WAN Governance

Evidence on behalf of unified [VPN + Cloud] networking (3/3): verdict time

…/…

While in a pure VPN architecture, it is often possible to manage the traffic from Datacenters in such a way that critical applications get guaranteed performances and recreational ones become harmless to the business (see for example Ipanema's virtual ip|engine concept), this becomes per definition impossible with Cloud computing: the goal here is to relocate application servers to centers that enterprises do not want to control anymore.

The level of virtualization is such that in most cases it would be illusory and counter-productive to try knowing the physical location of servers and databases. The rise of the unified [VPN + Cloud] network places now the branch office at the center of the game, each branch being a small independent hub for its own traffic, would it remain inside the private network or go to the Cloud.

Features to be implemented in the branch are then:

• Deep understanding of application traffic, including encrypted Web flows that usually support applications that are delivered in SaaS (Software as a Service) mode;
• Ability to adapt in real-time the resource allocation to the instantaneous demand, according to the business criticality and the technical requirements of each of the active applications;
• For multi-homed branches, dynamic network load-sharing to maximize application performance and availability (selection of the most appropriate access) in the one hand, and cost reduction (optimized resource usage) on the other hand.

Of course, all of this has to be performed in a fully automated manner, in order to relieve the technical teams from painful operational duty. Finally, the TCO (Total Cost of Ownership) of such a system must remain marginal compared to the cost of the application services it is in charge to guarantee the performance over the unified [VPN + Cloud] network.

In our mind, there is a way for enterprises to fulfill simultaneously all these conditions and get the performance guarantees they seek: the alliance of the WAN Governance approach and its top-down drive of application performance from a applications SLAs, with the ability of Autonomic Networking to use such objectives to automatically adapt to dynamic traffic demand and varying network conditions.

No doubt [VPN + Cloud] computing and Anything-as-a-Service will be increasingly important to the enterprise's performances. Let's bring them the unique verdict they deserve: lifelong performance guarantee!

Illustration: Paul Newman in Sydney Lumet's film "The Verdict", on 1982.

Evidence on behalf of unified [VPN + Cloud] networking (2/3): witnesses’ deposition

…/…

The branch office situation may be considered from two main viewpoints: the network and the applications. Let's start from the latter:

• First of all, applications are not equal for the business. Some of them must absolutely work perfectly well, in any circumstances: this is generally the case of the business transactional applications (ERP…) and applications related to customers relationship. Some others may just be "good enough": while the delivery of an email with a delay of a few seconds will likely not rise problems, a customer waiting too long at the department store cashier desk or when organizing his next holidays trip at the travel agency, will not only impact the enterprise efficiency but also the customer satisfaction.
• Then, enterprise's network is falling prey of cannibals: studies (e.g. OVUM) show that in many cases +50% of the traffic is of recreational nature (Facebook, YouTube…) while only 15% are dedicated to business critical applications (we found 9% in our own study). I believe this trend will confirm with the ever increasing richness of contents (digital photos came from 2, then 5 and now 15 Mpixels) and the arrival of the famous Y generation in enterprise.

From the network point of view, let's just do the two following considerations:

• The connection(s) between branch offices and the [VPN + Cloud] network(s) is shared between all users and all applications, whether recreational or business oriented, either hosted in the enterprise datacenter or in the (private or public) Cloud. This common resource must then be properly controlled and allocated.
• Multi-homed branches (e.g. one access to an MPLS private network and another one to the Internet, or two accesses to the Internet) may benefit from a very efficient and particularly robust configuration (network load-sharing). In this case, VPN-hosted as well as Cloud-hosted application servers can be reached from each access, with an intelligent per-flow selection of the access ensuring the best performance and availability for the service, anytime.

To complete the description, we can add the highly dynamic nature of the traffic that varies with the very immediate users' activity. Traffic may be estimated 'on average' (e.g. at peak time) but cannot be predicted in real time: while traffic jam can be predicted according to time in the day, it is still impossible to forecast whether a car will suddenly appear right-hand at the near crossroad in the next few seconds. It is then necessary to ally medium-term planning with real-time driving.

…/…

Evidence on behalf of unified [VPN + Cloud] networking (1/3): setting the scene

As Cloud computing and Anything-as-a-Service (AaaS) are clearly on the rise, I would like to defend the need of a unified [VPN + Cloud] networking approach for enterprises. Let me first express a few initial considerations:

1. The « Cloud » is not really new: for the last 20 years, enterprises have connected their central application servers to remote terminals through private (virtual) networks (leased line, Frame relay and now MPLS…).
2. VPNs are not close to disappear: they host the enterprises' specialized business applications and protect their confidential data. They more and more interconnect employees through interactive applications like wiki, forum, voice, tele-presence and enterprise video. Unless enterprises want to send their staff in the sky (which will likely create some justified emotion) all these flows will remain on the VPN.
3. "Cloud computing" as well as "AaaS" are certainly major IT evolutions with true benefits and long term existence. It follows in the Datacenter consolidation's wake that started a few years ago, completed by the power of server virtualization: Cloud computing, it is something like "virtual datacenter consolidation".

When the enterprise decides to deploy application in a Software-as-a-Service (SaaS) mode, or to deploy Virtual Datacenters to host its business applications, it has to handle 2 clouds: the plain old private cloud (VPN) and the new 'virtually private cloud'.

As applications concentrate on a few central locations, these clouds (or more precisely communications across them) are more and more crucial to users who do not care about technical considerations. For them, it is straightforward: there is 1 Information System and 1 Network. As the vast majority of collaborators are using the Information System remotely, they need the guarantee of good application performance over the network(s), anywhere, anytime.

Finally it has to be noticed that the cost of WAN is now very low compared to the cost of applications: the price for a VPN access to a small branch office with 10 users of 10 applications is similar to the license price for 1 application and 1 user (e.g. Salesforce or GoogleApps): not really worthy to take a risk here! When buying a family car, even if we decide for a low price model, who would build savings on the quality of the safety belts?

…/…

(PS: I'm not sure if you like clouds, but you should like this mist…)

 

Mehrabian’s rule, networking and teleportation

In the 70's, Albert Mehrabian studied the relative importance of verbal and nonverbal messages. The three elements of linking people together in a face-to-face situation are abbreviated as the "3 Vs": Verbal, Vocal and Visual. Mehrabian's findings on inconsistent messages of feelings and attitudes have become known as the 7%-38%-55% Rule: 7% of the message goes through words, 38% through tone of voice and 55% through body language.

Models for Call center activity where no visual contact is possible, propose 18% for words and 82% for voice quality (including tone, inflection, pitch, rate and volume). Note that these figures are consistent with Mehrabian's rule: when we remove its visual part (55%), words represents 7/(7+38) = 16% of the remaining communication channel.

While these figures have been discussed (mainly because of their dependency to the context), they clearly emphasis the importance of non verbal communication in human relationship. This is why it is so difficult to work closely with people you never met (or even hear if you're just interacting through email).

Let's compute how Mehrabian's law is distorted by networking:

• The average talking speed is 120 to 150 words per minute, while the average word length (in English) is of 5.1 characters, 6.1 if we add a separator. So a standard conversation takes approximately 150 words per seconds * 6 characters * 8 bits per characters (ASCII) = 7.2 Kbit/s.
• Voice transmission requires from 8 Kbit/s (low quality codec, like in mobile telephony) to 64 Kbit/s for good quality and even higher for hifi coding.
• Videoconferencing and tele-presence systems use from 250 Kbit/s (low quality) to 4 Mbit/s (High definition) and 15 Mbit/s for a multi-screen conference room.

   

For a "good enough" communication involving 7 Kbit/s for words, 64 Kbit/s for voice and 2 Mbit/s for a one-screen teleconference, we then find the 0.3%-3%-97% Rule that gives the % of the bandwidth allocated to verbal, vocal and visual purpose. It is interesting to see the efficiency of each channel:

• 7% of the information (words) is conveyed by 0.3% of the bandwidth --> Efficiency ratio of 7/0.3 = 23;
• 38% of the information (voice) is conveyed by 3% of the bandwidth --> Efficiency ratio of 38/3 = 13, approximately half of words' one;
• 55% of the information (image) is conveyed by 96% of the bandwidth --> Efficiency ratio of 55/96 = 0.57 (i.e. 40 time lower than words and 23 time lower than voice).

Would we accept the above figures, no wonder that historically people started with exchanging written messages which remain the most efficient communication channel. It took 1000s of years to come to phone calls that still have very impressive information to bandwidth ratio. 100 year later, imaging technology (codec, screens and camera) and networking (bandwidth, quality of service, application SLAs and Wan governance) make videoconferencing and telepresence possible.

Let's be ready for teleportation in a few 10s years…

(Illustration: see no evil, hear no evil, speak no evil at Toshogu, Nikko, Tochigi Prefecture, Japan)

PS: there is a notable exception to Mehrabian's rule: literature, music and painting/sculpture, when talented artists are able to bring us - just with words or sounds or colors or forms - in their own world and create into our brain images, impressions, feelings, souvenirs and thoughts. In the other hand, one can say that as this communication is not bi-directional, the rule just does not apply...

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.

(…/…)

When queues mean queues

As I am often asked to share practical examples of the direct relationship between application performance and enterprise business, I like to share here one of my favorite cases.

This particular business is in the retail market, with many supermarkets full of shoppers. The IT system is used for two very different purposes:

·         Logistics and administration (stock management, pricing, ordering, etc.)

·         Customers’ electronic payment between local terminals and a central application server.

Let's talk about the latter. While it is easy to understand the importance of this ePayment application’s availability for the business, it is even more interesting to drill down into the differences between ‘good’ and ‘bad’. It just so happened that the IT team was smart enough to build a model that directly relates customers’ average payment waiting time to the application response time. This study clearly shows a few thresholds (all have been reached in the past):

·        When application performs well, the waiting time depends mainly on the number of people queuing up. It is managed by varying the number of open payment terminals (opening all of them all the time would cost too much without real benefits, so it is important to match the number of operators to the number of customers present in the supermarket).

·         If the application transaction duration degrades to a certain level, then customers are obliged to wait in line for much longer: their Quality of Experience (QoE) is poor. They feel angry and start complaining loudly: this is not good for the image of the retail chain.

·         When it gets even slower, the most impatient customers start abandoning their shopping carts full of goods. Employees then have to throw out the fresh goods (meat, vegetables…) and replace everything else on the shelves: losses, extra work, less business (and probably lost customers)…

·         The bottom level of poor performance creates such a pressure on the supermarket employees that they are be forced to open the doors and let people carry their stuff away, without paying for it (could be good for the image, but not for the business).

We can see that good and stable application performance is not only a phrase and goal for technical IT and network engineers, but is directly proportional to the enterprise’s business, its image and finally its bottom line. I’m sure you will be able to adapt this story to your own business, and I will be very interested if you would like to share this with us.