WAN Governance

Why it pays to measure application performance

'On a scale of one to ten…' is a classic prompt that has been used to measure just about everything: emotions, the success of a new film or employee performance during reviews. And why not? The scale is concise. It's simple and succinct. It can capture complex data from a variety of sources into condensed, easily transferable and comprehensible numbers.

Now the 'one to ten' scale is helping classify how applications are performing across the WAN via the Application Quality Score (AQS) metric. AQS provides some basic numbers with big repercussions – numbers that could help businesses build better relationships between IT staff and senior managers, streamline processes, improve customer relations, guide wise investments, minimize downtime, and create some pretty interesting graphs.

What is AQS?

Today there is no standard indicator for application performance across networks – and with good reason: there are very few standard applications. AQS confronts this issue by operating as a composite indicator. By combining a variety of sub-metrics (such as round trip time, server response time, transaction activity and TCP retransmits) and unique 'one-way' network metrics (such as transit delay, loss and jitter), the 'one to ten' AQS score is a top level view that reflects the application performance that remote users experience over the WAN.

How AQS can help improve application performance

AQS acts like a compass, pointing IT departments to where an issue lies and then allowing them to find a solution. Poor application performance can be the result of the network, the server or the application itself. By creating a metric, AQS indicates which of these areas, and where within these areas, any issue is to be found.

For example, imagine the AQS indicates that the network is the cause of poor application performance. IT staff would be able to delve deeper. They would be able to answer two related network questions: is the network itself experiencing a fault, or is the network jammed with too much traffic from multiple applications? All this information is contained within the composite parts of the AQS metric. From there, IT staff could go about finding a solution: if the low AQS score is due to of a faulty network, then the network provider would be responsible for improving it; if the score occurred because of too much traffic over a perfectly suitable network, the IT department would have to unblock the traffic jam. They may choose to do this by determining which applications are business-critical (such as ERP, CRM, videoconference, etc.) and prioritizing them. Applications like YouTube or Facebook may be deemed as non-critical, in which case a company may allow an AQS score of 2 or 3. The same company may want an AQS score for critical applications, like SAP or VoIP, to be at 9 or higher.

To keep with the previous example, if the low AQS score indicated the problem was a result of too much business-critical traffic, then companies would know they need to direct funds towards increasing bandwidth, and better still, they'd have the information to make the business case.

Why IT people need AQS metrics

Conveying IT information about applications to senior officials can often be challenging. KPI, SLM, SLA, WAN optimization – these are all terms that managers and senior staff may not be familiar with. AQS helps make the language of IT more approachable and even more visually stimulating, owing to the visual nature of dashboards. The metric of 'one to ten' is simple and easy to understand. Without such measurements, it's difficult to set application-related goals. Under such circumstances, when issues arise, tension between IT staff and management inevitably follows. Management see IT as expensive and not performing well; the IT department is put in the uncomfortable position of being on the defensive. In contrast, when there is a metric, a clear indicator that simply and easily points to the heart of an issue, it's much more possible to positively respond to arising difficulties. IT staff have data and information they can provide to senior officials.

Additionally AQS improves user complaint handling. When people do complain, IT staff can see the information needed to help users, even if it is something as simple as being able to describe the problem and tell the customer when it will be fixed.

Consider a non-IT example: an individual misses her train. Not knowing when the next train is set to arrive may cause distress. Conversely, knowing how long the delay will be allows the individual to be proactive rather than reactive. Think of the AQS like a digital train board. Using the advanced information on application performance IT departments can better serve users.

(This article has been previously published on Business Computing World)

The changing Hub and Spoke model

The times, they are changing – even for the traditional hub and spoke models of networking. These changes come with serious repercussions: control structures change; companies democratize; security is threatened; efficiency is increased; and the need for control grows exponentially. How did this change happen? Why is the hub and spoke model shifting? Why did it exist in the first place? And perhaps most importantly, what is next?

Understanding the traditional hub and spoke model

Let's go back to the early 2000s. IT directors decided around this time to reduce the number of servers by consolidating everything into main datacenters. This meant placing information, organization and consultation in centralized offices. Consolidation 'fixed' a lot of the issues caused by having many servers across various branch offices, such as the expensive need for extensive hardware and software, and the overall complexity of unifying many separate bodies. Virtualization allowed companies to build datacenters that were flexible, that were at a reasonable cost, and that were fairly simple to maintain. WAN optimization enabled companies to have remote servers without impacting the user experience too much.

How the hub and spoke model is changing

Recently, Unified Communications and Collaboration (UCC) and cloud computing have changed how companies network. Unified communications has allowed companies to communicate through text, voice, and video. This means international, company-wide conversations can happen through one system and one network. As a result, communication has begun to shift away from the 'spoke-to-hub' model to a "spoke-to-spoke" model. Companies are increasingly using more and more applications which are located in the cloud, stored in datacenters all over the world. Branch offices access these applications independently, over the public Internet. In essence, the hub has become decentralized, while branches have become increasingly autonomous.

What the changing model means for IT staff

With increased independence comes decreased coherence, a lack of control, a risk of inefficiency, and endless challenges to the networks themselves. Security threats from the lack of a central monitor watching what's happening across the internet. Performance threatened as many applications compete to run across the network, clogging the system and risking business critical applications. Costs can balloon by users accessing applications without guidance.

IT staff must find a way of drawing together all these increasingly autonomous branch offices to create a streamlined company structure. Increased awareness is necessary to knowing and understanding what actions are happening in the branch offices. Then comes a need to prioritize and protect business critical applications. Traditional methods of traffic control, previously located in the central hub of a company, are too expensive and complicated to manage to distribute them at individual branch offices.

Yet with changing demand has come new technology; recent tools for WAN optimization have been designed for the local office. As simple to install as a cable box, but performing enough to allow quality control across many levels of the business. They allow IT staff to manage and prioritize what happens along networks. Small enough for the branch office, these tools enable centralized control; thus they span the old and new types of networking.

While the times may be changing, not all change is bad. In some cases, such as with the shifting network model, change can allow for improved efficiency, increased flexibility, and overall better business practice.

(A full version of this article has been previously published on Business Computing World)

Autonomic Networking is IT staff’s best friend!

In a recent Network World interview (« Why consumerization won't kill corporate IT"), Forrester analyst Matthew Brown said: "Some people are claiming that as technology is easier to consume and provision in a self-service manner, then the role of IT goes away. That is based on a principle that all IT does is perform low-value activities around maintaining the store. In fact if you look at IT you can see that what they should be doing is more value-added activities such as building new applications and services."

We couldn't agree more, if… Oh, actually there is one "if"... Consumerization trends, such as cloud platforms, bring more complexity to network management – not less. As part of the transition to cloud computing, companies are moving to a hybrid model where some IT needs are best served by cloud platforms and others simply are not. And as IT professionals are figuring out where the cloud can be rewarding and improve IT operations, application delivery still remains their primary focus. Anything and everything go over the network, sharing the very same access to the branch where users work. It's not rare to see recreational traffic using 80% of the line. And non-MPLS usage is doubling every 12 months, for business as well as for personal usage. Local Internet breakouts and hybrid networks that are necessary to maintain the network cost at an acceptable level despite the traffic inflation, participate also to the confusion.

In such a shaky and dynamic environment, to guarantee users' experience is becoming a nightmare that will occupy a huge amount of time and energy (for poor and uncertain results) - and finally prevent IT to move up in the added-value activities Brown was speaking about… Unless enterprises use automated and dynamic tools that are intelligent enough to differentiate each app flows and individually control them – providing application performance continuity whatever the end-users' activity.

With Autonomic Networking - now available on very affordable platform like Ipanema's nano - IT gains full control and optimization of application performance for each of their branch offices, whatever the IT architecture. Hand free, automated, self-adapting… it couldn't be easier. Mission-critical operational and financial applications can now compete with bandwidth-hungry social media and video applications – without impacting users' satisfaction and business efficiency.

And IT people can finally stop running behind problems and act proactively, implementing more strategic tasks like WAN governance – in a word: more aligned to the business.

Autonomic networking: IT staff's best friend!

Béatrice

Illustration: Yan Marcziewski (2007 - no title)

Consolidation: it’s the apps that matter

A post from Nigel Pink, VP and general manager UK, Ipanema Technologies

As a result of acquisitions multinational enterprises often have a plethora of different IT systems. Data centre and server consolidation simplifies complex IT infrastructure and reduces IT cost but increases complexity when delivering applications over the network.

The costs of distributed enterprise infrastructure include hardware, software, maintenance, security, and backup. These costs dramatically increase with the number of disparate systems. Global businesses operate branch offices and data centres at various domestic and international locations and, in addition, are increasingly employing mobile workers. To consolidate IT infrastructure, enterprises typically choose between one or two data centres (particularly for disaster recovery) to which existing applications as well as new applications are migrated. Even though these applications are now hosted in increasingly remote data centres they continue to be used in the same way as before. Practically, this means that enterprise information no longer flows locally in the LAN, but rather over the WAN resulting in slower response times and a considerable increase in data traffic. Ultimately, if unprepared, IT consolidation projects can unwittingly cause application performance to be degraded, impacting user productivity.

More bandwidth isn't the answer to user satisfaction
Bandwidth can be an expensive IT cost. Removing server, hardware and maintenance costs through consolidation loses much of its impact if you then need to double network bandwidth capability to support the distributed infrastructure. As we have found working with lots of organisations bandwidth can be consumed disproportionately by non-critical applications. At one oil company a single user was consuming over 90% of the bandwidth at a site through the use of recreational video applications. A WAN must have controls built in that can allocate resources based on the criticality of the application in question. For example, you must be able to specify objectives such as 'any end user session, anywhere, anytime, must have at least 50kbps if using SAP'. The Ipanema system works at this granular level and that's how we can help organisations to specify SLAs for each application at the individual user level. In the industry this is sometimes referred to as Application Performance Management (APM) but at Ipanema we prefer the term WAN governance.

Consolidation and the WAN
Consolidation of servers and applications means more data traffic flowing across the WAN. As a result application performance suffers due to delay, jitter, and packet losses. According to a study we commissioned in the UK, the cost of poor application performance can amount to additional costs upwards of £500,000 per year, due to reduced productivity, unfilled orders, and decreased customer satisfaction. The research also showed that network bottlenecks resulting from centralisation can generally be eliminated by increasing bandwidth. Although 57% of respondents confirmed this, 75% feel that increasing bandwidth alone does not solve the application usage problems that occur from increased centralisation of IT resources.

Significantly, 93% said that they wish for improved network visibility and control of the data traffic flowing over the WAN.
Access to centrally stored remote information, groupware systems such as Outlook, Sharepoint or Lotus Notes, and important proprietary applications like SAP are all vulnerable to performance issues post-consolidation. WAN Governance allows applications to operate according to their level of business criticality and can guarantee the performance of critical applications at the individual user level.

Governing WAN Infrastructure – it's more than acceleration
Current enterprise services run smoothly in the local environment, but were not designed for operation over a WAN. Using the WAN to access applications significantly increases bandwidth requirements between sites.

To solve this problem a variety of WAN optimisation tactics can be deployed including; compressing data and local caching at various locations, accelerating the transfer protocols used on the network, changing the network topology of information flows. However this tactical acceleration approach alone isn't enough. Imagine a Ferrari travelling round the M25 at top speed – it looks impressive. But now try it in rush hour and the Ferrari can't work to even a fraction of its potential. The same is true of an acceleration system deployed on a congested corporate WAN. Acceleration is part of the battle but it must be viewed within the wider context, which is where WAN Governance comes in.

The WAN governed as a whole thus becomes a solid asset on which the business can rely at all times. When conducting consolidation projects we urge the industry to consider the delivery of applications up front. After all users won't appreciate the IT efficiency savings you've achieved but they'll sure notice if their applications grind to a halt.

(This article has been previously published in Datacenter Management)

Illustration: The gluttonous Gargantua, as depicted by François Rabelais and etched by Gustave Doré.

Running faster with brakes

When the enterprise's network is labouring under too much data, what's the best course of action?

• Just buy more capacity?

This is the natural approach and will ultimately be necessary. But adding bandwidth alone proves to be an end-less game: TCP/IP traffic does a good job at filling-up pipes as much as possible ("flow elasticity") and enterprise non-critical traffic increases faster than the business traffic. It is current that non-critical applications like YouTube or facebook use 80% plus of the enterprise bandwidth - what's the use to just add more resources in such a situation?

• Would network acceleration solutions that can avoid this?

Network acceleration solutions have two effects – one good and one bad. The good impact is that the amount of data is somehow reduced (usually by 20% to 50% according to the traffic characteristic). The less good news is that the extra bandwidth will still be used in most of the cases by recreational applications… So it's putting off the inevitable. And the cost of WAN optimization solutions makes them applicable only for expensive bandwidth, typically international lines.

• What's the other option?

Running faster with brakes! Taking control about the way each user and each application make use of the bandwidth allows enterprises to get sure that the resource that is in place will be used for the sake and the wealth of the company. At Ipanema, we advocate the usage of WAN Governance which allows to dynamically driving each application flow from a business criticality and performance objective point of view and using the AQS (Application Quality Score) to get sure that the application is performing well. This way, business critical applications are getting always the right amount of resources – not less but not more - while the less critical one will adapt from the remaining resource. This allows enterprises to operate their business across very heavy loaded network without experiencing any poor performance and to postpone bandwidth upgrades by an average of 2 to 4 years. Even more importantly, it works particularly well in the complex public and private cloud computing environment of modern networks where the traffic is less predictable than ever.

Whatever the amount of bandwidth you have, would it be fixed or variable, the first thing to do is to ensure its usage is aligned on what matters for your enterprise: the business.

Béatrice.