One of the benefits of doing the EMA Radar Report: AIOps- A Guide for Investing in Innovation was getting data from all 17 vendors on critical areas ranging from deployment and adoption challenges, to cost and pricing, to architectural and functionality insights across everything from heuristics, to automation, and data assimilation.
Listen to EMA's Dennis Drogseth on the AI+ITOPS Podcast
Administration and Deployment
In the area of deployment and administration, EMA found that on average AIOps vendors indicated between 1-1.5 full-time employees (FTE) were required for ongoing administration in an enterprise with about 10,000 employees. This didn’t include initial deployment or any significant extension in breadth of coverage or functionality.
In 31 interviews, these estimates were generally borne out. Three vendors at the high end estimated between 2.5 and 3 FTEs, whereas the three vendors at the low end estimated between less than 0.5 FTEs.
Heuristics
The great majority of AIOps platforms have heuristics that can "learn" their environments dynamically, without added administrative intervention. On average, they can do this in a little more than one week for 5,000 managed entities.
EMA then asked vendors to weight their AI/ML heuristics on a scale from 0-2, with 2 being a featured heuristic value, 1 being present, and 0 being absent. The top 10 heuristics getting a 2 weighting were:
1. Correlators
2. Anomaly detection
3. Machine learning and baselining for event pattern recognition
4. Topology-based analytics
5. Prescriptive analytics
6. Predictive algorithms
7. Comparators
8. Streaming analytics
9. Optimization algorithms
10. Object-based modeling
Data Assimilation
On average, AIOps vendors could assimilate between 1 million and 10 million metrics within five minutes. When we asked about what data types were in play, we saw:
1. Events (performance related)
2. Time Series
3. Log files
4. Events/ Time Series (security related)
5. Transaction (application performance)
6. Configuration/topology
7. Unstructured data
8. Agent data (systems)
9. Byte code instrumentation
10. Comma delimited files /CSV files
Third-party toolset integration
Significantly, all 17 vendors have some level of third-party toolset integration out of the box, or in parallel, none claim to do "all their own monitoring." In fact, the average AIOps platform has supported integrations for more than 50 different third-party toolsets, with four vendors indicating 100 or more.
These integrations can have powerful political and practical advantages, easing stakeholder reluctance by eliminating the need to break away from their existing tools completely. Additional values include toolset consolidation as IT organizations begin to observe redundancies while also realizing which toolsets are most valuable.
The most common toolset integrations were application performance monitoring (APM) tools tied with CMDBs or extended configuration management systems. Service desk integration for trouble ticketing followed and third-party event management systems came in fourth. Automation integrations were also key, with IT process automation (runbook), and workflow across IT in the lead.
A few use-case views
We had three use-case scenarios. And for each use case we examined a number of factors ranging from domain reach, stakeholders supported, real-time data currency, and heuristics to enable not only awareness of anomalies, but predictive and prescriptive recommendations. Vendors were positioned separately on a per-use-case basis.
When we asked about the top benefits for incident, availability and performance management all vendors led with the following six items, which were also born out in deployment interviews:
■ Faster time to repair problems
■ Proactive ability to prevent problems
■ Improved OpEx efficiencies within IT
■ Less time spent writing rules
■ Real-time insights and historical trends on IT services
■ Reduction/consolidation, minimalization of tools
When we asked what changes each vendor could trace for change impact and capacity optimization, we got the following top five:
■ System configuration service impact analysis
■ Application release changes
■ Service impact analysis (in general)
■ Virtualized infrastructure service impact analysis
■ Containers and microservices service impact analysis
For business impact and IT-to-business alignment, we asked about relevant data sources and saw these as the top five:
■ Enterprise operations data
■ IT warehouse for advanced trending
■ Business application owner data
■ Executive dashboard
■ Security/audit compliance systems
To wrap up
This is just a taste of the data that emerged from our AIOps Radar research. The report contains considerably more detail, while still being a condensation of 105 data-rich slides.
Doing this has been an adventure for me, for EMA as a whole, and I believe for the vendors involved, as well. I do hope you can check out the report and see for yourself as to why.