Dependencies don't come for free. Even if all of the individual services are built to a very high quality, the reliability of the whole falls below the reliability of an individual service. For example, if the Hotel Search service is dependent on 5 other services (for simplicity, we will prune the dependency graph here), and it is built to 99.99%, availability, the actual availability of the overall Hotel Search feature is 99.99% which translates to about 99.94%. To consider the impact, lets see what 99.99% uptime means in terms of availability:

• Daily: 8.6 seconds
• Weekly: 1 minute 0.5 seconds
• Monthly: 4 minutes 23.0 seconds
• Yearly: 52 minutes 35.7 seconds

 With 99.94%, the availability is as follows:

• Daily: 51.8 seconds
• Weekly: 6 minutes 2.9 seconds
• Monthly: 26 minutes 17.8 seconds
• Yearly: 5 hours 15 minutes 34.2 seconds

This means that with just five dependents, all of a high quality, the overall system degrades with an extra five hours of downtime per year.

With this impact in mind, let's look at mitigation strategies. But before that, there is another interesting issue that service dependencies bring out.