Measuring the actual not the theoretical
Picture yourself sitting at a road traffic controller’s (imaginary) dashboard. There would be red lights showing problems at traffic lights, roundabouts, road closures, accidents etc, and green lights indicating everything is fine.
These systems are great for reporting the state (or availability) of the infrastructure just like a service management dashboard. What they fail to show is how long a particular driver’s journey is going to take. And so it is again with service management dashboards. In essence, there is an obsession with availability which completely ignores the user experience: the journey.
When I first started out in the application performance monitoring space, the majority of tools developed were for monitoring systems and application availability from a server perspective, using tools such as Tivoli, BMC Patrol, and Microsoft MOM etc. Later, products like HP Openview would monitor network devices, displaying maps showing “red lights” when there was an issue. Sound familiar?
While these tools were great at telling you whether or not an application, server or network component was available, they didn’t help you understand the user’s experience of accessing and using that application.
And as we recognize today, service management is not just about how available the technology is but making sure the customer is happy with their experience of it. An application or service can be showing on the monitoring panel as being 100 percent available but responding so slowly as to be regarded as being unavailable as far as the user is concerned. At best, this could be regarded as an annoyance but if application performance is really poor there are productivity implications for the end-user organisation and possibly financial penalties for the service management provider. Indeed, even for ‘internal’ applications poor performance simply leads to non-use.
Add to this, a move to datacentre consolidation/virtualisation, cloud computing, branch office initiatives and home service applications and you will find more and more applications today are expected to run across WANs, Satellite, 3G, GPRS, WiFi etc and this adds a very significant extra dimension to the far simpler problem of internal monitoring of application performance and availability. After all, before the external network arrived most of the factors in application delivery, number of users, server capability (CPU, Memory, Disk…) were understood and controlled. Now it’s like making a journey from Manchester to London – how long will it take? Who knows?
I remember the software industry entering into the uncontrolled world of network delivery of their applications under guises such as application service provider (ASP) or on-demand software and later renamed SaaS (software as a service). Generally these were supported by the service industry in the form of ‘hosting’. Equally, companies may have decided to externally ‘host’ some of their applications. Both imply a network (usually WAN, sometimes Internet) delivery of the application with all the uncertainty that brings.
Now we have cloud computing which is fundamentally SaaS, platform as a service (PaaS) and infrastructure as a service (IaaS). All will be ‘hosted’ and delivered over networks as uncontrolled as the Internet. This is an important marketplace with estimated market revenue in the USA of over $4 billion. If you are in doubt about this brave new world, look at the success of salesforce.com.
So, companies today are looking for more from their service management providers. There are more mobile applications needed for remote workers, to be delivered on mobile devices. It has been estimated that more than 52 percent of corporate workforces are now working outside of headquartered offices and so there is an increasing need to guarantee the performance of applications no matter what devices are being deployed. Recent adverts for the iPhone bear witness to this trend with an increasing number of business-oriented applications being made available on this device.
Unfortunately, there is a whole world of difference between how an application runs in a LAN and how it runs over the WAN, satellite, mobile 3G/GPRS networks etc and it’s not just about increasing bandwidth in the event of performance problem as ISPs would have you believe. There are other important factors such as latency (delay), losses and congestion in the ISPs core network or in the client end of the network and the service management providers need to have a focus on this.
Networked applications are very much at the ‘Core’ of the things we do every day, both professionally and personally. Examples include accessing our bank accounts online, booking travel tickets, social networking (Facebook), accessing web-based CRM systems such as Salesforce, controlling traffic lights, modern IP based public CCTV, and smart phone applications. So, keeping tabs on the performance of networked applications is just as essential in service management as it is to the end user.
So, from being of negligible interest network performance must now rank with the other major system resources of CPU, Memory, Disk I/O and storage. It’s the least controllable too, so at the very least we must make it visible and understandable.
While solutions already exist that enable service providers and IT departments to understand application quality of service and response times, typically these have been large, appliance-based systems that are prohibitively expensive. Some designs are centred on network protocol analysers or ‘packet sniffers’ that capture huge quantities of data, some of it commercially sensitive, and much of it irrelevant to pinpointing and resolving service availability problems. So what’s the answer?
Recognizing the cost-conscious times we operate in, one possible solution is AppQoS Live! This is an example of a networked applications monitoring solution that can be readily accessed, is easy to use and can be deployed on any modern server or even laptops and PCs, without the need for installation. It enables any IT specialist or consultant to investigate issues such as: application usage; application response time; network performance; and quality of service, all in real time.
It initially starts delivering information regarding application response times, transfer rates, bandwidth usage, etc, from a ‘satellite’ level perspective but then allows its users to drill down in order provide detailed analysis of the root cause of the poor performance. For example, it can help determine:
• Whether it is a network or server issue;
• Whether the application really is performing badly / worse than normal;
• Who is effected (single or multiple users);
• Where in the network the problem exists;
• What is running over the network;
• Whether other applications or traffic types are responsible;
• Is the other traffic relevant or undesirable;
• What other requests are being issued to the server.
When placed into a rack mount server, laptop, or PC, the software transforms it into a powerful application monitoring appliance. There is no need to install any software and the operating system (Windows, Linux, Unix etc) or other applications on the host machine are not impacted in any way.
Next, this system would be plugged into the network segment of interest and with no configuration, would provide concise, organised and easy-to understand views of key network activity data within minutes. Data gathered would be stored in a folder on any existing hard drive or on other writeable or flash media such a USB stick or SD card. This data could be exported for inclusion in reports, presentations etc. If it is no longer needed, it is simply removed from the host machine which then can be used for other activities.
The real benefit is that there is no need for a dedicated appliance in order to achieve sophisticated networked applications monitoring, so significant savings can be achieved. You can choose the hardware (within limits) from your regular suppliers. At the same time, on the portable front, having a team of consultants equipped with copies of the CD also means they can respond rapidly when service level issues arise, rather than waiting for the right appliance to be set up or re-configured. As a result, it should be possible to resolve issues more efficiently, leading to improved customer satisfaction levels.
If you’re offering any kind of hosted, centralised or managed service then managing the service in a cost-effective way is important. Solutions like AppQoS Live! allow for greater utilisation of the standard hardware you have invested. At the same time, the issue of dedicated hardware appliance obsolescence is completely overcome. This coupled with the substantially lower acquisition cost, less hardware management costs, lower training costs and ease of deployment, amount to a substantially lower TCO and hence a greater ROI than alternative and proprietary hardware based appliances, a real win for anyone offering a service where networked applications are part of the equation.
