Overview of Infrastructure Visibility
You can determine the root cause of application issues by looking at application, network, server, and machine metrics that measure infrastructure utilization. For example, the following infrastructure issues may slow down your application:
• Too much time spent in garbage collection of temporary objects (application metric)
• Packet loss between two nodes that results in retransmissions and slow calls (network metric)
• Inefficient processes that result in high CPU utilization (server metric)
• Excessively high rates of reads/writes on a specific disk or partition (hardware metric)
Infrastructure Visibility enables you to isolate, identify, and troubleshoot these types of issues. Infrastructure Visibility is based on a Machine Agent that runs with an App Server Agent on the same machine. These two agents provide multi-layer monitoring, as follows:
1. The App Server Agent collects metrics about applications and identifies applications, tiers, and nodes with slow transactions, stalled transactions, and other application-performance issues.
2. The Network Agent monitors the network packets sent and received on each node and identifies lost/retransmitted packets, TCP bottlenecks, high round-trip times, and other network issues.
3. The Machine Agent collects metrics at two levels:
• Server Visibility metrics for local processes, services, and resource utilization.
• Basic machine metrics for disks, memory, CPU, and network interfaces.
This multilayer monitoring enables you to determine possible correlations between application issues and service, process, hardware, network, or other issues on the machine. Figure 4-26 illustrates the Agent Monitoring Metrics.
Figure 4-26 Agent Monitoring Metrics
Network Visibility
Network Visibility monitors traffic flows, network packets, TCP connections, and TCP ports. Network Agents leverage the APM intelligence of App Server Agents to identify the TCP connections used by each application. Network Visibility includes the following items:
• Detailed metrics about dropped/retransmitted packets, TCP window sizes (Limited/Zero), connection setup/tear down issues, high round-trip times, and other performance-impacting issues
• Network Dashboard that highlights network KPIs for tiers, nodes, and network links
• Right-click dashboards for tiers, nodes, and network links that enable quick drill-downs from transaction outliers to network root causes
• Automatic mapping of TCP connections with application flows
• Automatic detection of intermediate load balancers that split TCP connections
• Diagnostic mode for collecting advanced diagnostic information for individual connections
Network Visibility extends the application intelligence of AppDynamics APM down the stack from the application to the network. With “app-only” visibility, it can be easy to mistakenly blame (or not blame) the network when an application issue arises. Network Visibility can help reduce or eliminate the guesswork involved in identifying root causes. Network Agents and App Agents, working together, automate the work of mapping TCP connections to the application flows that use them. Network Agents can identify intermediate load balancers (which often split TCP connections) and correlate the connections on either side of these devices. Figure 4-27 illustrates the agent-based Network Visibility approach.
Figure 4-27 Agent-based Network Visibility approach
The agent-based approach of Network Visibility provides these advantages over standard approaches to network monitoring:
• More cost-efficient than using network monitoring appliances, which often view traffic from a few central locations
• Especially useful for distributed environments and multitier applications that span multiple network segments
• Works in cloud and hybrid networks, unlike most network-monitoring solutions
