WAN Governance

The expansion of the digital universe (Application demand variability – Part 2)

We recently talked about the impact of seasonality on application traffic over networks: daily rhythm, weekly period, monthly activity and yearly recurrence induce massive changes about the amount and the nature of the traffic supported by networks. These variations are similar to the waves and tides that modify the surface of the oceans. They shake boats, rock shores and endanger crews, but after sometimes calm and low tides come again in a sort of "eternal return"…

There is another force to cause inflating traffic over networks, superposed to the seasonal swell: the expansion of the Digital Universe. The same way as our physical universe expanded at different speeds during its past life and is continuing to expand - with a velocity of 75 km/s/Mpc at that moment, according to the last findings of the WMAP project - our Digital Universe is steadily growing: currently estimated around 3 Exabytes (3*10¹⁸), experts believe it to grow to 40 Exabytes by 2020, reaching one Zettabyte (10²¹) before 2030. As an example of the digital inflation, let's remind the web page size yearly increase that has been recently discussed here.

All these data must be captured, collected, aggregated, stored before being used for many different purposes; they circulate many times though local and wide area networks. Soon, real-time communications will bring their participation to the growth of the network traffic: a medium quality desktop video communication, simply set-up with one or two clicks, uses 5 times the average bandwidth per employee that is currently provisioned in most of enterprises' WAN. Would it be a high definition communication? The ratio will reach 25!

In the past few years, the average traffic inflation on enterprise networks has been measured at +30% per year, i.e. approximately doubling every three years. Despite we cannot clearly forecast what will happen in the future, it seems that all the conditions for a faster inflation rate are in place: big data, generalization of video traffic, Internet usage at the workplace, etc.

There is no reason to be pessimistic: this data tsunami might be good news for business efficiency and people's quality of life. It just raises a new challenge to IT managers, about the best way to handle it in an efficient and economically savvy manner.

…/…

Béatrice

Illustrations: fishermen in the "raz de Sein" (French Britanny) and Digital Universe (Source: EMC2)

Application demand variability – Part 1

Microsoft's Rolf Harms and Michael Yamartino outlined in "The Economics of the Cloud" the main reasons for the seismic shift to cloud computing the industry is living. They split their study in supply-side, demand-side economy and finally multi-tenancy economy of scale. The networking industry is well used with the first and the last topic: actually it has been addressing them since a long time thanks to technologies like MPLS and VLANs which bring to telecom operators the level of automation and virtualization they need to provide cost efficient and secure Virtual Private Networks (VPNs) services across a shared infrastructure to their large customer base. Illustrating cloud computing, Figure 1 - Economics of scale applies well for VPNs versus fully private networks – a lively topic of discussion in the 'last century' but clearly no more a today challenge: apart from very specific situations like military and high security networks, who really think about deploying his own network infrastructure?

We propose to discuss here the impact of the second aspect: the increasing variability of the demand and its impacts to enterprise networks' performance and economics.

Seasonal application demand variability

Randomness

Employees make use of a large variety of applications in a quasi-random manner: voice call, unified communications, file transfers, ERP and of course the Web. Each of these application flows has also its own profile, depending on the application itself and on the actual activity of the end-user (data entry, thinking, gathering result, printing, etc.). On small and medium branch offices with 10s to a few 100s of employees, this behavior provides random traffic profiles like the one shown in Figure 2.

Time of the day patterns

Daily cycles in employees' and customers' behavior have a strong impact on application traffic. Figure 3 shows the typical daily and weekly shape of application traffic in a B-to-B enterprise. On the daily graph, we can easily recognize the beginning of the day when people start working, the lunch time in the middle and finally the end of the day when people leave. Night activity remains smaller (whilst not null) with monitoring, batch and back-up applications. On the weekly graph, each of the five working days appears clearly with relatively similar activity whilst the week-end application traffic is flat.

Activity in B-to-C sectors is very different from Business-to-Business one. Figure 4 shows the Query volume of Bing Search for two countries (UK and Japan). In addition to the expected time lag between the two continents (one is sleeping while the other one is awake), we can see that there is a plateau of activity in the evening as people continue to use the Internet after dinner.

Industry specific variability

Applications are diversely used according to the business activity, organization and rhythms. For example, the monthly financial consolidation in a public company combines data from multiple cost centers for budgeting, reporting, auditing and publishing data; it usually implies a large number of interconnected systems to exchange a significant amount of data in a relatively short time (typically one or two working days).

Figure 5 shows a few examples of activity variability, like see seasonal peaks during summer holidays (car rental companies) or the Christmas period (www.amazon.com). Tax services have a huge peak of activity when people declare their incomes (for example each springtime for impots.gouv.fr) while sport events like the London Olympic Games in August 2012 might generate their own specific traffic increase (www.visitlondon.com).

Microsoft has even proposed a few estimations about the variability ratio according to the industry (Figure 6). While only specific sectors (tax, retail, news, etc.) see a very large peak /average traffic ratio, all are somehow affected by such event across the year.

 

…/…

Béatrice

March Madness: Lessons for Networks

March Madness, the largest NCAA college basketball tournament in the nation has a reputation for impacting work productivity at companies throughout the United States. The annual, month-long single-elimination tournament is in its 75th anniversary and is being broadcast by CBS online. Since the early round games take place during the workday, the bad news for businesses grows past productivity into how their WAN is being used during these days. The bandwidth capabilities of company networks have been pressured as many dedicated fans stream the tournament from their office computers throughout the week. While U.S. businesses anticipated a spike in demand for network resources during March Madness, data centers, Wide Area Networks and companies alike also prepared, lest they risk the performance of business critical applications.

Mad for Networks

During previous March Madness tournaments, there have been 2.4 million daily unique viewers following the games. The growth of access to the Internet at work will stress U.S. corporate networks like never before. Not only can games be watched on computers, but now mobile devices can stream content. According to a 2012 report, Challenger, Gray & Christmas, an outplacement firm, estimates that more than 2.5 million unique visitors per day each spent an average of 90 minutes watching games.

What These Challenges Mean

According to a Harris Interactive poll, 64 percent of Americans watch online video while at work. As the games occur during business hours, employees are highly likely go online to track their favorite teams. So take the stats mentioned above, and imagine the impact it could have on businesses should their employees flock to watch March Madness and their networks not be able to withstand the increase in demand. According to a survey by Modis, a global provider of IT staffing and recruiting services, two in five IT professionals report March Madness impacted their network in the past, with about a third reporting system slowdowns or complete crashes. If companies aren't prepared, their business critical applications will be affected as they compete against YouTube and online video for limited resources. IT departments need to know how to guard against this. According to a leading analyst house, application performance problems really matter. Losing a mere 5 minutes per day costs 1 percent of overall productivity! It's also important to consider that applications are a huge investment, costing roughly an average of $360/employee/month when maintenance and up-front cost is considered. With all this at stake it makes sense to protect that investment.

Why Bandwidth isn't the Complete Answer

There are two solutions: companies can purchase more bandwidth before March Madness (or other known high-interest events) or companies can use what they have more effectively through Application Performance Guarantee solutions, which control and dynamically guarantee critical applications performance across networks. Adding more bandwidth isn't the solution, as more bandwidth is rarely enough. Non-applications are bandwidth hungry and the more bandwidth you have, the more bandwidth they consume. They will systematically cannibalize critical application performance. And the added resources will never guarantee critical application performances. Application traffic tries to use up all the available bandwidth. Simply increasing it is like filling a bottomless pit: expensive and never enough to satisfy the ever-growing usage demands. The additional traffic may also hinder the performance of the business critical and often resource thrifty applications.

It's Not Size — It's Sophistication

The answer lies in having solutions that allow an IT department to control applications as they flow across the network. Not deploying an Application Performance Guarantee solution implies uncertain business application performance and control that may involve lost of productivity and end users complaints – Enterprises' carry on running IT in a reactive mode that is neither optimum for the business nor satisfying for the IT department. Business applications continue to suffer from bad and/or unstable performance while the network is still over-sized. Unsatisfied users and business managers continue to complain. The WAN is unnecessarily upgraded without solving application performance and end-users experience troubles.

Managing Traffic

If we think of networks as roads, and applications as cars, an Application Performance Guarantee solution might be a police officer. It can direct cars into appropriate queues. It can slow cars that are less critical to the business (for instance, the YouTube car) and prioritize those that the business depends on (such as the Salesforce or SAP cars). This then allows vehicles to get to their destination in a timely, secure fashion – regardless of the amount of traffic on the 'road'.

As March Madness concludes, it is just one example that demonstrates the need for Application Performance Guarantee solutions. These technologies will continue to grow, helping enterprises to face Internet traffic growth challenges in an easy and user-friendly way by authorizing Internet and video traffic, while ensuring business applications perform at their best and that IT costs remain reasonable.

(this article is a reprint from Industry Perspectives)

David White is President North America for Ipanema Technologies. White is a senior executive with more than 25 years experience in sales, marketing and business development, with a background in WAN Optimization.

33% more

To finish 2012 and start 2013, I would like to share some interesting statistics from HTTP archives. The average web page over the Internet is now reaching 1286 Kbps (Dec 2012) from 726 Kbps (Dec 2010), a yearly growth rate of 33%. Would this ratio remain identical, the average page will grow up to 3Mbps by end of 2015, more than doubling every 3 year.

Interestingly, this is the very same traffic growth as Ipanema is noting from its customers base, meaning that Web applications are getting richer at the same speed as the rest of the enterprise VPN traffic.

It also means that doubling the bandwidth every two to three year will not accelerate the applications, rather just being sufficient to maintain the same performance for a richer and more diverse content.

Dec 2010

Dec 2011

Dec 2012

The whole WAN Governance team joins me to wish you, your family and your colleagues an excellent 2013 year!

 

 Béatrice

(Illustration: "toujours plus" by Gilbert Garcin, 2009.)

Business apps first – but let Internet shine in

The live broadcast of a major ongoing sports event… A colleague ingenuously watching a business TV channel... An internal training webcast that has too much success… The high definition photos from the recent corporate party just uploaded on Flickr… A few among many reasons for corporate networks to be saturated, impacting users' experience with direct impacts on productivity.

• Like the air we breathe, application traffic gets into every free interstice. Throwing bandwidth is like filling a bottomless pit: it will never be sufficient to satisfy the ever growing usage. Moreover the additional traffic might well affect the performance of the business critical and resource thrifty applications.
• Blocking Internet and social media is quite impossible (see our recent post). The Marketing and Sales department need Facebook to interact with customers. The training department posts e-learning video on YouTube. Explaining your generation Y - and soon Z - employees that corporate network supports only business usage is pure waste of time (don't you think it will finally cost even more if they start using their smartphones?).
• Understanding and deciding about the level of resources – and money – the company allocates to business and to recreational usage is a challenge. However it is of the IT department responsibility to justify to the Line of Business about the cost/performance balance.

How to face these challenges in an easy and user friendly way – authorizing Internet and video while ensuring business applications perform at their best?

Autonomic networking is a fantastic answer to this challenge. Its ability to dynamically learn about the traffic and control each user's flow according to its criticality and its performance objectives automatically ensures the harmonious coexistence between business critical applications and recreational traffic:

• Optimized – Autonomic networking radically improves network efficiency by properly organizing and streamlining all flows that share the network. More traffic can be transferred over the infrastructure without impact on business applications' user experience. As a result, bandwidth upgrades are differed by an average of 3 years.
• Open – Confident about their business applications, enterprises can implement an open policy for Internet and social media usage. There is no longer a need to block the Internet, allowing business usage of the public Internet and satisfying employees that cannot envisage an unplugged lifestyle anymore.
• Justified – The application usage is understood. A crystal clear relationship between the level of resource and the application SLAs is established, allowing to rightsize the network and to share with the business the responsibility of the cost/performance trade-off.

Guaranteeing the performance of business applications while accepting Internet and social media traffic without costly upgrades leverages Company's strategy to improve its efficiency and productivity through better collaboration among employees, open culture and closer interactions with its customers and partners.

Béatrice Piquer-Durand

Illustration: Sunset on Rouen by William Turner (1775-1851)