Awesome

mirakc

A Mirakurun-compatible PVR backend written in Rust

Features

• Mirakurun-compatible REST API
  • Enough to work with applications such as EPGStation
• Standalone recording
  • No need to install additional applications anymore
• On-air time tracking
  • No need to worry about recording failures anymore
• Timeshift recording
  • No need to add recording schedules anymore

Quick Start

If you already have a TV recording system built with Mirakurun and EPGStation, you can simply replace Mirakurun with mirakc.

It's recommended to use a Docker image which can be downloaded from Docker Hub. See docs/docker.md about how to make a custom Docker image for your environment.

Launch a mirakc Docker container

Create config.yml:

epg:
  cache-dir: /var/lib/mirakc/epg

server:
  addrs:
    - http: '0.0.0.0:40772'

channels:
  # Add channels of interest.
  - name: NHK
    type: GR
    channel: '27'

tuners:
  # Add tuners available on a local machine.
  - name: Tuner0
    types: [GR]
    command: >-
      recpt1 --device /dev/px4video2 {{{channel}}} - -

filters:
  # Optionally, you can specify a command to process TS packets before sending
  # them to a client.
  #
  # The following command processes TS packets on a remote server listening on
  # TCP port 40774.
  decode-filter:
    command: >-
      socat -,cool-write tcp-connect:remote:40774

Create compose.yaml:

services:
  mirakc:
    image: docker.io/mirakc/mirakc:alpine
    container_name: mirakc
    init: true
    restart: unless-stopped
    devices:
      # Add device files used in `tuners[].command`.
      - /dev/px4video2
    ports:
      - 40772:40772
    volumes:
      - mirakc-epg:/var/lib/mirakc/epg
      - ./config.yml:/etc/mirakc/config.yml:ro
    environment:
      TZ: Asia/Tokyo
      RUST_LOG: info

volumes:
  mirakc-epg:
    name: mirakc_epg
    driver: local

And then launch a mirakc container:

docker compose up

You can also launch a mirakc container by using other tools like podman and podman-compose which support Docker-compatible images and compose.yaml:

podman-compose up

You should see the version string when running the following command if the mirakc starts properly:

$ curl -fsSL http://localhost:40772/api/version
{"current":"<version-string>","latest":"<version-string>"}

See docs/config.md for details of config.yml.

Motivation

In these days, you can build a TV recording system by yourself, using a SBC (single-board computer) and OSS (open-source software).

Usually, SBCs are low power consumption. For this reason, SBCs are suitable for systems which works all the time. On the other hand, SBCs are not so powerful unlike PCs in these days. Typically, a SBC have a multi-core CPU with low clock of less than 2GHz, and only 1..4GB memory.

Mirakurun is one of popular DTV tuner servers in Japan. It's fast enough, but consumes much memory. For example, it consumes more than 150MB even when idle.

When Mirakurun provides 8 TS streams simultaneously, it consumes nearly 1GB of memory which includes memory consumption of descendant processes. As a result, some of processes which was spawned by tuner commands will be killed in minutes if Mirakurun is executed on a machine like ROCK64 (DRAM: 1GB) which doesn't have enough memory.

mirakc is more efficient than Mirakurun. It can provide at least 8 TS streams at the same time even on ROCK64 (DRAM: 1GB).

Performance Comparison

After running for 1 day

The following table is a snippet from the result of docker stats at idle after running for 1 day on Raspberry Pi 4B (DRAM: 4GB):

NAME           MEM USAGE / LIMIT    MEM %  NET I/O          BLOCK I/O
mirakc-alpine  19.5MiB / 3.704GiB   0.51%  7.45GB / 38.3MB  246kB / 15.6MB
mirakc-debian  26.47MiB / 3.704GiB  0.70%  7.49GB / 34.6MB  221kB / 15.6MB
mirakurun      123.6MiB / 3.704GiB  3.26%  61.3GB / 290MB   83.5MB / 1.36GB

The environment variable MALLOC_ARENA_MAX=2 is specified in containers.

The mirakurun container installed a few packages at the start time. That increased the amount of I/O data but it seems to be less than several GB.

8 TS streams at the same time

mirakc is lower CPU usage and 75% smaller memory consumption than Mirakurun:

where + means that the value is an amount of increase from its stationary value.

See mirakc/performance-measurements about how to collect the performance metrics.

Limitations

• CS and SKY channel types are not tested at all
  • In addition, no pay-TV channels are tested because I have no subscription for pay-TV
• mirakc doesn't work with BonDriver_Mirakurun
  • Use BonDriver_mirakc instead
  • See the issue #4 for details

TODO

• Use multiple tuners in the EPG task in order to reduce the execution time
  • Currently, it takes about 40 minutes for collecting EIT sections of 12 GR services and 12 BS services

Acknowledgments

mirakc is implemented based on knowledge gained from the following software implementations:

• Mirakurun

License

Licensed under either of

• Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT License (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in this project by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.