How to Choose the Right Server to Avoid Single Point of Failure

Deploy Divergence - Avoid Single Point of Failure

For virtual hosts/VPS from the same vendor, purchasing multiple small servers in a few physical servers in a few data centers may not provide adequate protection against single points of failure. Only by deploying across data centers, vendors, or even planets/galaxies can single points of failure be maximally avoided. However, even doing so cannot guarantee the service reliability one hundred percent.

The Importance of SLA

If a server provider claims to achieve a 99.99% SLA, it is already quite good on Earth. Once the SLA is violated, some compensation plans will be provided. However, in reality, even if you use multiple servers to mitigate risks, when that 0.01% failure occurs, most servers may be affected. This level of downtime is not caused by human factors, and both single servers and multiple servers face the risk of downtime.

CDN Always Online

If you are concerned about single points of failure, you can learn about CDN Always Online, which may alleviate some unnecessary concerns.

Economic Considerations

If you are concerned about cost-effectiveness, you can consider splitting small servers to deploy single sites one by one. However, the load on each server may be very low and could result in some waste. Would it be better to go for a big server? You will face a similar problem, but you can console yourself by thinking that 'adding a few more sites will at least reach 50%, not a loss'.

Ease of management

In the presence of tools like RunCloud, the management cost of multiple servers compared to a single server should be manageable, but in any case, managing a single server is easier. If you have more than 5 servers, you may need a dedicated DevOps to help you, or you should be proficient in tools like Chef/Puppet yourself.

Buy big or buy small?

In cloud computing, the fundamental purpose is to purchase a sufficiently usable small server (+0.5 times redundancy). It can run as long as it is sufficient, and upgrade when expansion is needed. Good vendors can allow you to continuously upgrade from a 512MB memory VPS (for analogy only) to a dedicated/bare metal server that costs tens of thousands of dollars per month (also for analogy only). Leave enough redundancy and upgrade when the capacity is insufficient. Success does not necessarily require frequent downtime, the key is to have a sense of 'elastic scalability'.