1. Feature Guide
  2. DSR Features and Functions
  3. Routing and Transaction Related Parameters in the DSR
  4. Congestion Control
  5. Per Connection Ingress MPS Control

2.6.15.1 Per Connection Ingress MPS Control

The Per-Connection Ingress MPS Control feature provides the following:

  • A method to reserve/guarantee a user-configured minimum ingress message capacity for each peer connection.
  • A method for limiting the ingress message capacity for a peer connection to a user-configured maximum.
  • A method for multiple peer connections to have a ‘shared’ ingress message capacity.
  • A method to prevent the total reserved ingress message capacity of all active peer connections on a DA MP from exceeding the DA MP’s capacity.
  • A method for limiting the overall rate at which a DA MP attempts to process messages from all peer connections.
  • A method for coloring (Green or Yellow) messages ingressing a DSR.

There are six user-configurable capacity configuration set parameters for DSR Connections: Ingress MPS Minor Alarm Threshold, Ingress MPS Major Alarm Threshold, Abatement Time, Reserved Ingress MPS, Maximum Ingress MPS and Convergence Time. Additional details on some of these follow.

  • Reserved Ingress MPS:
    • Ingress capacity (in Messages per Second) reserved for use by the peer connection. It is not available for use by other connections on the same DA MP.
    • Min value: 0
    • Max value: Minimum (Connection engineered capacity, DA MP’s licensed MPS capacity)
    • Default: 0

When a DSR Connection’s ingress message rate is equal to or below its configured Reserved Ingress MPS, all messages ingressing the connection are colored Green. When a DSR Connection’s ingress message rate is above its configured Reserved Ingress MPS, messages below the reserved capacity are colored green and messages above the reserved capacity are colored yellow.

  • Maximum Ingress MPS:
    • Maximum ingress capacity (in Messages per Second) allowed on this connection. Capacity beyond “reserved” and up to “max” is shared by all connections on the DA MP and comes from DA MP capacity leftover after all connections’ “reserved” capacities have been deducted from the DA MP capacity.
    • Min value: 10.
    • Max value: Minimum (Connection engineered capacity, DA MP’s licensed MPS capacity).
    • Default: Minimum (Connection engineered capacity, DA MP’s licensed MPS capacity).

A fundamental principal of Per-Connection Ingress MPS Control is to allocate a DA-MP’s ingress message processing capacity among the Diameter peer connections that it hosts. Each peer connection is allocated, via user-configuration, a reserved and a maximum ingress message processing capacity. The reserved capacity for a connection is available for exclusive use by the connection. The capacity between a connection’s reserved and maximum is shared with other connections hosted by the DA-MP. The DA-MP reads messages arriving from a peer connection and attempts to process them as long as reserved or shared ingress message capacity is available for the connection.

  • Convergence Time:
    • Convergence time in ms used by an algorithm for Ingress MPS rate computation by PCIMC.
    • Min value: 250ms
    • Max value: 4000ms
    • Default: 1000ms

Rate Convergence Time is used in message rate computation done using an algorithm called Sliding-Historic Metric where traffic history (that is message count) is stored for a configured time. The DSR maintains a sliding history using an array of traffic counts (ex: one second history), where each element in the array represents time in ms (for example: 50 ms) of elapsed time. Message rate is calculated as message per second (MPS). Rate Convergence Time allows the user to control the sensitivity of the request traffic bursts and allows them to tune accordingly.

When the ingress messages are above the connection Maximum ingress MPS rate the DA-MP enforces a short discard period, during which time ingress messages over the defined maximum Ingress MPS are read from the connection and discarded based on the message priority. This approach provides some user-configurable bounding of the DSR application memory and compute resources that are allocated for each peer connection, reducing the likelihood that a subset of DSR downstream peers which are offering an excessive/unexpected Request load can cause DSR congestion or congestion of DSR upstream peers. The discarding of ingress messages by the DSR results in the DSR Peer experiencing Request timeouts (when DSR discards Request messages) and/or receiving duplicate Requests (when DSR discards Answer messages).

It should be noted that the DSR is enforcing ingress message rate independent of the type (that is Request or Answer) or size of the ingress messages.

The figure below depicts a DSR DA MP hosting 3 connections with the attributes shown in the following table:

Table 2-6 DSR Ingress MPS Configuration Example 1

Connection Reserved Ingress MPS Maximum Ingress MPS MPS shared with other connections
Connection1 100 500 400
Connection 2 0 5000 5000
Connection 3 500 500 0

Figure 2-18 DSR Ingress MPS Configuration Example 1 – Normal Case


DSR Ingress MPS Configuration Example 1 – Normal Case

The DSR prevents the total Reserved Ingress MPS of all connections hosted by a DA MP from exceeding the DA MP’s maximum ingress MPS. The enforced limit for this is the DA MP’s licensed MPS capacity, which defaults to the DA MP’s maximum engineered capacity. The enforcement of this requirement on ‘configured’ connections versus ‘Enabled’ or ‘Active’ connections is a design decision.

This feature addresses the functionality to assist DSR overload and throttling algorithms in differentiating messages ingressing a DSR connection whose ingress message rate is above (vs equal to or below) its configured reserved ingress MPS.

When a DSR connection’s ingress message rate is equal to or below its configured reserved ingress MPS, all messages ingressing the connection are colored green. When a DSR connection’s ingress message rate is above its configured reserved ingress MPS, messages below the reserved capacity are colored green and messages above the reserved capacity are colored yellow. Message color is used as a means for differentiating diameter connections that are under-utilized versus those that are over-utilized with respect to ingress traffic. Traffic from under-utilized connections are marked “green” by the per-connection ingress MPS control (PCIMC) feature. As stated above, when a DSR connection’s ingress message rate is above its configured reserved ingress MPS, messages below the reserved capacity are colored green and messages above the reserved capacity are colored yellow. In the event of danger of congestion or of CPU congestion and based on the specified discard policy, traffic from over-utilized connections is considered for discard before traffic from under-utilized connections. Traffic discarded by PCIMC due to capacity exhaustion (per-connection or shared) is marked “red” and is not considered for any subsequent processing.

Figure 2-19 Message Coloring and Priority or Color based DA-MP Overload Control


Message Coloring and Priority or Color based DA-MP Overload Control