Only A Test...
As the Internet and convergence penetrate of our daily lives, the buzz
of applications increases: e-commerce, wireless connectivity,
personalization of everything from marketing pieces to Web sites to HTTP
404 errors, convergence of things once thought un-convergable. For most
people, convergence means the usable layer -- the interfaces and
applications that they deal with everyday. But for a few people behind the
scenes, convergence means something else. Put bluntly, it means a testing
nightmare.
Testing used to be so simple. I remember in high school shop when I
first used an oscilloscope. Up to that point I had used -- okay, I admit
it... I owned! -- an ohmmeter/voltmeter combo and a few other testing
accessories. But I had never imagined what a 'scope might be able to tell
you about the goings-on inside of a circuit.
Well, the days when you only needed a 'scope to troubleshoot are long
gone. In fact, specific applications have always required specific testing
equipment, but a generalist like me could get by with a few tools like a
'scope and an ohmmeter. Enter convergence. Enter the call
center-enterprise-e-commerce-DIY-online-auction-all you-can-eat super
bank-market-site-e-mail-service-provider thing (aka, the dot-com). And
enter their network. Now, try to test it.
Testing a converged network of any scale requires a wealth of specialty
knowledge well beyond what might be reasonably expected of any individual
testing person. Maybe even beyond the pale of a decent MIS department. So
what is the poor business person to do? His salespeople want to access
their e-mail from the road. Her agents must handle Web-based IVR, e-mail,
VoIP chat, and traditional phone calls transparently. But somehow things
aren't working. And figuring out why seems tough -- even tougher than
programming a VCR, or figuring out how to turn off that annoying pop-up
clipboard in Word 2000.
Well, at the heart of all of this are two things: convergence, and
testing to insure convergence is actually taking place. To paraphrase
Shakespeare: "Convergence is not convergence which alters when it
another converged application finds."
TOWARD A TESTING STRATEGY
As I said, testing the converged network requires knowledge. In a way,
that's exactly what testing products are. It's like you're buying a
hardened, productized form of knowledge. Knowledge of your network that
can be passed from user to user, that doesn't leave when another company
offers it better stock options.
A while ago, we ran some articles on testing that discussed the need
for businesses of all sorts to adopt more holistic policies involving
real-life and life cycle testing. Now, to get a feel for how users put
together their testing strategies, I asked a few people in the industry
whether these earlier cries have fallen on deaf ears. The response was
encouraging.
Brian Miller, manager of the contact center testing and monitoring
group with Empirix, had this to say: "There has been high interest in
automated quality monitoring solutions that are testing the system in the
same way that a customer would use it. Traditional monitoring solutions
won't necessarily identify performance problems that directly affect the
customer experience (back end slow downs, incorrect or garbled voice
prompts, etc.)."
As CTI technologies become more of a "must have" than a
"maybe," end-to-end testing is increasingly required. This is
not to say that end-to-end testing is easy to achieve, or even fully
possible in every situation. As Brian Mason, VoIP product manager for
SmartBits Product Line of Spirent Communications, points out, there are
two main hurdles to testing in this manner:
"The first hurdle is that the complexity of large networks today
makes it hard to test the entire network from end-to-end. Once a problem
is identified it is still hard to isolate the problem and to figure out
how to resolve it... And no single piece of test equipment today can test
an entire network. The second problem with testing a network from
end-to-end under real-live load conditions is tying up network resources.
Whereas, to test an individual component you may only have to tie up that
single device, and if you take it out of the network you do not use up any
network bandwidth. So while component type testing is not always the most
realistic, it is often the easiest and unfortunately what many companies
end up doing. A good compromise is to create as realistic a test lab as
possible that allows you to simulate real life situations."
24x7 SERVICE
ASPs are all the rage these days, and understandably so. What they
offer is the implementation of a basic business rule -- focus on your core
competencies. They do this by saying, "Let us handle your network. We
know it has to work all the time; we know it has to provide reliable
connectivity, bandwidth, and enhanced features. We'll handle it for you.
You just go about your business."
So, are different strategies employed when testing for enterprise/call
center type environments versus testing service provider environments?
What particulars, if any, might be of more interest to a service provider
than an enterprise or call center?
Brian Miller commented on the place of economics and technology in
testing: "A financial institution receiving five million calls per
month will be concerned about dropped calls because some percent of those
calls will be lost transactions or potential lost customers. This will
drive these companies to focus on testing solutions that can continuously
monitor the customer experience and then identify key failures in service
level. The test investment will focus on monitoring the automated
applications. On the other hand, a service provider or an enterprise that
has most of the calls going to an agent will focus much more on technology
that allows automated monitoring of performance through to the agent.
"Technology can also drive test strategy. Enterprises are
accelerating deployment of Web-enabled call centers that allow customers
to interact with agents through a variety of channels. This in itself
isn't new. What is new is the ability to opt out of one medium into
another and to integrate these channels into a common set of workflow
rules to route customer inquiries. Because all of these channels share a
common database and common workflow processing, it is important to test
their interaction. This requires coordination of virtual Web, voice, and
e-mail users, including the ability to ramp the load on any one of these
channels and measure the resulting performance impact."
Alan Sguigna, VP of marketing for Adtech Product Line, Spirent
Communications, had some comments more relating to protocols:
"Testing strategies for enterprise/call center applications versus
service provider environments are significantly different, particularly in
the areas of protocol support and feature test. For example, VoIP
signaling in the enterprise space is typically based upon the ITU's H.323
protocol, whereas the carriers are deploying protocols such as MGCP,
Megaco/H.248, or SIP. This is because service providers are looking for
simple protocols to support basic Class 4 and Class 5 services, whereas
the enterprises are looking for the rich multimedia support of H.323.
"The same distinction applies on the voice side of the VoIP
equation. Enterprises are looking for analog/digital phone set support,
with basic CAS and PRI signaling. Carriers are looking for SS7, GR.303,
and other such interfaces. In both cases they are looking for a robust
underlying IP or ATM infrastructure which can provide reliable voice QoS."
And Van Do, VP of marketing for Zarak Product Line, Spirent
Communications, also had some comments: "Generally, enterprise
networks, call centers, and service providers require the following common
testing features: check PSTN connectivity, connectivity to IP networks,
and performance under heavy load; emulation of incoming calls; and
termination of outgoing calls.
"The differences in testing strategies involve the intensity of
testing software applications. Enterprise networks and call centers
usually require more functional testing, which is the only way for
environments like call centers to ensure appropriate operation of the
software applications behind the scenes. Scripting from within the test
call becomes vital to performance assessment. In addition to the number of
calls serviced, there is concern with reaction time, correct prompts, and
correct processing of information. The service provider concerns itself
more with the quality of service, the number of calls it can connect, and
billing. Service providers apply the most intense of tests at the
integration stage of equipment and software. After integration,
maintenance tests are performed at a smaller capacity but at regular
frequencies."
WIRELESS & TESTING
And what about wireless? Sure, once the traffic hits the base station,
it's should be the same as any other PSTN traffic, right? But what about
when using an IVR? Could something happen in the air that might interfere
with service? I've occasionally had voice mail appear on my own phone a
day or so late. How might wireless affect enterprise/call center/service
provider needs, and what might be its special testing requirements?
Brian Miller was reassuring on this point: "To date, we have not
seen a huge effect on testing. Over time, as use of voice-based browsing,
unified messaging systems, and other applications that use advanced speech
recognition to convert voice inputs into digital becomes more prevalent,
verifying the functionality of these systems under a variety of user
conditions, including wireless callers, will become increasingly
important."
Van Do's comments were also reassuring: "When testing a wireless
interface, more emphasis is placed on testing the connection for quality
and ability to pass modem traffic. Other than that, the call is treated as
a wired call. There are too many variations of physical wireless interface
to be able to provide standard testing tools. However, equipment
manufacturers are usually clever enough to design in wired interfaces to
the wireless protocols, allowing test labs to verify the accessibility of
the network."
MORE THAN JUST A TEST
Developing a testing strategy depends heavily on your business needs and
how much or how little of your communications you decide to handle
in-house. But once in a place, an effective testing strategy helps keep
you up and running with minimal intrusion and, most importantly, minimal
surprises.
[ Return
To The November 2000 Table Of Contents ]
|