Many of the evolving ways that organizations use their IT resources are placing new and significant strains on wide-area networks (WANs). Chief among these trends: the dispersal of users, enabled by the ubiquity of wired and wireless networks. Users are no longer tied to an office location when they need to work. At the same time, many organizations are trying to reduce the number of data centers they operate by consolidation, often by virtualizing servers and other hardware. This reduction in the number of data centers and the growth in the number of users who work remotely greatly increase demands on the WAN.
Growing volumes of data also stress WANs. Big documents and images, video applications and Voice over IP (VoIP) have all become routine elements in daily work. Even traditionally organized companies — for example, those with a large central office and many branch offices — are relying more on their WANs.
The rapidly growing use of mobile computing has accompanied this dispersal of users. Mobile computing requires organizations to send and receive data from anywhere. This increases reliance on the untethered mobile devices and wireless networks in use. Mobile devices mostly operate without any onboard optimization software. Adding to the network challenges are newer technologies, such as desktop virtualization, and voice and video applications. Whether deployed on a mobile device or some other network endpoint, these technologies produce massive volumes of network traffic, which leads to congestion and saturation of network links. This can slow application response and degrade delivery of multimedia content.
Fortunately, WAN optimization gives IT managers a way to preserve an organization’s existing bandwidth and avoid expensive investments in more bandwidth while also improving application performance.
The Bandwidth Dilemma
One way that many IT managers try to address poor WAN performance is simply to acquire more bandwidth for the long-distance connections among data centers and branch offices. This may seem to be an intuitive answer, but it’s not necessarily the correct one.
Bandwidth is expensive. A T1 line can cost as much as $400 per month. Multiprotocol label switching (MPLS) circuits can run upward of $1,000 per month per 10 megabits of capacity. Additional lines require termination equipment, although sometimes carriers will provide it as part of the cost for long-term commitments. But IT shops that want to gain more efficiency from their WANs miss the point if their answer is to drive up monthly operational costs.
Most entities would prefer to reduce, not increase, their operations and maintenance costs, which would leave more money available for innovative application development related to the organization’s mission.
Further, added bandwidth may not actually provide the performance improvement the organization seeks. In many cases, a circuit will deliver less throughput than it is rated for (which is basically a theoretical limit). And latency over long distances or multiple transfer points can also reduce the realized bandwidth.
More important, merely adding bandwidth doesn’t address many of the factors that contribute to poor application performance. TCP/IP window size and scaling aren’t affected by the presence of additional bandwidth. Adding bandwidth doesn’t change the maximum bit volume that a server will accept before acknowledgement, so application performance won’t fundamentally change.
In addition, bigger pipes don’t address the jittery nature of TCP/IP traffic, nor do they accomplish the compression, buffering, data caching, load balancing and deduplication necessary to improve the speed of applications. To be sure, bigger pipes can speed transfer of large files in one direction, such as videos or big PDF documents. But they don’t fundamentally address the issues that can hinder applications delivered using HTTP.
Want to learn more? Check out CDW’s white paper, “Getting the Most Out of the WAN.”