System availability: 3 9s, 4 9s in SRE's mouth... What is it?

As a student of server development, the most common sentence I hear is how many 9s are there in your interface? 2 9s? 3 9s?How many 9s are stable?

Today we will start with system availability (Availability) and explain to you the names of indicators commonly used in the development process.

1. System Availability

1.1-What is System Availability

Concept: During a given time interval, the computer system is in a workable stateThe time scale of the state.

Here we focus on two nouns:

• Time interval
  : It means that the system availability is counted within the specified time range.

• Workable state
  : This leads to a formula.That is [system availability = system uptime (time to troubleshoot) / total system operation time]

1.2-the purpose of establishing "system availability"

On the one hand, system availability is strongly related to the construction goal of SRE. The construction goal of SRE is to minimize system failure time and improve system availability.

On the other hand, the concept of system availability seems simple, but there is no fixed calculation method, you need to define it according to your own system.

2. Two ways to measure system availability

At present, there are two main ways to measure system availability in the industry:

• System availability for time latitude.

• Request system availability for latitude.

The calculation methods of the above two availability are:

• System availability at time latitude:

Availability = Program Uptime / (Program Uptime + System Downtime)

• RequestSystem Availability of Latitude:

Availability = Number of successful requests / Total number of requests

3. Time LatitudeThe system availability of

The system availability based on the time dimension, its concept is: from the time dimension, to evaluate the uptime and system availability of the system.

The system availability of the time dimension is the X 9 we often mention.

X 9s indicate the ratio of the normal use time of the system to the total time within the specified time range in units of year/month/day.For example, we take 1 year as the time unit, we can get:

• 3 9s
  : (1-99.9%)
  365
  24=8.76 hours, which means that the maximum possible business interruption time of the system in one year of continuous operation is 8.76 hours.

• 4 9
  : (1-99.99%)
  365
  24=0.876 hours=52.6 minutes, which means that the maximum possible business interruption time of the system in one year of continuous operation is 52.6 minutes.

• 5 9
  : (1-99.999%)
  365
  24*60=5.26 minutes, which means that the maximum possible business interruption time of the system in one year of continuous operation is 5.26 minutes.

According to the above definition, we can summarize a table:

From the above table, we can also draw a

conclusion

:

• The higher the number of 99 in the system, the higher the availability of the system.

4. System availability of requested latitude

The definition of system availability based on requested latitude is relatively simple.

For example, based on 1000 requests, then:

• 90% system availability: means that 1000 of 1000 requests are allowed * (1- 90%) = 100An error occurred in the request.

• System availability 99%: Indicates that 1000 * (1- 99%) = 10 requests are in error out of 1000 requests.

• System availability 99.9%: Indicates that 1000 * (1- 99.9%) = 1 request error is allowed out of 1000 requests.

5. Factors to be considered in the establishment of system availability goals

5.1-Cost

The higher the system availability,The higher the system requirements for you, the higher the cost and price you will pay.

The higher the availability of the system, the higher the requirements for your server's storage, network, software, etc.

And some services do not need to achieve that high availability, so you can set a lower availability target for these services.

5.2-Business Tolerance

System availability also needs to consider business tolerance.

For services in the financial industry, the failure of any request may lead to loss of funds, so for such services, the tolerance for errors is relatively low, which requires system availabilityhigher.

For common social software, such as the comment area function/barrage function we use, for these applications, even if the request fails once, it is acceptable, and the next time the request is successful, it will be fine.Therefore, for these services, the service tolerance is high, and the system availability does not have to be very high.

6. Summary