Nik Rouda

Unstoppable Forces and Immovable Objects
Conflict is a fact of life in IT. Each year, a number of conflicting trends emerge, and weighing up the value of each "next big thing" can be a full-time job (as if IT managers didn't have enough to worry about as they try to keep the status quo of having IT running smoothly). This article will explore a few examples of conflicting pressures.

Current best practices lean heavily toward the centralization of IT resources, which means moving applications, servers, and storage from distributed branch offices to a very small number of data center locations. The advantages of consolidation include increased utilization, easier hands-on administration, and better physical security. Contrary to this goal is the widespread increase in remote and branch office workers who rely on the access and performance of these same IT resources to do their jobs. As they spread out ever further around the globe, away from the IT resources you're consolidating, the total distance between the users and their required tools increases. How can they be productive anywhere without putting the kibosh on your plans for centralization?

Another pair of conflicting trends is virtualization versus data protection. On the face of it, virtualization is supposed to go hand-in-hand with consolidation to remove the physical footprint, embodying the principal of freedom from service interruptions, planned or otherwise, due to hardware failures or site disasters. However, the centralization initiative that brought everything into a single data center also foments the problem of having all your IT eggs now virtualized in one basket of servers. A single site may now provide services to an entire enterprise, meaning that an impact to that site may also bring down the whole business at once. As a result, it requires adding a secondary site for disaster recovery.

Compound this further with the newly possible dichotomy of IT resources between your environment and public cloud service providers. IT managers might well want to get everything under one roof for increased control and manageability, but hosting applications, servers, or storage in the cloud might have an overwhelmingly compelling cost advantage, not least in elasticity. A single strategy might be simpler, but it doesn't necessarily enjoy the flexibility to deploy each component where it's best provisioned. How does IT decide if cloud services fit the company's business requirements, or if the cloud requires IT to re-architect to fit their models?

There's also the eternal oppositional pressure between investing in new capabilities and growth versus cost cutting or control for tight budgets. What's the normal outcome of all these mixed signals? As might be expected, the results are decidedly mixed, with inconsistent delivery to service levels and very often performance problems, if not locally, certainly for the remote users' productivity and satisfaction.

What to Do? What to Do?
There are a number of typical reactions to the challenge of how to deliver faster access to workers regardless of their location. Often the first idea is to simply add bandwidth to the wide area network (WAN.) This may help the peak throughput capacity of the network, but comes with substantial ongoing operating expenses, often with long-term contracts to service providers, multiplied by the number of offices upgraded. What's more, all the bandwidth in the world doesn't address the performance degradation caused by latency. The further the user is from their IT, the longer it takes to get there.

In reaction to this distance problem, many organizations redeploy local servers in key offices, largely defeating the original point of the consolidation project. Reintroducing local servers and storage many satisfy some people, but leads right back into IT sprawled across the enterprise.

Some might argue that these challenges can be overcome by using a quality of service (QoS) model on the WAN. The argument is that maybe QoS will help prioritize key applications, with the squeakiest wheels getting the oil. However, the downside is QoS often robs Peter to pay Paul, shifting the problem rather than really addressing the underlying issues.

Not least, IT must find a way to solve the disaster recovery dilemma. One approach is to tier storage, protecting only the most valuable data with replication across the WAN to the recovery data center. However, this raises more questions than answers, namely: how do you know which data is most important, how do you know what interdependencies exist, and how do you justify leaving any business data exposed to possible permanent loss?

About Those Five Considerations We Promised You...
A useful exercise may be to break out the primary factors that drive many of these issues. These might be defined as location, performance, scalability, cost and flexibility. Each one is a top-level consideration with some direct influences on each other, as well as impact on how the overall IT infrastructure will perform and be managed. Let's explore these in turn.

As they say in real estate, most important is "location, location, location." Where the applications, servers, and storage are hosted has significant implications for the outcomes, some of which we've already mentioned. The manageability and visibility is often dependent on how easily IT can see, diagnose, upgrade, and repair/replace any component. Beyond the value of close proximity to IT staff, location has direct ties to performance.

Performance is often a game of "find the bottlenecks," of which there can be many.  Assuming appropriately fast storage for the data, well-equipped servers to talk to the storage and that efficient applications running on those servers have been set up, then things are probably pretty rosy for the local clients. If you haven't already tackled this technology stack, well, we'll leave that aside as a trivial exercise for the reader. What about the remote user's experience? This is again usually bottlenecked by the capabilities of the WAN and the data center to scale.

Scalability should be evaluated for more total (throughput) capacity and total (connected) users. Scaling up matters if increased demands are placed on the network as the volume of data grows. Scaling out matters as more users demand support from a central location or if the workforce is grown organically or through acquisition. Either way, it's essential to build an infrastructure that is capable of handling the variable load efficiently with respect to costs....

Cost is another issue to take under consideration. IT catalog has a price as one-time capital or ongoing operating expense, or often both. As the economy recovers (or doesn't) it's important to work with what is already in place today, preparing for what's coming tomorrow. Throwing money at problems in terms of adding capacity is far less ideal than investing in efficiency, hence the impetus for new technologies like virtualization. Controlling costs as much as possible will set IT up for future success.

Flexibility is the art of not painting yourself into a corner. Choose technologies that have both a past and a future, and even though the cost is sunk, the ship sails on, requiring organizations to adapt as they go. Flexibility also allows IT to opportunistically seize new technologies that are promising new capabilities in terms of cost efficiency, scalability, performance, and service location, neatly bringing us full circle.

WAN Optimization - A More Elegant Solution
Looking at this another way, the common thread is the ability of the WAN to provide fast and cost-efficient connectivity across the enterprise. What can be done to improve the WAN? This is where WAN optimization technologies enter the stage, overcoming the limitations of distance to deliver data where and when it's needed.

WAN optimization comes in a variety of flavors, but the best approaches include all of the following functionality:

• Free up space on the existing WAN by deduplicating network traffic, finding and removing redundancy in email, file sharing, Web-based applications, replication, etc. This can often increase effective bandwidth capacity up to five times, and sometimes up to twenty-fold, delivering considerable savings on network costs.
• Streamline the network, tuning the protocols that run on the WAN. Most often, and for most traffic, this is TCP, but may also include Remote Desktop Protocol (RDP) for virtual desktops or others. This allows network traffic to ramp up faster, utilize more of the rated throughput capacity, and even prioritize key applications.
• Understand the applications, finding inefficiencies in chatter that are compounded by high latency in the WAN to slow things way down. The best products have the broadest range of application-specific optimizations, which when combined with the above can speed things up by 10x, 20x, or more, allowing operations that previously took hours to complete in minutes or seconds, independent of the distances involved.

Beyond these table stakes, a good WAN optimization solution should have inbuilt platform flexibility, with options for mobile workers, virtualized data centers, cloud computing environments, clustering for scalability and high availability, all those attributes we discussed as prime considerations for an effective architecture.

Resolving the Conflicts
Conflicting trends in IT can affect every organization's ability to deliver a consistent high-performance infrastructure for workers everywhere. Adding or rationing bandwidth, redeploying local resources, or tiering service levels don't fix the underlying problems. There are a number of considerations, but they seem to be interlocked in a vicious circle.

With WAN optimization, the network can be transformed from a barrier to an enabler of IT initiatives, solving many of the contradictions of emerging IT trends. Organizations can accelerate application delivery, while at the same time centralizing, consolidating and virtualizing servers in the data center. IT can reduce bandwidth requirements, while increasing the number and types of applications running on the WAN. Everything can be replicated, not just a random pick list, for comprehensive disaster recovery in an offsite location, and do so on the same network without impacting other business traffic. IT can adopt cloud computing services where they make sense for the business' strategy, and integrate the best choices seamlessly, even across the public Internet.

We've looked at five considerations to enable a more agile infrastructure, discussing the dependencies between location, performance, scale, cost and flexibility, and how organizations can increase agility with a new approach to the network that connects everything. Now it's time to decide how you want to resolve these issues and finally get ahead in the game. Who wants to play? Are you in or out?