Read the Original Article at http://www.informationweek.com/news/showArticle.jhtml?articleID=240006023
Respondents to our InformationWeek Next-Generation WAN Survey are a highly connected bunch: 44% are tasked with connecting 16 or more branch or remote offices to headquarters or primary data centers, and more than half spend 11% or more of their IT budgets on wide area connectivity. So we weren't surprised when one respondent asked an interesting rhetorical question: "Shouldn't there be a Moore's Law of WAN connectivity?"
At least, we think it was rhetorical.
IT's frustration with the pace of service provider innovation is understandable. As with so many technology areas, from storage to app dev, user demand for WAN bandwidth is rising at a faster clip than what companies are willing to pay for, and providers moving slowly on upgrades isn't helping. IT must move to MPLS even as carriers try to squeeze a few more years out of their ISDN circuits. WAN-optimization strategies are poised to lose effectiveness even as providers stall on running new last-mile fiber.
But for most respondents, the biggest WAN game changer is the public and private cloud. Extensive use of public cloud services demands that we shift from the conventional WAN hub-and-spoke design to a distributed model, where services are delivered over Internet connections at each location, thus reducing the load over the corporate backbone. Private clouds, in contrast, bring increased focus on the hub-and-spoke model to afford IT control over latency, quality of service, circuit quality, and overall performance.
Companies considering WAN investments must understand their cloud strategies, the performance levels the business expects, and what it's willing to spend. Well-architected private cloud services will be superior to public cloud services that rely on the Internet, but that quality comes at a price.
Data Center Demand
Fully 73% of respondents to our InformationWeek 2012 State of the Data Center survey expect demand for data center resources--generally including WAN capacity--to increase. Only 7% mention moving to a colocation facility.
Unfortunately, for 40% of our WAN survey respondents, carriers are leaving new business on the table. Just 60% say they have all the options they'd like; we saw unfulfilled demand for Fibre Channel (19%), dark fiber and copper (17%), and Carrier Ethernet services (15%).
The most common use for these protocols in the WAN is to connect data centers, and that brings us back to virtualization and private clouds. Ever since server hypervisors enabled IT to easily migrate server instances among physical hardware, there's been demand to shuffle virtual machines among data centers, often as part of a disaster recovery plan.
Growing dependency on data centers means ongoing calls for ever more bandwidth--sometimes IT isn't getting enough for its money, sometimes fiber isn't available, sometimes network termination equipment doesn't support high-speed services or entire geographic areas don't have enough capacity at any price.
This demand for high-bandwidth data center WAN connection may be met by advances in data center interconnection (DCI) technologies, like Overlay Transport Virtualization and Location Independence Separation Protocol from Cisco. It may also be met by MPLS-based schemes that enable IT to connect Ethernet networks across two or more data centers so that servers don't need to change their IP addresses or DNS entries, easing VM mobility.
The most surprising finding in our WAN survey is that 38% of respondents use Fibre Channel WAN circuits, with an additional 13% planning to adopt them within 24 months. Why is this surprising? Because Fibre Channel isn't suitable for WAN use. It's highly sensitive to jitter and delay and, typically, must have round-trip times of less than 50 milliseconds between devices and miniscule packet loss to be reliable for real-time data replication. These are very tough specifications for a carrier to deliver.
If you want Fibre Channel services, ensure that your provider has deployed DWDM rings in its colocation or hosting facilities. DWDM lasers can carry multiple data channels over fiber at different wavelengths to achieve very-high-density, high-performance backbone connections.
The alternative to native Fibre Channel over the WAN is Fibre Channel over IP, which is often used to add flexibility to DCI setups. Fibre Channel over Ethernet isn't suitable for WAN use but could be engineered for dark fiber once the final FCoE multihop standards are in place, and once vendors support the technology. Unfortunately, support is unlikely to materialize in the near future. The storage industry will stick with Fibre Channel where possible and resist adopting new technologies; even FCIP is finding it difficult to gain traction.
Perhaps we just helped explain the skyrocketing popularity of Ethernet?
Can't Get Much Satisfaction
One bit of bad news garnered from our respondents is that the median percentage of the typical IT budget spent on WAN services hasn't decreased--it's still in the 11% to 15% range, as in our 2010 survey. However, based on conversations with respondents, it's clear that we're getting a bit more for our money as some services decrease in price, so things could be worse. Still, for the majority of respondents, WAN services take a big bite: for 52%, that's more than 10% of the operational budget. And while just 35% expect to increase the amount they spend on the WAN in the next two years, 69% expect demand for WAN bandwidth to increase.
Are we at least happy with the service we're getting for all that money?
At the technical level, engineers are often critical of their service providers and can riff for quite a while about poor support, help desk cases that never get answered, and offshore technical support that doesn't have the adequate skills. Still, 81% of respondents are very (33%) or somewhat (48%) satisfied with their WAN service. The most commonly cited reasons for dissatisfaction: cost, unreliability, and providers being slow provisioning new services.
However, carriers shouldn't celebrate just yet. We asked respondents to rank their satisfaction in nine areas on a scale of 1 (very dissatisfied) to 5 (very satisfied); a 3 translates to "neither satisfied nor dissatisfied." Since the average value was less than 4 in every single area, we can say that, overall, respondents are in the dreaded customer service "zone of indifference" and feel no loyalty to their carrier partners. How much churn providers will see is largely a function of the WAN options in a given area.
For CIOs looking to chart a WAN strategy and reduce the budget hit, there are a few steps to take:
>> Audit your invoices: A typical service provider has many divisions, dozens of teams, and hundreds of products. Billing errors are surprisingly common and often involve large sums. We cover some typical errors in our full report, at informationweek.com/reports/ 2012wan.
>> Audit your services: Many WANs have grown ad hoc over a number of years, even as applications and user profiles changed and services were added and deleted. Check your network monitoring system for visibility into circuit capacity and performance to decide if downgrades are possible in some areas.
>> Find alternatives to local loop: Service providers are constantly upgrading their networks and may have deployed new equipment into your local exchanges. But the only sure way to locate those new services is to repeatedly request them from your provider account manager.
>> Consider WAN acceleration: Although the technology is complex to use and expensive to buy, it's common to compress your WAN traffic between two and four times using these devices. This lowers costs by delaying WAN upgrades while improving performance. However, plan on a maximum three-year ROI for acceleration.
The pace of change in service provider technology is painfully slow. Yes, the carriers face difficult technical and operational challenges. But businesses also have faced significant changes in the past few years, and we've adapted. Continue to hold your service providers to account, and demand better, faster WAN services. Maybe then the 44% of respondents who told us that the biggest factor preventing WAN upgrades is that the service is still too expensive will feel better about spending that cash.
Greg Ferro is a consulting network architect and senior engineer/designer. Write to us at firstname.lastname@example.org.