Daytone: A Performant Discrete-Event Simulator for Network Simulations

Developed with the Rust programming language, Daytone has been designed as a performant discrete-event simulator for network simulations. It uses stackless coroutines in asynchronous Rust to implement the actor model, where each actor can only interact with its counterparts using message passing. It excels at both small-scale and large-scale network simulations due to its ability to use either a single-threaded or a multi-threaded runtime; the single-threaded runtime minimizes overhead, while the multi-threaded runtime utilizes multiple CPU cores concurrently in the same simulation run.

To run a network simulation session using a configuration file, run:

RUST_LOG=debug daytone configs/simple.toml

where RUST_LOG levels can be error, warn, info, debug, and trace.

Running Unit and Integration Tests

To run both unit and integration tests, use the command:

cargo test --features test

Or if cargo-nextest has been installed (by using the command brew install cargo-nextest), run the following command instead to take advantage of its speed and more streamlined user interface:

cargo nextest run --all-features

Configuration Settings

In Daytone, all configuration settings are read from a configuration file when a simulation session starts, and the configuration file follows the TOML format for the sake of simplicity and readability.

Here is an example configuration file simple.toml:

# An example configuration file that shows a basic setup.

seed = 1000
edges = [[0, 1], [0, 2]]
hosts = [0, 1, 2]
duration = 20.0
ui_interval = 1.0
report_interval = 2.0
num_threads = 1
log_path = "./simple"

[switch]
port_rate = 8000
capacity = 100
weights = [1]
discipline = "FIFO"
drop = "RED"

[[flow]]
flow_type = "PacketDistribution"
graph = [[0, 1]]
[flow.traffic]
    initial_delay = 1.0
    size = 10000
    arr_dist = {type = "Exp", lambda = 1.0}
    pkt_size_dist = {type = "Uniform", low = 1000, high = 1500}

[[flow]]
flow_type = "PacketDistribution"
graph = [[1, 0]]
[flow.traffic]
    initial_delay = 2.0
    duration = 10.0
    arr_dist = {type = "Uniform", low = 1, high = 1}
    pkt_size_dist = {type = "Uniform", low = 1000, high = 1500}

[[collective]]
collective_type = "Broadcast"
flow_type = "PacketDistribution"
flow_count = 2
graph = [[0, 1], [0, 2]]
sources = [0]
sinks = [1, 2]
[collective.traffic]
    initial_delay = 1.0
    duration = 10.0
    arr_dist = {type = "Uniform", low = 1, high = 1}
    pkt_size_dist = {type = "Uniform", low = 1000, high = 1500}

The following introduces all the possible settings in the configuration file.

General

seed

The seed for the random number generator to ensure reproducibility of the simulation results.

• Valid value: Integer

• Required: Yes

• Example:

  seed = 1000
  

duration

The total duration of the simulation in seconds.

• Valid value: Floating point number

• Required: No

• Default: 1500.0

• Example:

  duration = 20.0
  

report_interval

The report_interval parameter specifies the time interval to log reports from all elements in the simulation.

• Valid value: Floating point number

• Required: No

• Default: f64::MAX (No reports when the simulation is in progress, all reports are logged at the end.)

• Example:

  report_interval = 1.0
  

ui_interval

Daytone provides a user interface, which for now includes a progress bar to visualize the progression of a simulation session. This ui_interval element specifies the time interval to advance the position of the progress bar.

• Valid value: Floating point number

• Required: No

• Default: duration / 100

• Example:

  ui_interval = 1.0
  

num_threads

The number of threads to be used in the simulation run. By setting this value to 1, the single-threaded runtime will be used instead of the default multi-threaded runtime.

log_path

Daytone generates three CSV files, sources.csv, sinks.csv, and switches.csv, containing statistics of a simulation session. log_path specifies the directory of the three CSV files.

• Valid value: String

• Required: No

• Default: ./output

• Example:

  log_path = "./test"
  

topology

Daytone supports arbitrary topologies. Besides widely-used topologies FatTree and Torus, any topology that can be specified as an undirected graph can be supported as well.

• FatTree

  To specify the FatTree topology, it is required to specific k. k is a multiple of 2.

  Example:

  [topology]
  category = "FatTree"
  
  [topology.torus]
    	k = 8
  

• Torus

  To specify the Torus topology, it is required to specific dimension dim, and node per dimension n. Only 1D, 2D, and 3D Torus topologies are supported. That is, valid values of dim are 1, 2, 3.

  Example:

  [topology]
  category = "Torus"
  
  [topology.torus]
    	dim = 2
    	n = 3
  

• Custom topology

  Use edges to specify an undirected graph as the topology, and hosts to specify hosts in the topology.

  edges is a vector of [integer, integer]. An [integer, integer] pair presents an edge in the undirected graph.

  hosts is a vector of integers

  Example:

  edges = [[0, 1], [0, 2]]
  hosts = [0, 1, 2]
  

Switch

In Daytone, all switches use the same setting which can be specified with the following attributes.

port_rate

The bit rate of each outbound port.

• Valid value: Floating point number

• Required: Yes

• Example:

  port_rate = 8000
  

capacity

The capacity (buffer size) of each outbound port.

• Valid value: Integer

• Required: Yes

• Example:

  capacity = 100
  

drop

The packet drop strategy that drops packets when the buffer is full.

• Valid value:

  Value	Meaning
  TailDrop	Dropping packets at the tail of the queue
  RED	Random Early Detection

• Required: Yes

• Example:

  capacity = 100
  

discipline

The scheduling discipline.

• Valid value:

  Value	Meaning	Notes
  FIFO	First In First Out
  DRR	Deficit Round Robin	Required to specify weights
  WFQ	Weighted Fair Queueing	Required to specify weights
  SP	Static Priority	Required to specify priorities
  VC	Virtual Clock	Required to specify vticks

• Required: Yes

• Example:

  discipline = "FIFO"
  

weights

• Valid value: Vector of integers

• Required: Yes if dispcipline = "DRR" or dispcipline = "WFQ"

• Example:

  weights = [1, 2, 3]
  

priorities

• Valid value: Vector of integer values, where the index of the vector is the flow class, and the value is the priority of this flow class

• Required: Yes if dispcipline = "SP"

• Example:

  priorities = [1, 2]
  

vticks

• Valid value: Vector of (integer, integer), where the first integer is the flow class and the second integer is the inverse of the desired rates for the corresponding flows, in bits per second

• Required: Yes if dispcipline = "VC"

• Example:

  vticks = [(0, 2), (1, 1)]
  

Flow

In Daytone, flows can be specified one by one:

[[flow]]
flow_id = 2
starts_before = [3]
starts_after = [1]
flow_type = "PacketDistribution"
graph = [[0, 1]]
[flow.traffic]
    initial_delay = 1.0
    duration = 2.0
    arr_dist = {type = "Uniform", low = 1, high = 1}
    pkt_size_dist = {type = "Uniform", low = 1000, high = 1500}

or by sets:

[[flow_set]]
first_flow_id = 10
flow_type = "PacketDistribution"
flow_count = 10
[flow_set.traffic]
    initial_delay = 0.0
    duration = 10.0
    arr_dist = {type = "Uniform", low = 0.0008, high = 0.0008}  # 10Mbps
    pkt_size_dist = {type = "Uniform", low = 1024, high = 1024}

The following table lists required, optional, or not supported attributes of a flow or a flow set.

Attribute	Meaning	flow	flow_set
flow_id	The id of the flow	optional	no
first_flow_id	The smallest flow id of the flow set	no	optional
starts_before	The ids of flows that cannot start until this flow / flow set ends	optional	optional
starts_after	The ids of flows that this flow / flow set must wait for them to end before it starts	optional	optional
flow_type	The type of the flow or flows of the flow set	required	required
flow_count	The number of flows in the flow set	no	required
graph	The pair of the source host and the sink host of the flow	required	no
path	The path of the flow	optional	no
traffic	The traffic of the flow / flow set	required	required

flow_id

The id of the flow. If not specified, the id of the first flow specified in the configuration file will be 0, and the subsequent ids of flows increase by 1.

• Valid value: Integer

• Required: No

• Example:

  flow_id = 1
  

first_flow_id

The smallest flow id of the flow set.

• Valid value: Integer

• Required: No

• Example:

  flow_first_id = 1
  

starts_before

The ids of flows that cannot start until this flow / flow set ends.

• Valid value: Vector of integers

• Required: No

• Example:

  starts_before = [3, 4]
  

starts_after

The ids of flows that this flow / flow set must wait for them to end before it starts.

• Valid value: Vector of integers

• Required: No

• Example:

  starts_after = [0, 1]
  

flow_type

The type of the flow or flows of the flow set.

• Valid value:

  Value	Meaning
  PacketDistribution	A flow whose packet source sends packets with specific distributions of inter-arrival times and packet sizes
  TCP	A flow whose packet source simulate the TCP protocol

• Required: Yes

• Example:

  flow_type = "PacketDistribution"
  

flow_count

The number of flows in the flow set.

• Valid value: Integer

• Required: Yes for a flow set

• Example:

  flow_count = 10
  

graph

The pair of the source host and the sink host of the flow.

• Valid value: [[integer, integer]]

• Required: Yes for a single flow

• Example:

  graph = [[0, 2]]
  

path

The path of the flow.

• Valid value: Vector of integers where each integer is a node in the path

• Required: No

• Example:

  path = [0, 1, 2]
  

traffic

For a flow, it must specify its traffic under [flow.traffic]. For a flow set, it must specify the traffic of its flows under [flow_set.traffic].

The following table lists attributes of traffic.

Note:

• initial_delay, arr_dist, and pkt_size_dist are required.
• Either size or duration is required.
• If flow_type = "TCP", [flow.traffic.tcp] or [flow_set.traffic.tcp] must be specified for the flow or flow set.

• Attributes:

  Attribute	Meaning	Valid Value
  initial_delay	The seconds the flow / flow set waits before producing its first packet	Floating point number
  size	The total size of packets of the flow / each flow of the flow set in bytes	Integer
  duration	The duration of the flow / each flow of the flow set in seconds	Floating point number
  arr_dist	The arrival distribution of packets in seconds	Valid distribution info
  pkt_size_dist	The distribution of packet sizes in bytes	Valid distribution info

Valid distribution info includes uniform distribution and exponential distribution:

• {type = "Uniform", low = 0.0008, high = 0.0008}
• {type = "Exp", lambda = 1.0}

• Required: Yes

• Example:

  [[flow_set]]
  flow_type = "TCP"
  flow_count = 100
  [flow_set.traffic]
  	initial_delay = 0.0
  	duration = 0.1
  	arr_dist = {type = "Uniform", low = 0.0008, high = 0.0008}
  	pkt_size_dist = {type = "Uniform", low = 1024, high = 1024}
  [flow_set.traffic.tcp]
  	cc_algorithm = "TCPReno"
  

traffic.tcp

For a flow, it must specify its tcp characteristics under [flow.traffic.tcp]. For a flow set, it must specify the tcp characteristics of its flows under [flow_set.traffic.tcp].

• Attributes:

  Attribute	Meaning	Valid Value
  cc_algorithm	The congestion control algorithm	TCPReno, TCPCubic

• Required: Yes

• Example:

  [flow.traffic.tcp]
  	cc_algorithm = "TCPReno"
  

Collective

In Daytone, collectives can be specified one by one:

[[collective]]
collective_type = "Broadcast"
flow_type = "PacketDistribution"
flow_count = 2
graph = [[0, 2], [0, 3], [1, 2], [1, 3]]
sources = [0, 1]
sinks = [2, 3]
[collective.traffic]
    initial_delay = 1.0
    duration = 10.0
    arr_dist = {type = "Uniform", low = 1, high = 2}
    pkt_size_dist = {type = "Uniform", low = 1000, high = 1500}

or by sets:

[[collective_set]]
collective_type = "Gather"
collective_count = 2
flow_type = "PacketDistribution"
flow_count = 2
sources = [[0, 1], [1, 2]]
sinks = [[2, 2], [3, 3]]
[collective_set.traffic]
    initial_delay = 1.0
    duration = 10.0
    arr_dist = {type = "Uniform", low = 3, high = 4}
    pkt_size_dist = {type = "Uniform", low = 2000, high = 2500}

The following table lists required, optional, or not supported attributes of a collective or a collective set.

Attribute	Meaning	collective	collective_set
collective_type	The type of the collective or collective set	required	required
first_flow_id	The smallest flow id of the collective or collective set	optional	optional
collective_count	The number of collectives in the collective set	no	required
flow_type	The type of the flows of the collective or collective set	required	required
flow_count	The number of flows in the collective or in each collective of the collective set	required	required
graph	The pairs of the source host and the sink host of the collective's flows	optional	no
paths	The paths of the collective's flows	optional	no
sources	The sources of flows	optional	optional
sinks	The sinks of flows	optional	optional
traffic	The traffic of the collective / collective set	required	required

collective_type

The type of the collective or collective set.

• Valid value: Brordcast or Gather

• Required: Yes

• Example:

  collective_type = "Broadcast"
  

first_flow_id

The smallest flow id of the collective / collective set.

• Valid value: Integer

• Required: No

• Example:

  flow_first_id = 1
  

collective_count

The number of collectives in the collective set.

• Valid value: Integer

• Required: Yes for a collective set

• Example:

  collective_count = 10
  

flow_count

The number of flows in the collective or in each collective of the collective set.

• Valid value: Integer

• Required: Yes

• Example:

  flow_count = 4
  

graph

The pairs of the source hosts and the sink hosts of the collective's flows.

• Valid value: Vector of [integer, integer]

• Required: No

• Example:

  graph = [[0, 1], [0, 2]]
  

paths

The paths of the collective's flows.

• Valid value: Vector of vectors of integers where each integer is a node in the path

• Required: No

• Example:

  paths = [[0, 1], [0, 1, 2]]
  

sources

For a collective, sources is the set of the source hosts of this collective's flows.

• Valid value: Vector of integers where each integer is a source host

• Required: No

• Example:

  sources = [0, 1]
  

For a collective set, sources is set of the source hosts of this collective set's collectives' flows.

• Valid value: Vector of vectors of integers where each vector is a collective's flows' source hosts

• Required: No

• Example:

  sources = [[0, 0], [1, 1]]
  

sinks

For a collective, sinks is the set of the sink hosts of this collective's flows.

• Valid value: Vector of integers where each integer is a sink host

• Required: No

• Example:

  sinks = [0, 1]
  

For a collective set, sinks is set of the sink hosts of this collective set's collectives' flows.

• Valid value: Vector of vectors of integers where each vector is a collective's flows' sink hosts

• Required: No

• Example:

  sinks = [[0, 0], [1, 1]]
  

traffic

For a collective, it must specify the traffic of its flows under [collective.traffic]. For a collective set, it must specify the traffic of its flows under [collective_set.traffic].

The following table lists attributes of traffic.

Note:

• initial_delay, arr_dist, and pkt_size_dist are required.
• Either size or duration is required.
• If flow_type = "TCP", [flow.traffic.tcp] or [flow_set.traffic.tcp] must be specified for the flow or flow set.

• Attributes:

  Attribute	Meaning	Valid Value
  initial_delay	The seconds the collective / collective set waits before producing its first packet	Floating point number
  size	The total size of packets of each flow of the collective / collective set in bytes	Integer
  duration	The duration of each flow of the collective / collective set in seconds	Floating point number
  arr_dist	The arrival distribution of packets in seconds	Valid distribution info
  pkt_size_dist	The distribution of packet sizes in bytes	Valid distribution info

Valid distribution info includes uniform distribution and exponential distribution:

• {type = "Uniform", low = 0.0008, high = 0.0008}
• {type = "Exp", lambda = 1.0}

• Required: Yes

• Example:

  [[collective]]
  collective_type = "Broadcast"
  flow_type = "PacketDistribution"
  flow_count = 4
  [collective.traffic]
      initial_delay = 1.0
      duration = 10.0
      arr_dist = {type = "Uniform", low = 3, high = 4}
      pkt_size_dist = {type = "Uniform", low = 2000, high = 2500}
  

traffic.tcp

For a collective, it must specify the tcp characteristicsc of its flows under [collective.traffic.tcp]. For a collective set, it must specify the tcp characteristics of its collectives' flows under [collective_set.traffic.tcp].

• Attributes:

  Attribute	Meaning	Valid Value
  cc_algorithm	The congestion control algorithm	TCPReno, TCPCubic

• Required: Yes

• Example:

  [collective.traffic.tcp]
  	cc_algorithm = "TCPReno"