WAN Governance

People matter, finally

In 1774, Claude-Nicolas Ledoux started the erection of The Royal Saltworks at Arc-et-Senans with goals not limited to build an efficient industrial complex to produce salt in large quantities. In top of a rational design to limit carriage and provide easy surveillance points, he installed comfortable houses and annexes for managers and administration officials, but also for workers and their families (including a chapel as well as private gardens to provide an appreciable complement to the usually limited and monotonous food).  

Technology and processes are not isolated from people that serve them; they need to take into account their motivation and their culture. In its turn, Application networking requires a global approach that embraces all the different aspects of networking, from performance to cost, as well as a way to improve the life of IT and network teams.

• When an end-user complains about application performance, IT teams usually lack visibility to understand properly the situation: is it a server, host or network issue? Is it the application itself that fails to deliver the right performance? Is the complaint justified or is the end-user in a bad mood because he has problems a bathroom that leaked during the night? Answers are then vague which adds to the frustration and facilitates the tendency to frequent unfair pin-pointing that make daily life difficult.
• When the CIO wants to clarify the impact of the network to the business, he gets approximate answers. Justifying the network strategy, performance and cost is painful. The network can easily be perceived as always too expensive while not performing enough. Deprived of high-level information, IT/network teams have hard time to convince about their choices; confidence is limited: it is difficult for them to become their management's trusted advisers.
• Being reactive to problems, support and troubleshooting is performed in urgent and critical situation, with stressed and angry users. Operations take most of the time of IT/network teams that have to run and put out fire at every place. This confines them to never-ending and quite demoralizing low-level tasks.

Enterprises that implement WAN governance processes and tools get a clear access to the performance impact of the different elements of the IT chain. KPIs and application SLAs enable their IT/network team to provide meaningful information to CIOs and other company executives. Associated to an Autonomic networking system, operations become proactive and most of performance brownouts are automatically solved.

Released from the 'quick and dirty' part of their job, IT and network teams can dedicate most of their time, talent and energy to higher level tasks, like long term planning, cost optimization, internal invoicing and application SLAs management. Providing high level information, working in the future and not only in the present time, they improve their value for the enterprise as well as their personal motivation.

People matter, finally. In one sense, modern Wan governance is to Application networking what Arc-et-Senans was to traditional saltworks, 250 years ago.

Optimize AND Maximize (2/2)

…/…

Desire for speed in application delivery comes from three significant drivers: the natural impatience of users, the richness of modern applications and the rationalization of IT architecture.

• We, as human being, are impatient animals. Knowing that our time is limited brings a huge psychological pressure on each of us. Waiting more than a few seconds behind a static screen (web specialists said 8 seconds 10 years ago, they say 5 seconds today, what will it be in 10 years when a new generation of young people accustomed with electronic gaming and all kind of frantic toys will enter the business?) creates a sense of frustration that makes users unhappy.
• Modern applications become richer and richer. Sophisticated Web design tools enable software editors as well as enterprise process designers to write applications with complex interactions between clients and servers. Fancy screens with embedded images and video increase the amount of information to be exchanged.
• IT centralization is a fundamental trend of today's modern IT architecture. Enterprises need simplification, for cost reason (less hardware, less operation staff, less software licenses...) but may be more fundamentally to simplify operations and concentrate on their main business. Virtualization, cloud computing and modern networking techniques (like Wan optimization and broadband accesses) make it possible.

Enterprises live in a permanently changing environment, where technology, finance, products, partners, customers, competition are moving faster than their production infrastructure and people's culture. IT has been and will remain a key enabler of adaptation to these fluctuations, and supports most of business processes of modern enterprises. This is the source for such a strong need for certainty in IT: managers expect a predicable efficiency that they can rely on and adjust to evolving requirements, at each minutes of every hour of every day.

The classical technical approach to enable enterprise face their 'optimize AND maximize' challenge is based on specialized solutions (e.g. acceleration, shaping, SLAs, etc.). While each of them provides what they have been designed for (no more, no less), none is able to solve the entire problem – for example accelerating an application would likely put the other ones at risk, just like a too speedy car will create accident on the motorway. Moreover, the cost and complexity of combining solutions are way too high to consider it as a good candidate in most cases.

It is then natural to select technologies that integrate all the required features in such a way that they could closely interact to provide the expected benefits (speed and certainty) at the same time. Combined under the 'always tuned' umbrella of a well-designed Autonomic Networking solution, they provide an exceptional association of stability with high level of performances. At least, a 'martingale' that works?

Optimize AND Maximize (1/2)

What does she prefer? Finishing her ice-cream till the very end (and - why not? - expecting a second one) or leaving some room for a reasonable dinner?

What would you prefer? Having a chance to arrive at the office in a record time or be safe and relax for your morning business review?

Optimize versus Maximize is a question that arises in most of the situations we live: family, business, industry, wildlife, etc. Fine tuning the delicate relation between these two close notions is at the core of multi criteria challenges that we daily face, as soon as external constraints (time, space, money, law, demography, natural resources…) oppose to our primary goal.

In aquaculture, maximizing fish growth is the main goal, but this must be achieved through optimizing cost of feed, labor, time, health, filtration among many other parameters of premiere importance.

In the famous car race "Le Mans 24 Hours", technical teams strive to maximize their top speed in the Hunaudières straight (ready for a tour?) and get a fair chance to win the race. On the other hand they also need to optimize their strategy and balance between speed, reliability, pit stops and pilots' tiredness, as the unusual race length will overstep classical limits.

100 million years ago, life evolution maximized Dinosaurs' size, strength and appetite. It did not optimize their ability to adapt to a fast changing environment – sadly for them, happily enough for our mammal ancestors.

Maximization has the beauty of a single, simple goal. It is usually meaningful in a (relatively) short term purpose, where we do not have to care too much about other criteria. It often goes along with some gambling mindset: when someone wants to maximize his profit, Las Vegas, Monte Carlo and Wall Street are good places for a stay.

In the same situation, optimization appears as a most complex grail as it takes into account many criteria at the same time. It is normally associated with a longer term vision, where stability and predictability matter. Optimizing your family incomes and being reasonably sure to be able to pay for your kids' high school will require more than roulette and black-jack.

Concerning application delivery over enterprise networks, we often read about WAN optimization and application acceleration. Even distorted by the magic effect of b……t marketing, words are still able to carry some underlying reality: the desire for speed and the need of certainty.

(…/…)

With rhyme and reason

 The other day, I found in Paul Jaillard's blog a sentence that nearly perfectly summarizes what we try to achieve with Wan governance: "When we concentrate on costs, quality decreases - When we concentrate on quality, costs decrease"(*).

   

A brutal financial approach is a sign of an organization in crisis. It has not been able to adapt in time to the economic situation, for internal (poor management, limited quality of products, unreasonable cost structure...) or external (technology revolution, credit crunch, regulation changes...) reasons.  Survival is now the only concern, the true purpose of the enterprise (manufacturing goods, servicing people, etc.) being shadowed. While in enterprises' life, it may happen that situation is so difficult that such emergency decisions must be taken, it is crystal clear that the cost of these savings is extremely high. We cut an arm and a leg to save the body, fair enough - but no wonder if the body will remain handicapped for a long time.

   

A pure performance approach might be the right one in early stage of an organization with deep pockets: it lies so far away from its goals that any mean to move it forward is welcome, whatever the cost. Here, the priority is to maximize, not to optimize. The bad news is that this policy can only apply for a short time, as the enterprise will either fail - and run out of cash - or succeed and reach its adult stage where external factors (market size, social environment, natural resources, etc.) will start to limit its expansion.

   

Mature organizations would rather adopt a thoughtful quality approach that balance expenses and benefits, short and long term impacts, direct and indirect improvements. They should praise managers that think and act about what is important for the enterprise (its customers, its products, its financial health…) not just for their own team or career. Of course, it may be seen as a more difficult road, as many parameters have to be taken into account and new questions will arise (starting with an accurate identification of the important needs). However, well aligned with the enterprise's goals and successfully implemented, it makes the difference between good and great organizations.

Applied to network governance, it is the philosophy at the core of autonomic networking and objective-based management systems that provide automated tools to allow teams in charge of the network to spend their time on the objective definition and achievements rather than on the day to day implementation.

The economic impact of good performance across WANs is huge: there is room for amazing improvements here: IT flexibility, users' productivity, bandwidth cost, operation workload, incident reduction, etc. To maximize the benefits, all these aspects have to be optimized globally. I expect to provide you with some examples and figures in further post that would illustrate how to save real money with WAN governance.

(*) From "Le client retrouvé" by Philippe Detrie

Once upon a time... (What are the key applications? 3/3)

Discussing how enterprises organize their applications, we previously saw in this study (Part 1 and Part 2) that on average:

• Enterprises differentiate 69 applications;
• They group these applications in 15 Classes;
• 3% of business critical traffic requires the same analysis accuracy than 54% of low critical traffic

We will now try to understand how the nature of the applications impacts the classification. We sorted the applications in three categories: Real-time (applications that are sensitive to delay variation); Transactional (application with a direct interaction between servers and people) and Background (applications that aim mainly to transfer large data between computers). Here are some typical examples of such applications:

• REAL-TIME: Voice, Videoconferencing, NTP…
• TRANSACTIONAL: SAP, CITRIX, RDP, Internet, Collaborative…
• BACKGROUND: FTP, Email, CIFS, Antivirus…

When we analyze how the type of these applications has been taken into account by enterprises, we found the following results:

• The number of User Class is balanced between Transactional and Background applications (6 to 7 Classes for each), while we get an average of only 2 classes for Real-time applications: the diversity of the application in an enterprise IT infrastructure comes from the data applications, not from the real-time one;
• The number of applications follows the same rule: 5 Real-time applications to be compared to 35 Transactional and 28 Background applications. This outcome is in sync with the former one: few real-time applications while transactional specialized applications blossom across the infrastructure…
• … This is also confirmed by the volume of traffic: 4% for Real-time, 50% for Transactional and 46% for Background. This result being more surprising, as I previoulsy thought that Background and Real-time applications would represent a volume much larger than the Transactional traffic.

Now, if we look at how business criticality and nature of application combine with the volume of the traffic across the enterprise's network we see that:

• Real-time applications have been assigned to the TOP criticality level. This is clearly due to the fact that they are not elastic applications. There is no gray zone: a phone call works or not, that's it. So when an enterprise deploys voice or videoconferencing across its data network, they have to work. Period.
• Transactional applications are spread across criticality levels and concentrate (in volume) on the LOW criticality level. Even with human to machine interactions, they are not all that important to the business: some can wait a bit in case of too large usage of the shared network;
• Background applications have only MEDIUM and LOW criticality. This was clearly expected. Even if a file or an email can carry very important data (and honestly this is not always the case…), to postpone them a few seconds will usually not make any harm to the business.

 

Will all of this be stable in the future?

I'm convinced that the need for thin-grain classification will remain or even amplify: the need for control and network governance is increasing fast. As for the volume breakdown, we can probably expect that Real-time traffic will grow significantly because of VoIP and Teleconferencing (you know, the saving on travels, etc…). On the other hand, we also see that Web applications use richer and richer contents, files size (thanks, Microsoft) and email attachments either. Imho, we must remain humble and accept that we do not know exactly what will happen; this is why we expect to run all of this in a yearly basis.