 |
Too Many Heroes working too many hours? |
 |
Virtualization performance inconsistency? |
 |
Users failing in the dark? |
 |
Whose Problem is it Anyway? |
The Age of Causality
Prelert arrives at an epoch of IT and Telecommunications service delivery. Never before has the provisioning and implementation of services been so cost efficient while the cost of service assurance spins out of control.
The adoption of highly virtualized server and switching infrastructures introduces a level of management complexity that cannot be addressed by traditional incident management technology and process.
Where in the past, operations could be sure to know about an issue before a user called allowing a level of customer experience containment, today, application errors occur without any warning leaving operations looking like the deer caught in the headlights when the users call.
Too Many Heroes working too many hours?
Where would we be without our Heroes? Frankly, we would still be looking for the needle in the haystack. When an application becomes unstable and users are affected, seldom does the symptom correlate with the cause, leaving multiple Heroes from separate IT management silos to collaborate, often for hours, to isolate the causality of the problem.
In the financial services and telecoms industries, service level agreements and capacity management regulations mean that the causality must be isolated and reported upwards – meaning expensive specialist resources tied up for many man hours – because until now, only subject matter specialists, working together, have been able to perform causality analysis.
Prelert automatically isolates the causality of application errors, in real-time, unencumbering the specialist resources.
Virtualization performance inconsistency?
So you’re realizing the economic value from virtualizing your service delivery platform, but now you’re finding that at key times your resource capacity is constrained slowing down your applications and frustrating your users and customers, however you’re unable to isolate the causality of the issues.
The problem is quite simple, when the incidents affecting your customers occur, it is very difficult to organize the collected management data and identify the significant related attributes at the specific time of occurrence across all the service delivery silos to enable you to determine the causality of the issues and avert the problems in the future.
Prelert consumes fault, trend and usage data from across the service delivery infrastructure and organizes it into significant Episodes of Causality. Prelert operators simply select the time period and affected applications to be presented with the significant causality episodes at that time period allowing them to understand how their System is working. In this way, virtualized environments can be optimized and organized to deliver the capacity necessary all of the time. Prelert can also warn when that capacity envelop is likely to be affected.
Users failing in the dark?
Unless you’re an anthropologist, it’s unlikely that you like to be the first to find something – especially when it comes to IT and Telecommunications problems. Let’s face it, if you call the help desk, the last thing you want to hear is “Nope, there’s nothing wrong with your application” – because it’s unlikely you would have called them if there wasn’t.
That’s how it is with Application Errors. In modern infrastructures, applications fail in the dark. Mostly because they’ve not actually failed at all, the application has simply become momentarily unstable and the user suffers the effects; popping and stretching on a VoIP call, no response to mouse clicks or typing in an application, no confirmation of a Trade or Credit Card transaction.
Mostly these issues are transitory, one moment they exist, the next performance is OK again, but sometimes the issue can be catastrophic and the users are ejected from the system.
There have been no ‘events’ in the fault management dashboard, no root-causes, because there have been no real failures – after all, our modern infrastructures are highly resilient to outright failures. The application behavior anomaly has been caused by a series of apparently unrelated changes of component state through the service delivery infrastructure.
Prelert offers an early warning to operations staff. Prelert delivers pre-emptive application behavior exception warnings as high severity alerts to the existing fault management and helpdesk dashboard allowing operations staff to catch the issues before the users call and proactively manage their customers quality of service.
Whose Problem is it Anyway?
There was a time, many years ago, when operations management was able to effectively work in silos. When incidents occurred they could be directed immediately to the appropriate team for isolation and resolution. Then came virtualization, mesh networks, caching, distributed applications, portals, VoIP, SOA…
Today, when an incident occurs that shows up as an application error, inevitably many operations groups are mobilized because the incident appears to be caused by needles in many haystacks. It looks like an application issue but the application is still working – so whose problem is it anyway?
We role out the Heroes from each of our silos of management and they must work together to, firstly, identify where the responsibility lies, then, isolate the causality, and only then, resolve the issue. The nature of application errors means that our time to resolve the incident is typically 300% greater in modern infrastructures.
Prelert isolates the needle to the appropriate haystack, ensuring that the appropriate operations silo is immediately mobilized to resolve the issue and the sequence of causality from the needle to the application error is presented allowing operations to review whether changes in the service delivery platform are appropriate to ensure the incident does not recur.
Example Usage Cases
Financial Services Market Data Application
You rely upon your stockbroker or e-trading service to be available real-time, 24/7 so that the moment you want to change your portfolio, it can be done. So imagine if the market data application that handles your position freezes as your portfolio turns red or, does not appear to complete that winning trade.
Prelert is helps investment banks protect their ability to trade, in real-time, 24/7, by automatically notifying operations to any issues which may adversely affect the behavior of their market data systems such as the Reuters RMDS. It’s a volatile world and Prelert enables operational stability.
Outsource Infrastructure Cost of Ownership Audit
Outsourcing services is a financially tricky trade-off. On the one hand the company with the infrastructure wishes to capitalize on the economies of scale of a large IT management partner while being secure in the knowledge that their planned capacity requirements are included within the contract, however the outsourcing partner wishes to ensure that they can accommodate the future capacity requirements of their customer within their costs envelop.
Today, the only approach that works is meticulously slick research and a very experienced ‘finger-in-the-air’ all backed-stopped by restrictive change management contractual terms. Often this leaves future problems for both parties and, nothing protects either company from the loss of in-depth knowledge of the infrastructure through staff attrition.
Prelert allows IT Management outsourcing companies to perform a non-invasive onsite analysis highlighting specific problem areas within the service delivery infrastructure which may give rise to future capacity or application behavior problems, before the contract is signed, allowing both parties to craft an agreement protecting both of their interests into the future. This analysis can be run periodically as and when new services or changes are required to the infrastructure to ensure IT spending is appropriate and optimized to the outcome.
Time Dependent Data Processing Service
When your business is dependent upon the timely processing of data harvested from and delivered back to your clients, and where minutes of accumulated delay withon your infrastructure translates into missed service levels and disproportionate financial penalties, the ability to see into the cloud becomes critical to your profitability and customer retention.
It’s all very well having alarms triggered if data is not produced within a specific time window, but heroic operations staff still have to hunt through several haystacks of management logs and telemetry to isolate the causality of the backlog and, then resolve the issue, adding precious minutes to the process and increasing the risk of missed SLAs and punitive penalties.
Prelert isolates the behavioral anomaly to a causality episode pointing the operations staff straight to the problem enabling them to resolve the issue before the service level is at risk.
VoIP and Video-on-Demand Services
Nothing more frustrating than a telephone call over VoIP where the person you are calling cannot decipher what you’re saying, or you cannot hear them. With Video, it’s more annoying; you’ve paid to watch a video stream or videoconference and the video and audio are out of sync and stuttering making the whole experience unwatchable.
Voice over IP and Video over IP are all heavily dependant upon the quality of the end-to-end service delivery infrastructure in real-time. It’s all about timely capacity (whether processing or bandwidth) from the service providers datacenter through to your terminal.
Monitoring the customer experience in real-time is possible with response time and synthetic transaction monitoring however isolating the causality of how an issue occurs back to a specific episode has been impossible due to the complexity and distribution of the service delivery infrastructure
Forensic analysis is problematic due to the difficulty of formatting the patchwork of service assurance data from a consistent time and service context suitable for an operations Hero to analyse, comprehend and distill.
Prelert allows operations staff to time travel through their disambiguated data in an organized and structured manner. Prelert sequences and highlights significant episodes of telemetry activity enabling operations staff to traverse from a CDR reporting Latency or Jitter directly through to the casual event; perhaps a network capacity load due to a high incidence of concurrent users – and visualize the process of causality whereby as a consequence the video server retransmits due to the network load, increasing the CPU of that server, causing a general slow down for all users from that server.
Normally the finger would be pointed at the server load, however the real culprit is the fact that the network capacity is not sufficient for peak loads of users subscribing concurrently meaning that the provisioning process should be off-loaded to another network.
SOA Portal Application Delivery
Applications that rely on the cloud to deliver user functionality are highly susceptible to infrastructure abnormalities. User communities of solutions such as SAP, Oracle Business Suite and Microsoft Dynamics, may suffer disruptions to their application without warning causing interruptions to their working day.
Prelert allows operations staff to time travel through their disambiguated data in an organized and structured manner. Prelert sequences and highlights significant episodes of telemetry activity enabling operations staff to traverse from a Call Detail Record reporting Latency or Jitter directly through to the casual event; perhaps a network capacity load due to a high incidence of concurrent users – and visualize the process of causality whereby as a consequence the video server retransmits due to the network load, increasing the CPU utilization of that server, causing a general slow down for all users from that server.
Normally the finger would be pointed at the server load, however the real culprit is the fact that the network capacity is not sufficient for peak loads of users subscribing concurrently meaning that the provisioning process should be off-loaded to another network.