Tackling the management of Modern & Hybrid IT Architectures – What is IT Operational Analytics
Speaking with a BSM product executive the other day, he shared his point of view concerning the evolution of IT Operations.
His point of view was that given the proliferation of device types and technologies that are pieced together in ever more fast growing applications, IT needs a BigData approach to add to the deterministic methods of monitoring.
I resume his guiding light into the diagram below.
In this context event triage is the minimum that most IT operations perform over time, dealing with “issues” reactively. Most Infrastructure, network and system management packages provide these capabilities.
Top left includes correlation techniques to prevent the “sea of red”, the plethora of alarms caused by event storms. By exploiting dynamically updated topological modelswhich describe dependencies between managed objects adaptive rules allow correlation to isolate root causes from various symptoms. Add to this the means to exploit patterns, calculate baselines that may have seasonality and use of clever algorithms that seek to define anomalies and you generate Predictive Analytics. www.hp.com/go/sha
Big Data techniques applied to IT operations management can start with gathering log, error, trace files, normalizing their content and then mining the information to detect patterns, seeking out abnormal configurations and comparing behaviors across similar IT resources. Given that the data is examined after the effect, this is still reactive, although if structured event correlated data is combined with the log files, then it is possible to detect IT resources that exhibit regular or irregular issues, study deeper their configurations and dependencies to define corrections that prevent reoccurrences.
Finally, top left. Proactive advanced analytics is typically, but not only, using high powered analytics and search engines to predict issues based on both structured and unstructured data. An interesting domain that pertains directly, I think, to Big Data. Providing the means to initiate “What if” analyses based on previous results, will assist in the definition of search patterns to produce pointers to obtuse application/business service behaviors, potentially before they actually impact any business users and revenues.
Exciting conversation, his point of view clearly defines a journey for most IT shops, gaining value from the masses of data stored across IT, predicting anomalies before business impacts, and driving configurations to consistency will all assist to reducing outages, decrease Opex and improve operational efficiencies. You can read about this exciting stuff and learn how HP is addressing at www.hp.com/go/opsanalytics