Go faster stripes

Poor app performance has a negative impact on a business’ bottom line. The examples are endless. Retail high street shoppers abandoning baskets if the checkout queue is too long and too slow. IT helpdesks flooded with calls if enterprise apps begin to crawl. Even complex B2B processes that link multiple apps and data sources together that eventually timeout – leading to transaction failures or time consuming retries. The world is increasingly reliant on application performance remaining consistent – even as user demand fluctuates.  

Apica, an APM software vendor conducted a survey among internet users in the UK, US and Sweden, to investigate changing attitudes towards a brand’s digital performance. The survey of 2,250 consumers reveals that nearly 40% of respondents won’t wait more than ten seconds for a website to respond before navigating away. One in nine users (11%) won’t even give a site five seconds before moving onto another website. Negative digital experiences are also likely to impact brand reputation with 83% of global respondents reporting that they would consider telling colleagues about a poor website or app experience.

Varied root cause

The reasons for poor application performance are as varied as the apps themselves. One of the common causes is poor application design with apps that are often built without consideration for scaling. For example, an application that pulls real-time data from a centralised database may perform fine when demand is low – but as the number of concurrent queries rises, the database becomes a chokepoint and performance across all users starts to drop.

Another common issue are apps that are running on a common infrastructure that are impacted by an adjoining app or resource constraints. For example, a critical enterprise app that is impacted when large file transfers are taking place. Or a slowdown that is triggered by a legacy batch processing tasks that needs to take place at a specific point during the day.

But often the biggest issue is the network layer. Most apps work within a client server methodology where the bulk of the app runs on a centralised server and the remote client is more a presentational layer – displaying information via a network connection from the core. As a result, high network latency, low bandwidth or sporadic outages can have a major impact on app performance.  Network issues can be the most difficult ones to pin down as most applications are network unaware and work in the same way irrespective of the state of the underlying connectivity.

There are a few well-known apps, for example Gmail, a widely used email client that will switch to “lite” mode that uses less bandwidth when it detects a low bandwidth network connection.

In many cases, the causes of chronic application performance issues are down to multiple factors that when combined lead to issues. And understanding the issues is the first step to solving the problem.

Visibility then action

For many organisations, app performance issues are first highlighted by the users of the app. A rising tide of calls to a help desk is one common indicator. For B2C, the wakeup call could be a major drop in user interaction, completed transactions or decline in requests for fulfilment of a service. In the B2C example, getting to this point is often far too late for many organisations as the revenue and reputational damage has already been done.

Many organisations choose to place ‘canaries in the mine’ and install agents on each application server within the infrastructure and collect various metrics and statistics from application servers. These are supplemented by client-side agents that measure user experience. Although comprehensive, this is not always possible for certain server environments and may come with a performance penalty. Alongside this approach are client-side java scripts and applets that can report back on browser-based application performance. These vary from open source tools offering basic page loading speed stats all the way through to full-blown Application Performance Management tools that offer detailed information to better pin down the cause of slow down - but at a price. Another option is to gather metrics from the network layer. This is more useful for enterprise applications where an organisation has access to the routing and switching equipment at the data centre and at branch offices.

Ascertaining that there is a problem is the key and the insights uncovered from this phase will be critical in working out a plan of action. One of the biggest causes of application performance issues is scale. An app that works fine with a few hundred users, but becomes unresponsive with a surge of thousands of users, is often down to the limitations of the underlying infrastructure. This is an area where the cloud has excelled in allowing server-side resources, databases and connectivity to scale up based on demand. However, legacy apps may need to be redesigned to take advantage of the cloud compute paradigm to gain inherent scale to overcome performance issues.

Overcoming network bottlenecks

When it comes to performance issues that are due to network connectivity, there are several potential solutions. However, not all networks are the same. For larger organisations, network connectivity is typically an aggregation of several different network types to meet the needs of different use cases.  A network architecture may use combinations of multi-protocol label switched (MPLS), leased line, Ethernet, fibre broadband, digital subscriber line (DSL), cellular 3G/4G and/or Wi-Fi that change depending on region or site function such as data centres, call centres or remote branch office.

The network admin has a number of tricks up the sleeve to remediate performance issues, including ensuring that compression and data deduplication technologies are built into the network, plus more active Quality of Service measures to define how certain categories of traffic must be handled across different types of network links and use cases. Although some of these elements are already built into the network elements; the more granular capabilities are defined within the broad category of Software Defined Wide Area Networking (SD-WAN).

Software Defined approaches

SD-WAN simplifies the management and operation of a WAN by separating the networking hardware from its control mechanism. This allows organisations to dynamically adjust network behaviour and resources based on exact traffic demand from applications or to meet critical workflow requirements.  This is a major advantage for enterprises as it is based around business rules, which may need to change quickly, rather than hard coded routing and switching tables.  SD-WAN adoption offers the biggest benefit in terms of simplification for enterprises that have a lot of disparate networking connections, application and data flows across private, public and cloud based applications.

In addition, many of the security functions such as secure access, firewall and VPN that are typically separate elements can be integrated into the SD-WAN function to provide more end-to-end visibility of every user and application. This approach is becoming more popular with the 2019 Software-Defined WAN Survey by IDC estimating that 95 per cent of enterprises expect to use the technology within the next two years, with almost 40 per cent having already deployed it in some form.

Replace or upgrade?

SD-WAN is not a magic bullet that will fix every app performance issue and there are also some trade-offs to consider. The reason for the longevity of WAN technologies such as MPLS is its perceived reliability and its powerful, built-in QoS capabilities. Although SD-WAN can apply some of this QoS to dumber networking connections such as last mile DSL, it won’t necessarily overcome the fundamental weaknesses of unmanaged circuits. SD-WAN also comes in two very distinct deployment methodologies. The first is a rip and replace option where an enterprise must replace its core routing and switches architecture to be able to gain all the benefits of being able to use lower cost network connectivity with MPLS-like QoS. For many organisations, this requires major CAPEX investment plus a potentially complex, multi-month project. The other issue to consider with a replacement strategy is that many of the SD-WAN solutions are proprietary and require all network elements to be from the same vendor which limits potential future flexibility and upgrade path.

 An increasingly popular approach is a hybrid or overlay SD-WAN which uses existing networking equipment that is supplemented with an SD WAN layer that adds more intelligence to existing networking architecture. This approach has the advantage of allowing a faster time to market and protection of existing CAPEX while delivering the most critical advantages such as policy-based traffic management, path selection of more granular QoS.  The overlay approach also has the advantage of not being tied to a single network infrastructure vendor which makes future upgrade options more flexible.

And finally,

Maintaining application performance is an ongoing concern that requires a continual cycle of testing and adaptation to ensure the best possible user experience. Irrespective of which combination of technologies and approaches are utilised, organisations must ensure that visibility is maintained as applications change and user demand fluctuates. Technologies like SD WAN offer an elegant solution to overcome common networking related issues and should be part of the consideration process for organisations of all shapes and sizes.