This is the keynote presentation by Bhanu Singh delivered at Cloud Expo Santa Clara. He talks about the importance of context in AIOps and how rich data makes a difference in AIOps effectiveness.
Learn more at https://www.opsramp.com
Also, follow us on social media channels to learn about product highlights, news, announcements, events, conferences and more:
Twitter - https://www.twitter.com/OpsRamp
LinkedIn - https://www.linkedin.com/company/opsramp
Facebook - https://www.facebook.com/OpsRampHQ/
8. 8
How Does AIOps Drive Actionable Insights?
Recognizing Redundant Alerts
• Topology based: Cascade of alerts from downstream dependencies
• Clustering based: Cluster of alerts from a common root cause
• Co-occurrence based: Alerts that occur together due to common root cause
Preventative/Proactive Monitoring
• Forecast-based alerting: metrics trending towards a specific limit
• Anomaly detection: out of normal system and application behavior
Learning-Based Resolution
• Route the incident to the right available team based on the historical
analysis, auto-creation, escalation, and notification.
9. Five Use Cases for
Contextual Data
and Actionable
Insights
9
10. Cascading Alerts Across Connected
Resources
Five Alerts, But
Only One Matters
Switch
A
Switch B
VM 1, 2, 3
10
11. Chain of Alerts With the
Same Root Cause
Three Alerts, But
Only One Matters
Server A
11
12. Chain of Alerts With Microservice
Dependency
Microservice-C
Microservice-BMicroservice-A
Microservice-D
12
13. Chain of Alerts Related to Hidden Cause
13
Three Alerts Tied To
One Common Issue
Server A
Server B
Server C
Config
Changes
17. Build a culture of data-driven
actions & outcomes
• Build skills and expertise
• Demonstrate opportunity through value
• Do not start with data
Focus on Business Objectives First1
17
18. 18
Shift From Process-Based to Data-BasedValues2
Staff up with key data
leadership and data
governance rules
18
19. Discover, understand, and rationalize all data sources
Create a data model that drives relationship
and context between technical and business services.
Leverage a metadata-driven data model to deliver context.
Create the right data via:
• Data Selection
• Data Cleaning
• Data Reduction
• DataTransformation
19
Get the Data Right!3
19
20. 20
Bring Qualitative and Quantitative DataTogether4
Selectively apply data-driven
business process, technology
innovation, and corporate
governance
20
21. Use automation to eliminate
tasks that are repetitive,
higher value, and have limited
human interaction
5 Start Small WithTargeted Use Cases
21
22. Why Embrace Contextual, Data-Driven IT Operations?
Create transparency,
team collaboration and
productivity
Drive change with
speed, quality and
confidence
Manage risk and
optimize cost
22
23. Bhanu Singh’s keynote at CloudExpo
Santa Clara
OpsRampTechTalk – Context is
Critical: How Richer DataYields Richer
Results
OpsRamp’s Service-Centric AIOps
solution
The State of AIOps Report
Useful Links
23
24. ThankYou
Bhanu P. Singh
Head of Engineering and Ops
Bhanu.singh@OpsRamp.com
linkedin.com/in/bhanu-singh/
@bhanu1527
Editor's Notes
Business Enabling technologies are changing Faster than Ever
Complexity is demanding automation as it is humanly impossible to monitor and manage distributed hybrid cloud apps and environment. The move towards an agile ops encompassing multi-cloud platform, containers and distributed apps architecture, combined with DevOps and continuous delivery practices challenges the effectiveness of traditional monitoring, infra and apps mgmt. tools. Factors challenging traditional tools are – Static to Dynamic Systems, Data to Insights (Future proofing technologies), More stakeholders, specialist vs. generalist
Machines and algorithm needs right data and its context to accurately operate and perform the tasks that meet the needs and outcomes of the business.
Automation is the only way for modern IT to transform its operating model and support the digitization initiatives of the business processes.
Machines are moving from physical tasks to “Thinking” tasks.
Automation is about applying algorithmic intelligence to eliminate mundane tasks and driving thinking into machine.
A lot of hype and companies are trying to solve all this use cases magically but the key is data context and for each uses data preparation and quality could be different to drive human equivalent or better results following predictable and repeatable model that’s gets better with maturity and over time.
Context could be who, where and when – Location, usage, timing, etc.
Inconsistent data modeling results in a number of problems
– Contextual blindness ( Give example of Alarm->Network Service)
-- Gap Stitching – What can’t be model can’t be effectively monitored, diagnosed, interpreted or predicted. For example compute data utilization (CPU) exposed via infra monitoring should be presented in the context of app supported and the performance and vice versa.
-- Problem often manifest at the deepest point
-- Raw data may have many missing links and information
-- Inaccurate and Fragmented data leads to expensive analysis and non-sensical results.
Chief Information and Security Officer (CISO)
Chief Data Officer (CDO)