19 Configuring and Managing Communication Service Traffic

Understanding the Plug-in Manager Execution and Evaluation Flow

The following section describes the execution and evaluation flow.

Understanding Plug-in Routing Logic

The plug-in routing logic matches data in a request, and data associated with a request, with routing logic and results in the selection of a plug-in instance. The request is handed off to the selected plug-in instance for further processing.

The routing logic offers a fine grained way to select a network protocol plug-in instance and makes it possible to select a plug-in instance, and thereby a network node, to be targeted for individual requests, based on all data available in the request.

In combination with the plug-in instance feature make it possible to single out a certain network node based on the above operands and thereby:

• Offer different QoS levels. Example: different network nodes offers different QoS, for example latencies for message delivery.

• Ensure that requests are routed to a node that can actually handle the request. Example: one SMSC is capable of handling binary content in the form of SM logos, while another is not.

• Maximize utilization of the network nodes. Example: Two network nodes offers exactly the same functionality, but one processes the request more expensively, so the selection is done based on the charging code the application supplied.

In combination with the plug-in instance feature make it possible to single out a network protocol plug-in instance even if the different plug-in instances connect to the same network node and thereby use different configurations for the request toward the network node.

For example, a network protocol plug-in offers the possibility to configure a priority parameter when sending request to the network node, by setting the parameter differently in different plug-in instances, different priorities can be used for different service providers. In the same way credentials can be mapped so the originator of the request is mapped to the request that is made to the network node.

<?xml version="1.0" encoding="UTF-8"?>
<route xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:noNamespaceSchemaLocation="route.xsd">
  <destAddress>tel:123.*</destAddress>
</route>

<?xml version="1.0" encoding="UTF-8"?>
<route xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:noNamespaceSchemaLocation="route.xsd">
  <destAddress defaultResult="true">.*</destAddress>
</route>

<?xml version="1.0" encoding="UTF-8"?>
<route xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:noNamespaceSchemaLocation="route.xsd">
  <and>
    <spId>sp1</appId>
    <appId>app1</appId>
  </and>
</route>

<?xml version="1.0" encoding="UTF-8"?>
<route xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:noNamespaceSchemaLocation="route.xsd">
  <and>
    <not><method>sendSmsRingtone</method></not>
    <or>
      <appInstanceId>appInst1</appInstanceId>
      <and>
        <method>sendSms</method>
        <param name="arg0.senderName" value="tel:123456"/>
      </and>
    </or>
  </and>
</route>

<?xml version="1.0" encoding="utf-8"?>
<xs:schema xmlns:xs="http://www.w3.org/2001/XMLSchema">
  <xs:element name="route">
    <xs:complexType>
      <xs:group ref="ConditionGroup" minOccurs="1" maxOccurs="1"/>
    </xs:complexType>
  </xs:element>

  <xs:group name="ConditionGroup">
    <xs:choice>
      <xs:element name="and" type="AndType"/>
      <xs:element name="or" type="OrType"/>
      <xs:element name="not" type="NotType"/>
      <xs:element name="spId" type="xs:string"/>
      <xs:element name="appId" type="xs:string"/>
      <xs:element name="appInstanceId" type="xs:string"/>
      <xs:element name="destAddress" type="AddressType"/>
      <xs:element name="method" type="xs:string"/>
      <xs:element name="param" type="ParamType"/>
      <xs:element name="xparam" type="ParamType"/>
    </xs:choice>
  </xs:group>

  <xs:complexType name="AndType">
    <xs:group ref="ConditionGroup" minOccurs="1" maxOccurs="unbounded"/>
  </xs:complexType>

  <xs:complexType name="OrType">
    <xs:group ref="ConditionGroup" minOccurs="1" maxOccurs="unbounded"/>
  </xs:complexType>

  <xs:complexType name="NotType">
    <xs:group ref="ConditionGroup" minOccurs="1" maxOccurs="1"/>
  </xs:complexType>

  <xs:complexType name="AddressType">
    <xs:simpleContent>
      <xs:extension base="xs:string">
        <xs:attribute name="defaultResult" type="xs:boolean" default="false"/>
      </xs:extension>
    </xs:simpleContent>
  </xs:complexType>

  <xs:complexType name="ParamType">
    <xs:attribute name="name" type="xs:string" use="required"/>
    <xs:attribute name="value" type="xs:string" use="required"/>
    <xs:attribute name="defaultResult" type="xs:boolean" default="false"/>
  </xs:complexType>
</xs:schema>

Specifying Address Ranges in Routes

Address ranges are specified using UNIX regular expressions. A few examples are given below:

• ^.* specifies a route that matches all addresses.

• ^[0-5].* specifies a route that matches all address strings starting with 0, 1, 2, 3, 4, or 5.

• ^.*[6-9]$ specifies a route that matches all address strings ending with 6, 7, 8, or 9.

• ^46.* specifies a route that matches all address string starting with 46.

• ^46.{8}$ specifies a route that matches all address strings starting with 46 that contain exactly 10 total digits.

• ^.*@.*\.com$ specifies a route matching all mail addresses in the com domain. The dot in .com must be written "\.".

In the examples:

• ^ indicates the beginning of the string.

• . matches anything except a new-line. That is, "a.b" matches any three-character string that begins with a and ends with b.

• * is a suffix that means the preceding regular expression is to be repeated zero or more times. That is, the "." in the expression ”^46.*" is repeated until the whole string is matched.

• $ is an indicator of end of line (or end of string).

The address scheme can be included in the expression. If not specified, any scheme will match. Examples:

• tel:^46.* matches all tel (E.164) numbers starting with 46.

• sip:.* matches any SIP address.

Understanding Plug-in Routes and Routing Logic

Plug-in routing logic is an extension of the plug-in routes and these PluginManagerMBean methods:

• setRouteConfig

• getRouteConfig

Plug-in routes added using the addRoute method, are converted into plug-in routing logic where a route is added as a logical OR given that the route that is being modified contains <or> and <destAddress defaultResult="true"> elements only.

See "PluginManagerMBean Reference" for information on using this MBean.

Creating a Plug-in instance

Following is an outline for creating an instance of a plug-in service using the PluginManagerMBean:

1. Find the plug-in service ID for the service you want to use to create the plug-in instance. The plug-in service IDs are listed under the heading ”Properties” in the sections describing the management of each plug-in. To get a list of plug-in service IDs, use the listPluginServices method.

2. Use the createPluginInstance method to create the instance. The Plug-in Instance ID supplied as a parameter will be used to identify the instance for all routing and administration.

3. To destroy an instance, use the destroyPluginInstance method.

See "PluginManagerMBean Reference" for information on using this MBean.

Administering Plug-in Routing Logic and Node IDs

You administer routes using these PluginManagerMBean operations:

• To add new routing logic: getRouteConfig

• To change or remove routing logic: getRouteConfig and setRouteConfig

• To list all existing routes: listRoutes

• To get information about a specific plug-in instance: getPluginServiceInfo

• To get a list of all registered plug-in instances: listPluginInstances

• To get a list of all registered plug-in services: listPluginServices

• To define the node ID: setPluginNodeId.

  A node ID is assigned to one or more Plug-in Instance IDs to enforce node SLAs. The node ID is then referred to in the node SLAs.

See "PluginManagerMBean Reference" for information on this MBean.

Adding Bulk Messaging Support to a Communication Service

If a communication service does not already support bulk messaging, you can add support for it. If you want bulk messaging applied to the message address, have the PluginFactory implement the BulkRequestFactory, and implement the createSingleAddressRequest method.

To make bulk messaging apply to other types of information in the request:

1. Create a new class that extends AddressRequestInfo to contain the information after the message is split.

2. Create a new xparameter or other identifier in the original bulk request.

3. Define a new schema for the information in the bulk request, and add a process for it in the createRequestInfo method in the PluginFactory.

4. Set the information back in the createSingleAddressRequest method in the PluginFactory.

OverloadProtectionMBean

The OverloadProtectionMBean consists of two MBeans, one for configuration and one for producing statistics.

OverloadProtectionMBean configures the following options:

• Enable overload protection attribute, EnableOverloadProtection

  A true-or-false flag to enable or disable the feature (Default: true)

• Enable load statistic collection attribute, EnableLoadStatisticCollection

  A true-or-false flag that enables or disables the collection of system load statistics. If it is false, the system stops collecting and updating load data (Default: true).

• Performance metrics to monitor, which include but are not limited to transaction number, Servlet session number, memory usage, CPU usage, and so on. Each metric has the following configuration items:

  • Name - name of the metric

  • Collector class - the class name of the PerformanceStatisticCollector for this metric

  • Critical metrics - true or false. If true, the metric can be the decisive factor in determining the overload status. If false, the metric is aggregated with other non-critical metrics to calculate the system load.

  • Valid on AT - true or false. Indicates whether the collector is valid on an AT server or not

  • Valid on NT - true or false. Indicates whether the collector is valid on an NT server or not. A collector must be valid on either an AT server, an NT server, or both.

  • Weight - a percentage between 0 and 100. If the metric is not critical, its load contributes to the system load according to the weight setting. The sum of all non-critical metrics should be 100 percent.

• Overload threshold

  The overload is defined in three levels: light, medium, and high. Normally, for each level, the exiting threshold should be a little bit less than the entering threshold to avoid a ping-pong effect. The following examples illustrate the three levels:

  • light - entering threshold of 85% and exiting threshold of 80%

  • medium - entering threshold of 90% and exiting threshold of 85%

  • high - entering threshold of 95% and exiting threshold of 90%

• Number of consecutive overloads leading to protection

  This option prevents the system from taking action on overload bursts, such as a one-second burst. Overload protection takes effect when the system is experiencing a heavy load and cannot recover soon. If the system can auto-recover, no action needs to be taken. The default setting is 8, which means that overload protection occurs when the system reaches the overload threshold for two consecutive collection windows.

• Time period for statistics collection

  Defines the timer for StatisticsCollector. Default is 4 seconds.

• Traffic throttling ratio

  Defines the traffic-rejecting ratio for each overload level. For example:

  • light: 20%

  • medium: 50%

  • high: 80%

When a change is made to the configuration, an event is issued and OverloadProtectionController and StatisticsCollector update accordingly.

Performance statistics exposed by OverloadStatisticsMBean are based on data received from OverloadProtectionController and include:

• Current system load for each metric

• Current system overload status

• Total time spent in overload for each level

• Total number of rejected requests during overload

Statistics data is not stored in a database. It is collected and updated from system startup.

StatisticsCollector

The StatisticsCollector should be a server service on each cluster node (in multi-tier, on each AT and NT machine) that is responsible for periodically collecting relevant performance data. It will use a set of collectors to do the job, as defined in the MBean, one collector for one metric. The collectors should follow the same interface so that it is easy to add new collectors. When the timer expires, the StatisticsCollector calls the getStats() method of each collector, then composes a StatisticsReport to send to the OverloadProtectionController.

Overload protection includes three out-of-the-box collectors:

• CpuLoadCollector collects CPU usage, oracle.ocsg.overload_protection.collector.CpuLoadCollector

• MemoryUsageCollector collects memory heap usage, oracle.ocsg.overload_protection.collector.CpuLoadCollector

• DafRequestCountCollector collects the DAF request rate and compares it with a predefined threshold to determine the load percentage, oracle.ocsg.overload_protection.collector.DafRequestCountCollector

With the MBean provided by WebLogic Server, the collector could collect some system metrics, like memory and CPU usage for all servers in the domain, so that it is not necessary to allocate this work to each collector. The OverloadProtectionController assigns a delegate across all collectors to collect data from WLS MBean during startup, by default the one stays on same node of the OverloadProtectionController. OverloadProtectionController should be able to change the delegate, if the current delegate is down.

If some Services Gatekeeper internal resources are important and can reflect the system load, you should create collectors to monitor the usage, one for each resource. The modules that hold such resources must provide an interface to enable the collectors to get the statistics.

OverloadProtectionController

The OverloadProtectionController is the central controller of overload protection, a singleton service that stays on one of the NT nodes in the domain. Each StatisticsCollector and TrafficThrottler should register itself with the OverloadProtectionController during startup, so that the controller knows how many instances are active. The major work of OverloadProtectionController includes:

• Receiving a periodic StatisticsReport from StatisticsCollector

• Updating the statistics data on the MBean

  The OverloadProtectionController aggregates and summarizes the StatisticsReport from each StatisticsCollector and periodically updates it to the MBean.

• Determining the system's overload status based on the StatisticsReport. The system load is calculated as follows:

  • For the non-critical metrics, the aggregated load is sum(weight * load)

  • The system load is the max value of the aggregated non-critical load and each critical metric's load

  If the system load changes from normal to overload for predefined consecutive windows, regardless of which overload level, then the system is deemed overloaded.

• Notifying the TrafficThrottler when the overload status changes

• Receiving the periodic ThrottlingReport from TrafficThrottler during overload and updating the MBean

  The OverloadProtectionController should aggregate and summarize the ThrottlingReport from each TrafficThrottler and then update to MBean periodically.

The communication between OverloadProtectionController and StatisticsCollector or TrafficThrottler is through RMI messages and an event channel. With WLS singleton service, the OverloadProtectionController can seamlessly switch between NT nodes without breaking service, if the current node is down.

TrafficThrottler

The TrafficThrottler stays on each AT node. It is the first security provider so that it can take effect at the very beginning. If the system is not overloaded, it does nothing. When the OverloadProtectionController sends OverloadStatusInd() to say that the system is in overload, the TrafficThrottler begins rejecting incoming requests based on the predefined ratio in the MBean, with code 503 Service Unavailable, and collects the throttling data and sends a periodic ThrottlingReport back to the OverloadProtectionController. When the overload level is up or down, it applies a different throttling ratio. When the system quits overload, TrafficThrottler stops throttling.