• About
• Features
• Get involved!
• Promote
• Funding and donations
• Help
• BETA status
• Contacts
• Feedback

Server reliability and uptime policy

Date of next review due: 1st October 2015.

This page details our arrangements for reliability and uptime which apply unless otherwise stated contractually.

Applicability

Details here apply to all of:

• the CycleStreets website (including area-specific subdomains);
• calls to our API via a registered API key, e.g. from mobile apps, external websites and multi-modal planner interfaces;
• user interfaces we have created which run on the server from the same codebase and database (e.g. Cycling Sorted);
• proxied sites which we host, running under different domain names.

Downtime, monitoring and switchover

Our main and development servers are monitored via a ping service and 5-minutely custom API monitoring, both resulting in SMS messages to the sysops/developers if no response or an invalid is receiving.

We aim to exceed 99% uptime.

Historical downtime since December 2008 has been well in excess of this:

• no periods of extended downtime;
• two several-hour period of downtime due to upstream network connectivity issues;
• occasional 10-minute scheduled downtimes to initiate backups during quieter periods.

Hot-running standby fallback server provision accompanies the live site and API. Switchover currently requires some manual intervention to ensure full database and data synchronicity.

DNS propagation periods are kept low for our domains to ensure quick repointing to a different server in an emergency scenario.

Non-emergency reboots are minimised and are scheduled for out-of-office periods; UK night-time periods are used when possible.

Response times

During the hours of 10-6pm, e-mail to our main contact address, info@cyclestreets.net, relating to uptime-related matters, will generally get a response within 2-4 hours if not immediately.

Outside office hours, we strongly aim to reports of uptime-related problems, though this cannot be guaranteed. Restoration following downtime is regarded as highest priority above all other work.

The sysops/developers make efforts to avoid holidays at the same time.

Phone-based support is not available unless otherwise arranged, e.g. via a support contract.

Backup

Code repository copies exist in multiple places so can be restored easily.

Backups of user material (files and database entries) such as planned/saved journeys, Photomap images and more, are made on a daily basis and transferred to fallback server(s).

All backups are made to a machine in a different datacentre, and additional off-site backups are in practice made every few months.

Backup data is transferred using secure protocols.

Servers

We use self-hosted colo server provision, with a support contract from our hosts, using servers in more than one geographical location (currently London and Cambridge).

Our hosts (Mythic Beasts) use Telecity and Redstone, which are both ISO27001 certified.

Servers are regularly patched for security.

All server-level access is via secure protocols.

Server responsiveness

We strongly aim to run routing data imports weekly. This is normally run on background servers to avoid lack of responsivity of main servers.

API keys

Use of an API key is subject to agreement with the API usage policy.

Support contracts

SLA-supported support contracts are available; please contact us.

(We do not normally issue API keys to Local Authority and similar official clients without such a contract.)

If your organisation requires increased resilience beyond what is outlined above, please contact us to discuss this; we are happy to consider additional resilience subject to covering of costs.