With every minute you can shave off root cause analysis, you get a minute closer to restoring the performance or availability of a process that's important to your business. But the plethora of monitoring tools used throughout your organization, each with its own root cause perspective about the IT environment, can lead to confusion, dysfunction and drawn-out debate when things go wrong. To get the most business value from these diverse views, you need to understand how they can work together.
Think of root cause analysis as a software stack, and the higher the layer is in the stack, the more meaningful it is from a business perspective. For example, in the Open Systems Interconnect (OSI) stack, understanding layer 1, the physical layer, is vital, but layer 7, the application, is more meaningful to the business.
Each layer in the root cause analysis stack is provided by unique monitoring functions, analytics and visualization. Here they are, top down:
- Business Service Root Cause Analysis
- Application-Driven Root Cause Analysis
- Network Fault Root Cause Analysis
- Device Root Cause Analysis
Think of adding each layer in terms of a geometrical analogy of human awareness cleverly explained by the Russian philosopher P.D. Ouspenski in his book Tertium Organum. As he explained, if you were one-dimensional, a point, you couldn't think of a line. If you were a line, you couldn't perceive two-dimensions: a square. If you were a square you couldn't understand a cube. If you were a cube, couldn't understand motion.
Let's see how each layer has legitimate root cause analysis and how each successive layer up the stack adds awareness and greater business value.
1. Device Root Cause Analysis
The device layer is the foundation, letting you know if a server, storage device or switch, router, load balancer, etc. simply is up or down, fast or slow. If it's pingable, you know it has a power source, and diagnostics can tell you which subcomponent has the fault causing the outage. For root cause of performance issues, you'll be relying on your monitoring tools' visual correlation of time series data and threshold alerts to see if the CPU, memory, disk, ports etc. are degraded and why.
But if servers or network devices aren't reachable, how do you know for sure if they are down or if there's an upstream network root cause? To see this, you need to add a higher layer of monitoring and analytics.
2. Network Fault Root Cause Analysis
The next layer is Network Root Cause Analysis. This is partly based on a mechanism called inductive modeling, which discovers relationships between networked devices by discovering port connections and routing and configuration tables in each device.
When an outage occurs, inference, a related Network Root Cause Analysis mechanism, uses known network relationships to determine which devices are downstream from the one that is down. So instead of drowning in a sea of red alerts for all the unreachable devices, you get one upstream network root cause alert. This can also be applied to virtual servers and their underlying physical hosts, as well as network configuration issues.
3. Application-Driven Root Cause Analysis
Next up is Application-Driven Network Performance Management, which includes two monitoring technologies: network flow analysis and end-to-end application delivery analysis.
The first mechanism lets you see which applications are running on your network segments and how much bandwidth each is using. When users are complaining that an application service is slow, this can let you know when a bandwidth-monopolizing application is the root cause. Visualization includes stacked protocol charts, top hosts, top talkers, etc.
The second mechanism in this layer shows you end-to-end application response timing: network round trip, retransmission, data transfer and server response. Together in a stacked graph, this reveals if the network, the server or the application itself is impacting response. To see the detailed root cause in the offending domain, you drill down into a lower layer (e.g., into a network flow analysis, device monitoring or an application forensic tool).
4. Business Service Root Cause Analysis
The best practice is to unify the three layers into a single infrastructure management dashboard, so you can visually correlate all three levels of analytics in an efficient workflow. This is ideal for technical Level 2 Operations specialists and administrators.
But there's one more level at the top of the stack: Business Service Root Cause Analysis. This gives IT Operations Level 1 staff the greatest insight into how infrastructure is impacting business processes.
Examples of business processes include: Concept To Product, Product To Launch, Opportunity To Order, Order To Cash, Request To Service, Design To Build, Manufacturing To Distribution, Build To Order, Build To Stock, Requisition To Payables and so on.
At this layer of the stack, you monitor application and infrastructure components in groups that support each business process. This allows you to monitor each business process as you would an IT infrastructure service, and a mechanism called service impact analysis rates the relative impact each component has on the service performance. From there you can drill down into a lower layer in the stack to see the technical root cause details of the service impact (network outage, not enough bandwidth, server memory degradation, packet loss, not enough host resources for a virtual server, application logic error, etc.).
Once you have a clear understanding of this architecture, and a way to unify the information into a smooth workflow for triage, you can put the human processes in place to realize its business value.
ABOUT David Hayward
David Hayward is Senior Principal Manager, Solutions Marketing at CA Technologies. Hayward specializes in integrated network, systems and application performance management – and his research, writing and speaking engagements focus on IT operations maturity challenges, best-practices and IT management software return on investment. He began his career in 1979 as an editor at the groundbreaking BYTE computer magazine and has since held senior marketing positions in tier one and startup computer system, networking, data warehousing, VoIP and security solution vendors.
The State of the Mainframe report from Syncsort revealed an increased focus on traditional data infrastructure optimization to control costs and help fund strategic organizational projects like AI, machine learning and predictive analytics in addition to widespread concern about meeting security and compliance requirements ...
The 2018 Software Fail Watch report from Tricentis investigated 606 failures that affected over 3.6 billion people and caused $1.7 trillion in lost revenue ...
Gartner predicts there will be nearly 21 billion connected “things” in use worldwide by 2020 – impressive numbers that should catch the attention of every CIO. IT leaders in nearly every vertical market will soon be inundated with the management of both the data from these devices as well as the management of the devices themselves, each of which require the same lifecycle management as any other IT equipment. This can be an overwhelming realization for CIOs who don’t have an adequate configuration management strategy for their current IT environments, the foundation upon which all future digital strategies – Internet-connected or otherwise – will be built ...
Many network operations teams question if they need to TAP their networks; perhaps they aren't familiar with test access points (TAPs), or they think there isn't an application that makes sense for them. Over the past decade, industry best-practice revealed that all network infrastructure should utilize a network TAP as the foundation for complete visibility. The following are the seven most popular applications for TAPs ...
Organizations are eager to adopt cloud based architectures in an effort to support their digital transformation efforts, drive efficiencies and strengthen customer satisfaction, according to a new online cloud usage survey conducted by Denodo ...
Globally, cloud data center traffic will represent 95 percent of total data center traffic by 2021, compared to 88 percent in 2016, according to the Cisco Global Cloud Index (2016-2021) ...
Enterprise cloud spending will grow rapidly over the next year, and yet 35 percent of cloud spend is wasted, according to The RightScale 2018 State of the Cloud Survey ...
What often goes overlooked in our always-on digital culture are the people at the other end of each of these services tasked with their 24/7 management. If something goes wrong, users are quick to complain or switch to a competitor as IT practitioners on the backend race to rectify the situation. A recent PagerDuty State of IT Work-Life Balance Report revealed that IT professionals are struggling with the pressures associated with the management of these digital offerings ...
Businesses everywhere continually strive for greater efficiency. By way of illustration, more than a third of IT professionals cite "moving faster" as their top goal for 2018, and improving the efficiency of operations was one of the top three stated business objectives for organizations considering digital transformation initiatives ...
One of the current challenges for IT teams is the movement of the network to the cloud, and the lack of visibility that comes with that shift. While there has been a lot of hype around the benefits of cloud computing, very little is being said about the inherent drawbacks ...