The Network Demand of Learning on Demand

By Monique Lucey

There seems to be a perfect storm brewing that will increase the need for on-demand learning: an influx of out-of-work education seekers, another looming round of the H1N1 virus, and reductions in faculty.

Combined, these forces are putting pressure on higher education institutions to boost their on-demand – or distance learning – offerings. However, this comes with a Catch-22 for IT teams: they need to support more applications without an increased budget.

A 2009 Babson Survey Research Group report found that more than 4.6 million students were taking at least one online course during the fall 2008 term – a 17% increase from the previous year. At the same time, 50 percent of institutions saw their budgets decrease due to the economy.

Now, let’s keep something in mind about on-demand learning: it is a multimedia-rich environment that can involve thousands of students from all over the world or right next door clamoring for the same information at the same time.

Consider the impact of a resurgence of the H1N1 virus, which is predicted for this spring. The Babson report says “substituting online classes for face-to-face classes is a component of 67 percent of H1N1 contingency plans.” So you might be used to having a certain number of users accessing your network in a face-to-face learning environment, and all of a sudden, you are going to have to ramp up to support the entire student body accessing data on-campus and remotely with the same infrastructure! If you’re scratching your head, you are not alone.

Here’s another scenario: Unemployment is at an all-time high. As people continue to hit a brick wall in finding work, chances are they are going to want to beef up their education to pass the time and improve their resumes. Many will want to do so via online learning.

Also, depending on your university’s location and the local unemployment rate, your leadership might partner with government to expand course offerings for citizens. Many of these will be offered over the Internet because academic institutions are stretching their faculty resources by moving courses online.

The Babson survey found that 87% of nearly all public institutions say the economic downturn has increased demand for their existing online courses and programs. At the same time, the report says, “virtually all institutions report an increase in demand for financial aid, and about one-half say that the size of the institutional budget has decreased as a result of the economic downturn.”

So where does that leave you? You certainly can’t say no and that it’s too much for the network, and users will start to see application slowdowns and possibly outages. There’s no way you’re going to force your organization to miss out on the revenue that will be generated by these courses. Yet, you know you can’t continue to pile on more users and resource-intensive programs without the infrastructure beginning to buckle.

Over the next few blogs, we’re going to help you strategize how to support this surge in on-demand learning without breaking the bank or breaking your network. We’ll discuss challenges and solutions at all levels, including the data center, campus LAN, satellite campus, and mobile/remote.

You’ll see that rather than crumbling, your network will thrive and, therefore, so will your institution.

Talk Is Not Cheap, It’s Cost-Effective

By John Gray

In the previous blog, we started our discussion about energy efficiency – a key component of the next-generation data center – by addressing the need for real numbers regarding consumption. But just as important as having this information is communicating it to the right people.

Sound familiar? Your university needs to bring in a new admissions system, so the network team specs out the proper hardware based on performance requirements and then has it shipped to the data center. It isn’t until the gear arrives in the data center that the facilities team is aware of the additional power and cooling needs of the new infrastructure. This is despite that fact that the facilities team is most often the team responsible for ensuring that there is enough energy and HVAC to keep the infrastructure going. And it is usually the facilities team that has to account for electric and other utility costs in its budget.

Doesn’t sound very fair, does it? Or very cost efficient? What if that new gear needs just enough power that the electric grid would be considered at capacity? Or what if the HVAC system needed to be upgraded to account for new hardware? Both situations could delay the rollout of your admissions software, and therefore jeopardize student enrollment.

The better strategy is to identify who in the data center and the overall physical plant are responsible for supplying energy and HVAC to the building as well as the entire organization. Establish a virtual energy-efficiency team with these folks so that when you are making hardware purchases, you can run the real-world numbers you’ve gathered by them.

Armed with this information, this “green” team can determine whether the data center has enough energy and HVAC resources to handle any increased demand. If not, you might be able to scale back your plans before users are impacted by delays or poor performance.

This team could also begin to research the Energy Star program and Green Grid tools noted in the previous blog to gain an understanding of the actual consumption within the data center. It may become clear that if you upgrade to newer, more energy-efficient servers and switches, you could extend the life of your existing data center and avoid an expensive expansion.

Together you could also develop guidelines for purchasing and deploying new gear that could speed rollouts. For instance, by having the facilities teams involved at the time of purchase, they could have floor space, power and cooling all mapped out by the time the new equipment arrives.

Most importantly, you’ll be tuned in to any changes in how power and cooling are supplied to the physical plant. For instance, you’ll know if there are plans to make part of the data center run off solar energy or to use more energy-efficient cooling systems. The cost savings from going green might be able to help fund strategic, revenue-generating projects such as new courses and cutting-edge research.

It’s also likely that your superiors will be impressed that you took the initiative and reached out to the facilities teams to save money, streamline business processes, and improve student and faculty satisfaction. In this time of shrinking endowments and tight budgets, that’s a very good thing.

Ga-ga for Green

By John Gray

In this last part of our discussion about evolving to become a next-generation data center, we’re going to tackle the ubiquitous topic of energy efficiency. We know that you’re getting bombarded with questions about how “green” your data center is, and as you seek budget for resource-intensive applications such as real-time distance learning and advanced collaboration, you’ll need real answers.

Deploying these projects in a time of tight budgets, shrinking endowments and limited data center resources (i.e., lack of real estate and power grid resources) will take some strategic planning on your part. Somehow you have to roll out systems that will contribute to the bottom line without sending power and cooling costs into the stratosphere or requiring so much more data center space that you have to build a whole new facility. Both of these approaches negate whatever revenue might be gained from these exciting new projects and are considerably environmentally un-friendly.

Add to this that Nemertes Research says it won’t be long before data centers are maxed out energy-wise. “By the end of 2010, more than 50 percent of large data centers will reach maximum power consumption levels, leaving consumers in a major energy crisis as these organizations will be unable to scale data operations to support their business needs,” says Andreas Antonopoulos, senior vice president at the research advisory firm.

So how can you avoid this outcome? The answer is twofold: Know the real energy consumption your data center infrastructure—including servers and switches—will require, and communicate that information to your data center and physical plant facilities teams.

Let’s start with having real numbers. Antonopoulos warns that the lack of data center resources will mean that “power efficiency and accurate data on power consumption is absolutely vital for the future.”

“To avert this crisis, organizations need to not only measure their own data on power consumption, but also, they need to begin demanding that vendors disclose all product efficiency data in order to serve as a key criterion for customers who are choosing the tools that will best provide them with sustainable next-generation data centers,” he says.

Too often, network teams base their projected power and cooling architectures on vendor data sheets. But these on-paper numbers do not necessarily reflect how the products will perform under the rigors of your specific data center environment. The inevitable discrepancy can present serious problems. If you overprovision resources, you’ll end up spending more than you need and impacting the energy efficiency of your data center. Under-provisioning will leave you scrambling to bring in more power and cooling infrastructure, which can impact network performance, drive up costs and potentially force you to unnecessarily add on to your data center.

What you should be using to gauge the energy impact of network gear on your data center are real-world numbers. This way, you can accurately assess whether you can eke more from your raised-floor data center. Since the cost of a new data center is estimated to start around $20 million, working with real-world power and cooling requirements is essential.

The U.S. Environmental Protection Agency will take a big step in making this possible come June when it plans to extend its Energy Star program to data centers. According to the IDG News Service, the program will offer incentives for organizations to make their data centers more energy efficient and offer tools to track the results of conservation projects over time.

“Data centers that take part will use an online tool that ranks their efficiency on a scale of 1 to 100. Those that score 75 or higher can request an audit from the EPA, which then awards the Energy Star certification,” the article states.

The author points out that the Energy Star program is also recognized in Europe and that China and India have agreed “in principle” to use the EPA’s system to rank products. Andrew Fanara, the program’s lead at the EPA, is quoted as acknowledging that avoiding a patchwork quilt of regional programs is important, especially for multinational companies.

Some vendors have even started certifying the energy efficiency of their own products. For instance, independent testing firm Miercom offers its “Certified Green” program to provide vendors with an objective assessment of the environmental impact and business case effectiveness of their products as compared with national indices.

Organizations hoping to nail down some stats on their energy efficiency today can use the online resources at The Green Grid. There you’ll find a power configuration efficiency estimator, a data center design guide, a presentation on real-time energy consumption measurements in data centers and other valuable tools to help you better understand how to manage your energy demands.

You can also alleviate your concerns about energy efficiency via a flattened architecture, which we’ve discussed in previous posts. By moving from a three-tier strategy to a two-tier one, you can reduce the number of devices and interconnects in your data center consuming floor space as well as power and cooling resources.

Being proactive about energy efficiency ensures that your organization won’t have to choose between spending money on power-hungry network infrastructure that might tap out the grid and force you to construct a new data center facility versus purchasing new research lab equipment that could help find a cure for cancer. You’ll be able to show university administrators that not only are you choosing to be green, but you’re also saving green—money, that is—in the process.

In our next blog, we’ll talk about the second half of the energy discussion: planning power and cooling needs with your data center and physical plant facilities teams.

10G Ethernet: Paving the way to FCoE

(This week we continue our series on the data center.)

By John Gray

If you peer inside the data center of a large higher education institution, you’re bound to see at least three separate fabrics being supported: Ethernet for the LAN, Fibre Channel for the storage area network, and the specialized Infiniband for high-performance computing (HPC). Keeping with our discussion on the characteristics of the next-generation data center, managing three separate fabrics is hardly the path to simplification.

However, it’s understandable why you’re holding onto this type of architecture – to unify, you’d need an underlying network technology that could handle the unique demands of all three environments. Gigabit Ethernet alone is often not robust enough to deal with the I/O requirements of HPC and large-scale SANs. Fibre Channel and Infiniband are expensive technologies that need highly trained staff. In a time of tight budgets, the double whammy of costly hardware and training is not a recipe for success.

Over the next few years, you’ll see the adoption of a unified network fabric based on Fibre Channel over Ethernet (FCoE) as it evolves and becomes more commonplace in network gear. FCoE enables large data centers dealing with tremendous traffic and storage loads to preserve the benefits of Fibre Channel, which has a strong installed base, while taking advantage of the ubiquity of Ethernet deployments and common skill sets.

Although work on the FCoE standard was finalized last year, there are still significant components that need to be completed to ensure the reliability that users have with traditional Fibre Channel. For instance, there are protocols in the works still to deal with congestion notification, enhanced transmission selection and priority-based flow control. All of these are working their way through the IEEE’s standards process.

As you wait for these piece parts to be settled out, it’s important that you start to look at the most critical element of FCoE’s success – 10G Ethernet. For FCoE to be able to handle the load as gracefully as its counterparts, it needs the workhorse that is 10G switching. Think about the I/O you’ll need to handle virtual learning and other multimedia traffic, backup and storage.

The migration to 10G Ethernet in the near term offers you many benefits beyond the future support of FCoE. Since 10G delivers 10 times the bandwidth of Gigabit Ethernet, colleges and universities can reduce the Gigabit Ethernet NICs they need in highly virtualized environments. Rather than use as many as four to eight Gigabit Ethernet NICs in each server, you can deploy just two 10G Ethernet NICs and achieve full redundancy and availability while increasing bandwidth per virtual machine.

Consolidating around 10G, Gigabit Ethernet network I/O also dramatically reduces the number of Gigabit Ethernet ports, upstream switch ports and cables you need to deploy and manage. And it puts you in line with the goal we’ve outlined over the past few blogs – a flatter network architecture. As a bonus, fewer ports and switches lead to reduced power and cooling.  Both of these outcomes are the essence of the next-generation data center.