A
wide area network (WAN) basically connects two or more
local area networks (LAN). While this may
sound like a relatively straightforward goal, the implications
of WAN connectivity have had a profound influence on networking.
Early on, a WAN was typically proprietary, connecting corporate
networks in different geographic locations, but WAN connectivity
has proliferated with the growth of the Internet. Today, the WAN
has evolved into a global network tying remote offices to
datacentres, connecting remote
backup and
disaster recovery sites, and even connecting
customers and partners to shared corporate resources.
Still, the global WAN is hardly universal -- it has evolved as a
mishmash of cabling, switching and routing technologies supported
by a vast array of service providers that are guided by disparate
local and national regulations. It's easy to get WAN connectivity,
but selecting the "right" connectivity requires careful
consideration of practical issues, such as bandwidth, cost, latency
and reliability.
Private vs. public
Traditional WANs connected two or more locations using dedicated
infrastructure (aka leased lines) provided by a telecommunications
provider. Leased lines are generally expensive, but they provide
good overall service levels and reliability. Leased lines also tend
to be secure because they are not accessible to users outside of
the locations that are connected.
Private WAN connectivity is still available but has been largely
overwhelmed by public WANs (e.g., the Internet) where connectivity
is shared between multiple users on the same infrastructure
supported by Internet Service Providers (ISP). This reduces costs
for the individual users and allows network connectivity literally
around the world. However, performance declines as more users
compete for service. Global access also carries global security
risks where users can attempt to access unauthorized or protected
data resources.
Bandwidth and cost
Bandwidth is the first consideration for
most WAN deployments, denoting the available data transfer rate
between points -- usually expressed as bits per second (bps).
Higher bandwidth is crucial because it allows more data to be
passed in a given time. For example, a 3 Mbps connection can
potentially pass twice the data moved by a 1.5 Mbps WAN
link.
WAN connectivity is typically organized into bandwidth
"classes." Low-speed WAN connections often include
frame relay at 56 Kbps or 1.5 Mbps speeds.
DSL and cable connections are available in a variety of service
levels up to about 3 Mbps. High-speed/high-reliability WAN
connectivity is also available for business users, including T1
service to 1.5 Mbps and T3/DS3 service to 45 Mbps. T3 service
can be shared between multiple users with lower bandwidth needs
(dubbed fractional T3). It's important to note that additional
bandwidth costs more money, and WAN service is a recurring cost
for organizations. For example, a business cable WAN connection
can run up to $250 per month, while a T3 line can run over
$10,000 per month. Bandwidth increases can become a significant
hit to the monthly budget.
But this tradeoff between bandwidth and cost puts companies into
a quagmire. Corporate data volumes are spiraling upward, growing at
rates that sometimes reach 100% annually. As the amount of data
increases, it takes longer to transfer that data between locations.
Thus, companies have to shoulder the cost of higher bandwidth or
employ data reduction technologies to shrink the effective data
volume.
Latency
Latency is an important factor to consider
because electronic data does not reach its destination
instantaneously. Not only does it take finite time for network
data to traverse the physical distance between two points
through its copper/optical transmission medium, the data packets
must also travel through an array of hubs, switches, routers and
other network traffic management devices -- each device adds a
delay, slowing the effective response time. To exacerbate
matters, a typical application may require extensive packet
handshaking, significantly worsening the effective delay.
Latency can have a detrimental effect on remote application
performance, forcing users to wait (sometimes minutes) for a single
query response. This can also have an adverse effect on remote
mirroring and replication setups. Latency is such an important
issue that synchronous (real time) data replication can only be
supported within a fairly short physical distance. If real-time
replication is not required, asynchronous replication can support
global data transfers.
Reliability
Reliability is vital for a WAN, but no WAN connection is 100%
reliable. Data can be interrupted at any point due to
cable/hardware malfunctions, poor configurations, natural or
man-made disasters and so on. Reliability is also affected by the
quality of the service. For example, a typical SOHO broadband cable
connection is generally less reliable than a T1 or T3 line where
service levels are guaranteed.
Any WAN interruption can impact business operations, so it's
important to consider the potential effects of various outages and
establish contingency plans accordingly. For example, a service
interruption can interfere with the corporate remote replication
process. If the interruption is brief, data may simply be cached
locally and resynchronized when the WAN link returns. Interruptions
may also cause users to switch over to local application versions,
such as a database, running locally cached data. Longer
interruptions may cause the replication process to switch over to a
local secondary storage system.