SLA inversion

With an SLA inversion , a highly available technical system is dependent on other systems, which means that the average time in which the system is available falls below the availability defined in the Service Level Agreement (SLA).

Calculation of availability

The overall availability of a system S , which is dependent on external systems E , is calculated from:

${\displaystyle P\left(S\vert \mathrm {verf{\ddot {u}}gbar} \right)=\left(1-P\left(S\vert \mathrm {ausgefallen} \right)\right)\cdot \prod _{n}P\left(E_{n}\vert \mathrm {verf{\ddot {u}}gbar} \right)}$

For example, if a system is 99.99% available, but depends on two external systems with 99.9% availability each, the overall system is only 99.99% 99.9% 99.9% = 99.79% available. The system thus falls on average for

${\displaystyle 1\mathrm {a} \cdot (1-99{,}79\ \%)=18{,}4\mathrm {h} }$

in the year.

Troubleshooting

SLAs should be agreed for business processes and not for overall systems. The development department can then prioritize important business processes over less important business processes. In addition, the development department cannot make SLAs for third-party systems. However, it is possible to decouple the systems from one another so that the system continues to function with limited functionality if external systems have failed. This purpose is served, for example, fuses and enkoppelte middleware , such as an enterprise service bus .

swell

1. ↑ Michael T. Nygard: Release It! Design and Deploy Production-Ready Software. O'Reilly, 2007, ISBN 978-0-9787392-1-8 , 4.10 SLA Inversion (English, 326 pages).