as-required basis, but never having it
nailed-up and unused.
● Being able to control users and groups of users in terms of
what applications, services and protocols they are allowed to use
to as finite a degree as possible, both in automated, timetabled
and real-time situations.
● Being able to analyse the traffic to as finite a degree as
possible, drilling up and down the traffic information - think of
it as a tree - and focusing on exactly which elements are
problematic.
● Having extensive reporting options so that all the captured,
real-time snapshot data can be presented in a format that any level
of user, from administrator to board director, can make use of.
Of course, another benefit is that you can simply speed up the
transfer of data across a wide area network or the internet, rather
than throwing expensive bandwidth at it. Even if you do the latter,
the inherent latencies of the internet will render much of that
money spent on bandwidth invalid.
So, faster is better, rather than bigger in this case. And here
is where the DBam guys have really excelled. Forget everything you
think you know about
data compression, caching-based acceleration and the related
levels of expectation.
Instead, think of the whole concept of data acceleration being
re-invented from the ground up. Forget compression ratios of 10:1
and even 50:1 - they are history.
During extensive testing of the DBam technology, where we tried
every which way to break it, instead all we proved was that these
boys have made a very significant technological breakthrough. MP3
files cannot be compressed, right? Wrong. We got 10,000:1
acceleration with all compressed and uncompressed files.
No, that is not a misprint. Nor is it April Fools' Day. Seeing
is believing.
● Steve Broadhead is founder and director of
Broadband-Testing Labs
Interview with Shane O'Hanlon, chief technology officer
at DBam Systems
What do you see as the main problems network managers are
encountering in trying to manage and maximise their wide area
network and internet connections?
"Fundamentally, the problems with maximising performance stem
from application design and TCP protocol stack implementations in
modern operating systems.
"The inability to conclude if performance issues are network
utilisation, network quality or application faults causes
administrators to be persistently questioning the core of business
communications.
"The first step in any application-performance enhancement
project is to analyse the network and reach a conclusion as to the
cause of application performance degradation.
"We still see cases of customers installing bigger, fatter wide
area network connections in an attempt to resolve issues. Is this
bandwidth-oriented approach the right solution?
"For a small group of applications, purchasing increased
throughput capacity is a cost-effective solution. For the majority
of applications it is not. Simply stated, this is because in one
form or another applications ensure the sender or receiver of data
acknowledges the receipt of data.
"This limits the amount of data that can be in transit between
sender and receiver, and purchasing more bandwidth does not reduce
the distance between sender and receiver.
"Another limiting factor to this approach is that modern
operating systems cannot receive significant benefit from increased
network capacity without some form of TCP and application tuning,
which is a very difficult process to put into place."
Who will benefit most from the DBam data acceleration
technology, and in what ways?
"We hope developing countries will use the technology to enhance
communications that would otherwise be limited by infrastructure
cost. Also, homeworking with enhanced access to business
applications is a benefit this technology will afford users.
"I believe our technology will form an inherent part of a
paradigm shift in IP communications, utilising the advances in
computational powers of computers to increase the performance of
applications across the wide area network."
Is this acceleration technology applicable to streaming and
broadcast traffic, such as video-over protocols like Mpeg-2 and
Mpeg-4, as well as static files?
"Yes, we have successfully increased the quality of Mpeg streams
for a number of customers. A variety of techniques are used to
transport the data, enabling acceleration on not just applications
that use TCP, but also UDP and others."