Maple

An energy assessment by Groundlake.

Date: July 2021

Aims

Maple is a single server that acts as a central point for running self-hosted personal network services.

The aims of running the server are:

• To control the hardware and software stack as much as possible, as an experiment in learning and understanding this stack.
• To provide a flexible piece of infrastructure that can have software installed and uninstalled without need for third party administrators or policies.
• To consider how self-hosting impacts both positively and negatively on hosters, administators and users.
• To allow cheap or free use of self-hosting software.

Summary of suggested actions

• Consider upgrading PHP to 7.4 for stable, or 8.0 for unstable

• Consider automated upgrades to server software, and the following components:

• 6work.exmosis.net:

  • Wordpress major upgrades to core
  • Auto-convert images to webp
  • Install or change cache plugins to improve Lighthouse assessment

• 6wall.exmosis.net: wallabag upgrades to service code

• 6read.exmosis.net: ttrss upgrades to service code

• Consider generating a static HTML version of 6work.exmosis.net?

Review: Services and Stack

The Maple server is currently responsible for running the following services:

• Self-host 6work.exmosis.net (Wordpress)
  • Dependencies: PHP, nginx, MySQL, Let's Encrypt
• Self-host 6read.exmosis.net (ttrss)
  • Dependencies: PHP, nginx, MySQL, Let's Encrypt
• Self-host 6wall.exmosis.net (wallabag)
  • Dependencies: PHP, nginx, MySQL, Let's Encrypt
• Central file synchronisation server for local network (syncthing)
• Publishing synchronised content to external website (rsync)
• Proxy point for synchronisation: diary, photos, markor, pad (rsync)
• Daily alert service for file change alerts (custom: PHP)
• Local file backup (rsync, custom: bash)
• Backup databases (mysqldump)

Dependent services are:

• PHP 7.3
• nginx 1.14.2
• Let's Encrypt
• MySQL: MariaDB 10.3.27

Dependent operating system is:

• Raspbian 10 (Buster)

Review: Manual Processes

Top-level services

Service	Repeated Manual Processes	Possible Improvement Impact
Wordpress	Major updates to core Wordpress (monthly, 20 mins)	Low: Automate major upgrades to core
ttrss	Updates to service code (monthly, 30 mins)
wallabag	Updates to service code (monthly, 30 mins)
syncthing
rsync
mysqldump

Custom services

Service	Repeated Manual Processes	Possible Improvement Impact
Daily alert service: PHP
Local file backup: bash

Dependency services

Service	Repeated Manual Processes	Possible Improvement Impact
PHP 7.3		Upgrade to PHP 7.4 for stable, 8.0 for unstable
nginx 1.14.2
Let's Encrypt
MySQL: MariaDB 10.3.27

Operating system

Service	Repeated Manual Processes	Possible Improvement Impact
Software updates	Updates applied

Review: Caching and optimisation

Wordpress

• Caching plugin enabled: yes
• Image optimisation plugin enabled: no: Install relevant plugin to reduce image sizes and transfer times

PHP

• OPCache enabled: yes
• Latest version: no: Update to the latest version to benefit from optimisations

nginx

• gzip enabled: yes

MySQL

Review: Energy usage and efficiency

This assessment uses a standard set of review metris to identify how the server uses energy:

• Energy rates, measured in Watts (W) and kiloWatts (kW)
• Energy usage, measured in kiloWatt-hours (kWh)
• Data sizes, measured in kilobytes (kb), megabytes (mb) and upwards
• CPU usage: server load

As standard, we measure usage across different contexts relating to the services above:

• Hardware: Overall energy used by the server hardware directly.
• Software: If possible, how energy is split between services and used by custom code.
• Client: Page weight, data transferred, and distance travelled. CPU intensity.

As energy usage varies widely across time and activity, this assessment aims to triangulate a generally accurate sense of usage through a variety of means.

Hardware and system

The Maple server runs on a Raspberry Pi Model B (+?).

Energy measurement method:

• Software: powerstat, powertop

• USB meter (whole hardware system)

• Energy rate on idle: xW

• Energy rate under stress: xW

Duration-based samples used in assessment: Single day, single week. We suggest a monthly assessment is also appropriate, especically given monthly manual processes identified for services, above.

• Energy used on sample day 1: kWh
• Average daily energy use during sample week 1: xkWh (total xkWh / 7)

System load:

Load measurement tools:

• sar (sysstat package), top, nmon

Information:

• No of CPUs: 1

Readings:

• Sample 1 (snapshot, average load):

  • Average load sample 1: 0.9-1.0
  • Average draw: 5.1V, 0.35A = 1.785W (extrapolated to 1 day = 42.84Wh)

• Sample 2 (duration, day 1 (Wednesday 3pm)): 7800 mAh = 7.8 Ah. At 5.1V x 7.8 = 39.78Wh. Difference to average: -3.06Wh

• Sample 3 (duration, day 2 (Thursday 3pm)): 6875 mAh = 6.875 Ah.

• Sample x (duration, day x (Saturday, 4pm)): 28900 mAh

Steps run:

• sysstat package installed via apt
• powertop package installed via apt
• (linux-perf package installed via apt, not used)
• nmon installed via apt

End users

We identified 3 key web-based user interactions to assess, based on the services and needs identified above:

1. Page load of 6work.exmosis.net homepage
2. Page load (logged in) of 6read.exmosis.net homepage
3. Page load (logged in) of 6wall.exmosis.net homepage

Review: Hardware production

We are currently investigating production processes for the Raspberry Pi line.

Review: Energy supply

The maple server sits within a home network, attached to the main power supply for the home.

The energy provider for this location is the company Pure Planet, and the tariff used is based on a commitment to 100% renewable energy. For electricity, the company employs Renewable Energy Guarantees of Origin (REGOs) to supply this.