Net DIM - Generic Network Dynamic Interrupt Moderation

Author:

Tal Gilboa <talgi@mellanox.com>

Assumptions

This document assumes the reader has basic knowledge in network drivers and in general interrupt moderation.

Introduction

Dynamic Interrupt Moderation (DIM) (in networking) refers to changing the interrupt moderation configuration of a channel in order to optimize packet processing. The mechanism includes an algorithm which decides if and how to change moderation parameters for a channel, usually by performing an analysis on runtime data sampled from the system. Net DIM is such a mechanism. In each iteration of the algorithm, it analyses a given sample of the data, compares it to the previous sample and if required, it can decide to change some of the interrupt moderation configuration fields. The data sample is composed of data bandwidth, the number of packets and the number of events. The time between samples is also measured. Net DIM compares the current and the previous data and returns an adjusted interrupt moderation configuration object. In some cases, the algorithm might decide not to change anything. The configuration fields are the minimum duration (microseconds) allowed between events and the maximum number of wanted packets per event. The Net DIM algorithm ascribes importance to increase bandwidth over reducing interrupt rate.

Net DIM Algorithm

Each iteration of the Net DIM algorithm follows these steps:

  1. Calculates new data sample.

  2. Compares it to previous sample.

  3. Makes a decision - suggests interrupt moderation configuration fields.

  4. Applies a schedule work function, which applies suggested configuration.

The first two steps are straightforward, both the new and the previous data are supplied by the driver registered to Net DIM. The previous data is the new data supplied to the previous iteration. The comparison step checks the difference between the new and previous data and decides on the result of the last step. A step would result as “better” if bandwidth increases and as “worse” if bandwidth reduces. If there is no change in bandwidth, the packet rate is compared in a similar fashion - increase == “better” and decrease == “worse”. In case there is no change in the packet rate as well, the interrupt rate is compared. Here the algorithm tries to optimize for lower interrupt rate so an increase in the interrupt rate is considered “worse” and a decrease is considered “better”. Step #2 has an optimization for avoiding false results: it only considers a difference between samples as valid if it is greater than a certain percentage. Also, since Net DIM does not measure anything by itself, it assumes the data provided by the driver is valid.

Step #3 decides on the suggested configuration based on the result from step #2 and the internal state of the algorithm. The states reflect the “direction” of the algorithm: is it going left (reducing moderation), right (increasing moderation) or standing still. Another optimization is that if a decision to stay still is made multiple times, the interval between iterations of the algorithm would increase in order to reduce calculation overhead. Also, after “parking” on one of the most left or most right decisions, the algorithm may decide to verify this decision by taking a step in the other direction. This is done in order to avoid getting stuck in a “deep sleep” scenario. Once a decision is made, an interrupt moderation configuration is selected from the predefined profiles.

The last step is to notify the registered driver that it should apply the suggested configuration. This is done by scheduling a work function, defined by the Net DIM API and provided by the registered driver.

As you can see, Net DIM itself does not actively interact with the system. It would have trouble making the correct decisions if the wrong data is supplied to it and it would be useless if the work function would not apply the suggested configuration. This does, however, allow the registered driver some room for manoeuvre as it may provide partial data or ignore the algorithm suggestion under some conditions.

Registering a Network Device to DIM

Net DIM API exposes the main function net_dim(). This function is the entry point to the Net DIM algorithm and has to be called every time the driver would like to check if it should change interrupt moderation parameters. The driver should provide two data structures: struct dim and struct dim_sample. struct dim describes the state of DIM for a specific object (RX queue, TX queue, other queues, etc.). This includes the current selected profile, previous data samples, the callback function provided by the driver and more. struct dim_sample describes a data sample, which will be compared to the data sample stored in struct dim in order to decide on the algorithm’s next step. The sample should include bytes, packets and interrupts, measured by the driver.

In order to use Net DIM from a networking driver, the driver needs to call the main net_dim() function. The recommended method is to call net_dim() on each interrupt. Since Net DIM has a built-in moderation and it might decide to skip iterations under certain conditions, there is no need to moderate the net_dim() calls as well. As mentioned above, the driver needs to provide an object of type struct dim to the net_dim() function call. It is advised for each entity using Net DIM to hold a struct dim as part of its data structure and use it as the main Net DIM API object. The struct dim_sample should hold the latest bytes, packets and interrupts count. No need to perform any calculations, just include the raw data.

The net_dim() call itself does not return anything. Instead Net DIM relies on the driver to provide a callback function, which is called when the algorithm decides to make a change in the interrupt moderation parameters. This callback will be scheduled and run in a separate thread in order not to add overhead to the data flow. After the work is done, Net DIM algorithm needs to be set to the proper state in order to move to the next iteration.

Example

The following code demonstrates how to register a driver to Net DIM. The actual usage is not complete but it should make the outline of the usage clear.

#include <linux/dim.h>

/* Callback for net DIM to schedule on a decision to change moderation */
void my_driver_do_dim_work(struct work_struct *work)
{
      /* Get struct dim from struct work_struct */
      struct dim *dim = container_of(work, struct dim,
                                     work);
      /* Do interrupt moderation related stuff */
      ...

      /* Signal net DIM work is done and it should move to next iteration */
      dim->state = DIM_START_MEASURE;
}

/* My driver's interrupt handler */
int my_driver_handle_interrupt(struct my_driver_entity *my_entity, ...)
{
      ...
      /* A struct to hold current measured data */
      struct dim_sample dim_sample;
      ...
      /* Initiate data sample struct with current data */
      dim_update_sample(my_entity->events,
                        my_entity->packets,
                        my_entity->bytes,
                        &dim_sample);
      /* Call net DIM */
      net_dim(&my_entity->dim, dim_sample);
      ...
}

/* My entity's initialization function (my_entity was already allocated) */
int my_driver_init_my_entity(struct my_driver_entity *my_entity, ...)
{
      ...
      /* Initiate struct work_struct with my driver's callback function */
      INIT_WORK(&my_entity->dim.work, my_driver_do_dim_work);
      ...
}

Tuning DIM

Net DIM serves a range of network devices and delivers excellent acceleration benefits. Yet, it has been observed that some preset configurations of DIM may not align seamlessly with the varying specifications of network devices, and this discrepancy has been identified as a factor to the suboptimal performance outcomes of DIM-enabled network devices, related to a mismatch in profiles.

To address this issue, Net DIM introduces a per-device control to modify and access a device’s rx-profile and tx-profile parameters: Assume that the target network device is named ethx, and ethx only declares support for RX profile setting and supports modification of usec field and pkts field (See the data structure: struct dim_cq_moder).

You can use ethtool to modify the current RX DIM profile where all values are 64:

$ ethtool -C ethx rx-profile 1,1,n_2,2,n_3,n,n_n,4,n_n,n,n

n means do not modify this field, and _ separates structure elements of the profile array.

Querying the current profiles using:

$ ethtool -c ethx
...
rx-profile:
{.usec =   1, .pkts =   1, .comps = n/a,},
{.usec =   2, .pkts =   2, .comps = n/a,},
{.usec =   3, .pkts =  64, .comps = n/a,},
{.usec =  64, .pkts =   4, .comps = n/a,},
{.usec =  64, .pkts =  64, .comps = n/a,}
tx-profile:   n/a

If the network device does not support specific fields of DIM profiles, the corresponding n/a will display. If the n/a field is being modified, error messages will be reported.

Dynamic Interrupt Moderation (DIM) library API

struct dim_cq_moder

Structure for CQ moderation values. Used for communications between DIM and its consumer.

Definition:

struct dim_cq_moder {
    u16 usec;
    u16 pkts;
    u16 comps;
    u8 cq_period_mode;
    struct rcu_head rcu;
};

Members

usec

CQ timer suggestion (by DIM)

pkts

CQ packet counter suggestion (by DIM)

comps

Completion counter

cq_period_mode

CQ period count mode (from CQE/EQE)

rcu

for asynchronous kfree_rcu

struct dim_irq_moder

Structure for irq moderation information. Used to collect irq moderation related information.

Definition:

struct dim_irq_moder {
    u8 profile_flags;
    u8 coal_flags;
    u8 dim_rx_mode;
    u8 dim_tx_mode;
    struct dim_cq_moder __rcu *rx_profile;
    struct dim_cq_moder __rcu *tx_profile;
    void (*rx_dim_work)(struct work_struct *work);
    void (*tx_dim_work)(struct work_struct *work);
};

Members

profile_flags

DIM_PROFILE_*

coal_flags

DIM_COALESCE_* for Rx and Tx

dim_rx_mode

Rx DIM period count mode: CQE or EQE

dim_tx_mode

Tx DIM period count mode: CQE or EQE

rx_profile

DIM profile list for Rx

tx_profile

DIM profile list for Tx

rx_dim_work

Rx DIM worker scheduled by net_dim()

tx_dim_work

Tx DIM worker scheduled by net_dim()

struct dim_sample

Structure for DIM sample data. Used for communications between DIM and its consumer.

Definition:

struct dim_sample {
    ktime_t time;
    u32 pkt_ctr;
    u32 byte_ctr;
    u16 event_ctr;
    u32 comp_ctr;
};

Members

time

Sample timestamp

pkt_ctr

Number of packets

byte_ctr

Number of bytes

event_ctr

Number of events

comp_ctr

Current completion counter

struct dim_stats

Structure for DIM stats. Used for holding current measured rates.

Definition:

struct dim_stats {
    int ppms;
    int bpms;
    int epms;
    int cpms;
    int cpe_ratio;
};

Members

ppms

Packets per msec

bpms

Bytes per msec

epms

Events per msec

cpms

Completions per msec

cpe_ratio

Ratio of completions to events

struct dim

Main structure for dynamic interrupt moderation (DIM). Used for holding all information about a specific DIM instance.

Definition:

struct dim {
    u8 state;
    struct dim_stats prev_stats;
    struct dim_sample start_sample;
    struct dim_sample measuring_sample;
    struct work_struct work;
    void *priv;
    u8 profile_ix;
    u8 mode;
    u8 tune_state;
    u8 steps_right;
    u8 steps_left;
    u8 tired;
};

Members

state

Algorithm state (see below)

prev_stats

Measured rates from previous iteration (for comparison)

start_sample

Sampled data at start of current iteration

measuring_sample

A dim_sample that is used to update the current events

work

Work to perform on action required

priv

A pointer to the struct that points to dim

profile_ix

Current moderation profile

mode

CQ period count mode

tune_state

Algorithm tuning state (see below)

steps_right

Number of steps taken towards higher moderation

steps_left

Number of steps taken towards lower moderation

tired

Parking depth counter

enum dim_cq_period_mode

Modes for CQ period count

Constants

DIM_CQ_PERIOD_MODE_START_FROM_EQE

Start counting from EQE

DIM_CQ_PERIOD_MODE_START_FROM_CQE

Start counting from CQE (implies timer reset)

DIM_CQ_PERIOD_NUM_MODES

Number of modes

enum dim_state

DIM algorithm states

Constants

DIM_START_MEASURE

This is the first iteration (also after applying a new profile)

DIM_MEASURE_IN_PROGRESS

Algorithm is already in progress - check if need to perform an action

DIM_APPLY_NEW_PROFILE

DIM consumer is currently applying a profile - no need to measure

Description

These will determine if the algorithm is in a valid state to start an iteration.

enum dim_tune_state

DIM algorithm tune states

Constants

DIM_PARKING_ON_TOP

Algorithm found a local top point - exit on significant difference

DIM_PARKING_TIRED

Algorithm found a deep top point - don’t exit if tired > 0

DIM_GOING_RIGHT

Algorithm is currently trying higher moderation levels

DIM_GOING_LEFT

Algorithm is currently trying lower moderation levels

Description

These will determine which action the algorithm should perform.

enum dim_stats_state

DIM algorithm statistics states

Constants

DIM_STATS_WORSE

Current iteration shows worse performance than before

DIM_STATS_SAME

Current iteration shows same performance than before

DIM_STATS_BETTER

Current iteration shows better performance than before

Description

These will determine the verdict of current iteration.

enum dim_step_result

DIM algorithm step results

Constants

DIM_STEPPED

Performed a regular step

DIM_TOO_TIRED

Same kind of step was done multiple times - should go to tired parking

DIM_ON_EDGE

Stepped to the most left/right profile

Description

These describe the result of a step.

int net_dim_init_irq_moder(struct net_device *dev, u8 profile_flags, u8 coal_flags, u8 rx_mode, u8 tx_mode, void (*rx_dim_work)(struct work_struct *work), void (*tx_dim_work)(struct work_struct *work))

collect information to initialize irq moderation

Parameters

struct net_device *dev

target network device

u8 profile_flags

Rx or Tx profile modification capability

u8 coal_flags

irq moderation params flags

u8 rx_mode

CQ period mode for Rx

u8 tx_mode

CQ period mode for Tx

void (*rx_dim_work)(struct work_struct *work)

Rx worker called after dim decision

void (*tx_dim_work)(struct work_struct *work)

Tx worker called after dim decision

Return

0 on success or a negative error code.

void net_dim_free_irq_moder(struct net_device *dev)

free fields for irq moderation

Parameters

struct net_device *dev

target network device

void net_dim_setting(struct net_device *dev, struct dim *dim, bool is_tx)

initialize DIM’s cq mode and schedule worker

Parameters

struct net_device *dev

target network device

struct dim *dim

DIM context

bool is_tx

true indicates the tx direction, false indicates the rx direction

void net_dim_work_cancel(struct dim *dim)

synchronously cancel dim’s worker

Parameters

struct dim *dim

DIM context

struct dim_cq_moder net_dim_get_rx_irq_moder(struct net_device *dev, struct dim *dim)

get DIM rx results based on profile_ix

Parameters

struct net_device *dev

target network device

struct dim *dim

DIM context

Return

DIM irq moderation

struct dim_cq_moder net_dim_get_tx_irq_moder(struct net_device *dev, struct dim *dim)

get DIM tx results based on profile_ix

Parameters

struct net_device *dev

target network device

struct dim *dim

DIM context

Return

DIM irq moderation

void net_dim_set_rx_mode(struct net_device *dev, u8 rx_mode)

set DIM rx cq mode

Parameters

struct net_device *dev

target network device

u8 rx_mode

target rx cq mode

void net_dim_set_tx_mode(struct net_device *dev, u8 tx_mode)

set DIM tx cq mode

Parameters

struct net_device *dev

target network device

u8 tx_mode

target tx cq mode

bool dim_on_top(struct dim *dim)

check if current state is a good place to stop (top location)

Parameters

struct dim *dim

DIM context

Description

Check if current profile is a good place to park at. This will result in reducing the DIM checks frequency as we assume we shouldn’t probably change profiles, unless traffic pattern wasn’t changed.

void dim_turn(struct dim *dim)

change profile altering direction

Parameters

struct dim *dim

DIM context

Description

Go left if we were going right and vice-versa. Do nothing if currently parking.

void dim_park_on_top(struct dim *dim)

enter a parking state on a top location

Parameters

struct dim *dim

DIM context

Description

Enter parking state. Clear all movement history.

void dim_park_tired(struct dim *dim)

enter a tired parking state

Parameters

struct dim *dim

DIM context

Description

Enter parking state. Clear all movement history and cause DIM checks frequency to reduce.

bool dim_calc_stats(struct dim_sample *start, struct dim_sample *end, struct dim_stats *curr_stats)

calculate the difference between two samples

Parameters

struct dim_sample *start

start sample

struct dim_sample *end

end sample

struct dim_stats *curr_stats

delta between samples

Description

Calculate the delta between two samples (in data rates). Takes into consideration counter wrap-around. Returned boolean indicates whether curr_stats are reliable.

void dim_update_sample(u16 event_ctr, u64 packets, u64 bytes, struct dim_sample *s)

set a sample’s fields with given values

Parameters

u16 event_ctr

number of events to set

u64 packets

number of packets to set

u64 bytes

number of bytes to set

struct dim_sample *s

DIM sample

void dim_update_sample_with_comps(u16 event_ctr, u64 packets, u64 bytes, u64 comps, struct dim_sample *s)

set a sample’s fields with given values including the completion parameter

Parameters

u16 event_ctr

number of events to set

u64 packets

number of packets to set

u64 bytes

number of bytes to set

u64 comps

number of completions to set

struct dim_sample *s

DIM sample

struct dim_cq_moder net_dim_get_rx_moderation(u8 cq_period_mode, int ix)

provide a CQ moderation object for the given RX profile

Parameters

u8 cq_period_mode

CQ period mode

int ix

Profile index

struct dim_cq_moder net_dim_get_def_rx_moderation(u8 cq_period_mode)

provide the default RX moderation

Parameters

u8 cq_period_mode

CQ period mode

struct dim_cq_moder net_dim_get_tx_moderation(u8 cq_period_mode, int ix)

provide a CQ moderation object for the given TX profile

Parameters

u8 cq_period_mode

CQ period mode

int ix

Profile index

struct dim_cq_moder net_dim_get_def_tx_moderation(u8 cq_period_mode)

provide the default TX moderation

Parameters

u8 cq_period_mode

CQ period mode

void net_dim(struct dim *dim, struct dim_sample end_sample)

main DIM algorithm entry point

Parameters

struct dim *dim

DIM instance information

struct dim_sample end_sample

Current data measurement

Description

Called by the consumer. This is the main logic of the algorithm, where data is processed in order to decide on next required action.

void rdma_dim(struct dim *dim, u64 completions)

Runs the adaptive moderation.

Parameters

struct dim *dim

The moderation struct.

u64 completions

The number of completions collected in this round.

Description

Each call to rdma_dim takes the latest amount of completions that have been collected and counts them as a new event. Once enough events have been collected the algorithm decides a new moderation level.