Autonomic request flow manager settings

Use this panel to configure the autonomic request flow manager (ARFM). The ARFM manages incoming messages for applications: it determines whether and when to allow the messages to be serviced by middleware servers. For HTTP and Session Initiation Protocol (SIP) traffic, the management actions take place in the on demand routers. For Java Message Service (JMS) and Internet Inter-ORB Protocol (IIOP) messages, the management actions take place in the application servers.

To view this administrative console page, click Operational policies > Autonomic managers > Autonomic request flow manager.

The autonomic request flow manager (ARFM) contains two parts: a controller and a gateway. The ARFM function is implemented for each cell by a controller plus a collection of gateways in the on demand routers (ODRs). The gateways intercept and queue the incoming requests, while the controller provides control signals, or directions, to the gateways and the placement controller. The components work together to prioritize incoming requests.

[8.5.5.3 or later] To manage HTTP traffic, you can use a node-based algorithm. To enable the node-based ARFM, set the cell custom property arfmQueueMode. With the node-based ARFM, queuing and CPU overload protection is performed at the node level. There is no separate controller and gateway involved.

Depending on your administrative role, you are allowed specific privileges when you configure the autonomic request flow manager. This list shows the administrative roles and privileges for configuring the autonomic request flow manager:

Monitor
Can view the information.
Operator
Can view the information on the configuration tab.
Configurator
Can change the configuration.
Administrator
Has all privileges.

Enabling security When security is enabled, some fields are not editable without appropriate security authorization.

Aggregation period

Each ARFM gateway broadcasts aggregated statistics periodically, and this field specifies the period. The default value is 5 seconds.

[8.5.5.3 or later] This property does not apply to node-based ARFM.

Gateway support reports the following statistics:
  • The runtime charting in the WebSphere® administrative console.
  • The operation of the ARFM controllers.
  • The operation of the application placement controller.

When you set the aggregation period, set the value high enough to support the collection of a sufficient number of performance samples. Gateways collect samples for each request. To produce a good statistical measure, a few hundred samples are necessary. For example, requests that are associated with a service class run in 250 milliseconds, and on average 10 requests run concurrently. The concurrency value is calculated automatically, based on the cluster size and the resources in the environment. You can see the concurrency value on the visualization panels under Runtime operations in the administrative console.

As a result, the service class handles about 40 requests per second. Setting the aggregation period value to 15 seconds results in the collection of 600 samples for each aggregation period. The metrics that are provided by a 600 sample survey are useful and reliable.

Setting an aggregation period value too low results in unreliable performance metrics. Performance metrics that are derived from fewer samples are less reliable than a higher number of samples. Because the ARFM controller is activated when new statistics are produced, setting an aggregation period value too long results in less frequent recomputation of the control settings. Therefore, the product becomes less responsive to sudden changes in traffic intensity and patterns.

Control cycle length minimum

Defines how often the ARFM controller is activated. The default value is 59 seconds.

[8.5.5.3 or later] This property does not apply to node-based ARFM.

Controller activation is the process of evaluating input and producing new control settings as a result of the received input. The activation process for an ARFM controller initiates when new statistics come from one of its gateways, and the elapsed time since the previous activation is greater than or equal to the control cycle minimum length, or if the controller was never activated.

Smoothing window

Defines how sensitive the ARFM controller reaction is to the incoming gateway statistics, by allowing a concatenation of gateway statistics. The default value is 12.

[8.5.5.3 or later] This property does not apply to node-based ARFM.

The ARFM controller of any gateway uses a running average of the last few statistic reports from that gateway. The smoothing window controls the number of reports that are combined. A low smoothing window setting makes the controller more sensitive and supports quicker reaction. However, a low setting also creates a sensitive reaction to noise, or anomalies in the data.

The product of the smoothing window and the aggregation period is approximately the same as the actual control cycle length, which sometimes is slightly greater than the configured control cycle minimum length.

Maximum queue length

Bounds the length of each ARFM queue to a maximum number of requests that are possibly held in queue.

The ARFM has a separate queue for each combination of on demand routers, node groups, service classes, and deployment targets. When a request arrives and the queue is full, the request is rejected. A lower parameter in this field increases the possibility that a request is rejected due to short-term traffic bursts, while a higher parameter allows requests to linger longer in the queues. Queued requests use memory. The default setting is 1000, but test this setting to determine which one best matches your environment.

[8.5.5.3 or later] The node-based ARFM has a separate queue for each node and each cluster. This property refers to the total number of queued requests allowed.

Memory overload protection: Maximum percentage of the WebSphere Application Server heap size to use

Specifies the maximum percentage of the heap size to be used for each application server. This property applies to HTTP and Session Initiation Protocol (SIP) messages. The default is 100%.

Maximum CPU usage

Specifies a maximum CPU usage percentage for middleware nodes. The ARFM considers the entire cluster as a whole when calculating the CPU usage. When the CPU usage exceeds this percentage, the cluster is considered to be overloaded. The ARFM considers the entire cluster as a whole when calculating the CPU usage. The default is 90%.

[8.5.5.3 or later] The node-based ARFM considers CPU usage on a node-by-node basis. When the CPU usage exceeds the maximum usage percentage, the node is considered to be overloaded. The default is 90%.

Rejection policy

A rejection policy prevents a CPU from being overloaded by rejecting incoming HTTP or SIP messages that are not part of preexisting dialogs or sessions.

  • Reject no incoming messages, allow all messages into the queue regardless of their potential to time out or breach the response time threshold of their service policy goal. This option has a rejection threshold value of -1. This value is the default.
  • Reject incoming messages that are not part of a preexisting dialog or session if the predicted response time of the message exceeds the response time threshold of its service policy goal. This option has a rejection threshold value of 0%.
  • Reject incoming messages that are not part of a preexisting dialog or session if the predicted response time of the message exceeds the response time threshold of its service policy goal by more than the following percentage. For this option, specify a Rejection threshold value as a percentage. For example, the default value of 400% means that a message with a service policy response time threshold of 500 ms is rejected if its predicted response time was greater than 2.5 seconds.


File name: arfm.html