About two thirds of
application deployments in UK enterprises are considered
failures when assessed against the original roll-out plan,
writes Owen Cole, technical director at F5. Failures are most
prominent in businesses with offices in multiple locations and
usability problems increase in areas of high latency and
poor connectivity, for example in South America, Africa and the
Middle East.
Adding to the problem of application performance is the fact
companies are
consolidating datacentres to comply with regulation, which
takes applications and data further away from the user. The growth
in companies
putting applications on the web to make them easier to use when
shared across different countries also takes them further away from
the user, causing yet more latency issues.
Businesses are investing millions of pounds in applications
today and failing to deliver them properly costs time and money,
with hundreds of hours wasted when projects are abandoned.
Why are they failing?
The majority of applications rolled out today don't meet the
original specifications. One of the main problems is that
developers don't foresee how applications are going to be used in
the real world. At the point where they think they've finished the
application, many more security and network problems can arise.
When issues do arise, the network manager blames the application
developer and the application developer blames the network manager.
The most common reaction is for the network manager to
buy more bandwidth, but
this isn't necessarily going to help because the reason the
problem exists is almost always multi-faceted.
It could take up to six months to investigate and fix
performance problems once the application has been deployed and
this is not only very expensive but often impossible when
contractors who worked on the job initially have moved on in a
transient technology market. If the application deployment was
planned more effectively at the outset both money and effort could
have been saved.
All in the planning
Companies deploying applications for mission-critical projects
must create a virtual team to plan the roll-out and involve all
interested parties in the planning stage. This team must include
the application developer, network manager and security manager.
The team must look at the application it wants to roll out and
decide what it will deliver. For example, if a company is rolling
out a collaboration product with Microsoft SharePoint, it needs to
consider the distances that exist between the new and existing
applications and how this is going to influence changes in the
network profile.
The amount of time it takes a page to load must also be
considered. A typical page for a local user could take a second,
for a remote user in Europe it could take five seconds, but for
someone in South America it could take five minutes. This can't be
reduced but can be planned around by better design in the
application. This option can be complicated, however, and the other
alternative is to add devices to the network to accelerate
application performance. Application acceleration uses a number of
technologies to improve application performance and response time
over network connections. It can assist in mitigating issues such
as
WAN latency,
packet loss
and bandwidth congestion. It also addresses application challenges
that adversely affect performance such as "chatty" protocols, for
example
HTTP, CIFS, differences in TCP/IP stack implementations, and
the lack of distinction in web applications between cacheable and
non-cacheable content.
Whichever approach is taken to application delivery, the main
thing companies need to recognise is that it is a big problem.
Continuing to invest in applications without thinking about their
delivery can affect employee productivity levels and in today's
environment, where businesses need to be as agile and competitive
as possible, application delivery must surely move up the up the
priority chain.