Aug 22, 2011
Network managers responsible for their firms’ WANs find themselves at a bit of at impasse these days. With ever-increasing demands upon their network due to large file capacity, video and audio, not to mention the demands of cloud access, they have to decide whether to purchase more bandwidth to accommodate these additional demands or make do with what they have.
The dilemma is this: purchasing more bandwidth poses an immediate financial cost, but failing to expand and forcing their WANs to labor under too much data would likely result in financial implications in the long-run due to latency and packet loss. Demands of the cloud and disaster recovery would further compound these long-term financial implications.
So, what’s a network administrator to do?
The knee-jerk reaction is to spend the money and buy more bandwidth capacity, the reasoning being that the long-term financial implications of not expanding could be more costly than not expanding at all, but that may not solve the problem.
Why? Because bandwidth is only one-third of the problem addressed by WAN optimization. There are also latency issues due to distance and quality issues due to congestion (i.e., packet loss). Buying more capacity only helps when bandwidth is the limiting factor. If latency or losses are the limiting factors, then more capacity will not make applications run faster or with higher quality.
In addition, in many instances more bandwidth is not what is really needed. What is needed is better bandwidth utilization. By properly prioritizing traffic and allocating resources, you can see significantly better performance with the bandwidth you have.
One of the best solutions to improve bandwidth utilization is WAN optimization (WANop).
WANop solves latency and loss issues that cannot be avoided on many WAN links, especially as more enterprises move to MPLS and the cloud. In addition, WAN optimization helps utilize existing WAN bandwidth better and it is often cheaper to deploy WANop appliances with deduplication than it is to upgrade a WAN link. The actual ROI depends on location, the size of the link, the telco rates, etc.
WANop devices have become more robust over the last few years, which means as a solution it can do a pretty good job on its own and sometimes avert the need for network managers to invest in additional products.
Take, for example, the WANop solution from Silver Peak. It has TCP acceleration techniques to mitigate latency, network integrity features to fix packet delivery problems (i.e., dropped and out of order packets) and deduplication for bandwidth savings. In addition, it incorporates QoS and traffic shaping to avoid doing this in a separate device and includes IPsec to eliminate the need for separate VPN concentrators.
The Silver Peak solution also focuses on solving problems within the network, not at the application layer. Therefore, it works on all IP traffic – today and tomorrow. This is in stark contrast to caching solutions, for example, that only work on a single application. The result is that one product can optimize an entire global network.
But how do you decide if WANop is the right solution? If bandwidth is your only problem, then a TCO analysis is required to make a determination of how much WANop costs and how much bandwidth savings it provides versus the cost of more bandwidth.
Silver Peak recently started offering both subscription and perpetual pricing to make this comparison easier. The former is an annual fee, so it can be easily compared to ongoing telco bandwidth expenditures. The latter is a onetime fee that depreciates over time.
If enterprise WANs also have to deal with latency and loss or are worried about the performance of real-time traffic over the WAN, like voice, video, VDI, SQL and replication, then this is a moot discussion. Organizations simply cannot just “add capacity” to address these issues; they need an optimization product.
As long as multiple applications are sharing a WAN — or are using shared WAN technologies like MPLS and cloud — congestion is inevitable. This is especially true for international WANs where traffic is traversing multiple carrier’s networks! There is no way to avoid congestion completely without spending a fortune on dedicated WANs, so the key is to mitigate the impact of congestion when it does arise.
Silver Peak uses QoS to prioritize traffic and traffic shaping to allocate bandwidth. Both of these help to handle congestion. In addition, Silver Peak has unique Forward Error Correction (FEC) and Packet Order Correction (POC) to gracefully handle dropped and out of order packets (respectively) that inevitably result during periods of congestion. This is actually one of Silver Peak’s unique differentiators.
The typical MPLS network has about 0.05% packet loss. The typical Internet deployment (i.e., VPN or public cloud) has about 1% packet loss. Both of these levels of loss make it very hard to deploy real-time traffic, no matter how much bandwidth is available. The more users understand the fact that bandwidth isn’t their only problem when they are using shared WANs, the better they will be.
Still not sold? Here is a great calculator that shows how latency and loss affect effective throughput just as much as bandwidth: http://www.silver-peak.com/calculator/. The results are pretty compelling.