Network Director is integrated with Cloud Analytics Engine – Helps customers in troubleshooting Network Congestion
Oct 28, 2015
Network Congestion is generally caused by oversubscribed physical interfaces of a network element (switch). This oversubscription can cause queuing, latency sometimes packet loss and can affect network throughput and degrade the overall QoS of the network. Troubleshooting network congestion efficiently improves overall network performance and helps in understanding the causes to prevent problems in the future.
How to troubleshoot: Typically network administrators use Network Traffic Analysis (NTA) to troubleshoot and resolve network congestion issues. Network Traffic Analysis is the process of collecting, processing, analyzing and deriving traffic patterns. Correlating various network traffic patterns of various statistics helps in effectively troubleshooting and resolving and even preventing (predictive analysis and capacity planning) network congestion.
Juniper’s solution: Juniper’s network management product offerings enable the network administrators to troubleshoot network congestion by integrating Network Director (ND) and Data Learning Engine (DLE) a component of Cloud Analytics Engine (CAE). NTA feature in Network Director is designed to help administrators to troubleshoot oversubscribed physical interfaces by trending Applications and Conversations on the physical interface in question. The NTA feature includes the following use cases:
Top N Applications (identified by protocol and destination port) on a physical interface in a selected time interval.
Top N Conversations (unique pair of source and destination end points) on a physical interface in a selected time interval.
How does it work?
Juniper’s NTA uses Data Learning Engine’s (DLE) collection and processing engine to gather application and conversation information. DLE receives flow samples via s-flow (see http://sflow.org) protocol periodically from the network elements (switches). DLE then processes these flow samples to derive application and conversation information. Applications and Conversations samples are exposed via REST API interface. Network Director uses the REST API interface to fetch information trends.
Identifying Congestion: The Device & Port Utilization’ heat map on the main dashboard of Network Director presents the bandwidth utilization of all the physical interfaces. Bandwidth utilization is computed based on the packets transmitted and link speed of the physical interface in a given period of time. Color-coding of physical interfaces as per the percentage utilization of bandwidth allows network administrators to easily navigate to the interfaces that are experiencing congestion.
Enabling s-flow: Network Director automatically enables s-flow for the physical interfaces when the administrators either access NTA feature or the interface experiencing high bandwidth utilization (above a configured threshold). This automatic configuration of s-flow allows monitoring of network without user intervention. Network Director utilizes the capabilities of thresholds feature to trigger automatic configuration changes to the switches. This configuration includes details of DLE (host name and port) to which the s-flow samples need to be exported.
Top Applications: Network Director presents the top applications by number of bytes transferred by each application for a given period through the physical interface in question. The top applications screen presents two sets of information. The table lists the applications sorted by number of bytes transmitted (most bytes transmitted at the top). The top portion is a line chart for each application with average transmitted byes at regular points in time for the selected period.
Top Conversations: The top conversations screen presents the number of bytes transmitted between any two end points in the network. The bytes transmitted between end points are for all the applications. The information presented here is very similar to the information presented in the top applications screen.
Now that the top applications and conversations are identified on congested interfaces, the following corrective actions can be taken to improve the efficiency of the network:
Make necessary configuration changes to the switches to distribute the application traffic (e.g. QoS).
Make necessary changes to end point connections to evenly distribute the network traffic across all interfaces on the switch.
Leverage aggregated interfaces to meet the bandwidth needs between end points.
Identify the patterns in application and end point traffic to plan network capacity for future network needs.