Working with SLAs

Steve Lydford · 19 May 2015
WmRogers_guarantee, by Nancy Gluck

If you spend any time designing applications which use third-party platforms and services, you will have spent time considering Service Level Agreements (SLAs). The tl;dr part of any SLA, the bit the everyone is actually interested in, is the uptime guarantee. However, this seemingly simple number has some hidden meanings that you should be aware of when designing applications and reporting your SLAs to customers.

Uptime guarantees

This one seems obvious, but of course an uptime guarantee is not actually a guarantee of uptime in the purest sense of the word. It is merely the point at which any penalty clauses in the uptime guarantee section of the service contract come into force. Subtle, but very different. Especially worth considering if the financial reimbursement offered if less than the cost per period of downtime to your business.

Amount of permitted downtime

As mentioned previously, uptime guarantees form the basis of SLAs. Typically these are values in the region of 99% and above. While 99% uptime may sound a lot, it actually means that the service you are using may be unavailable for over 3.5 days per year and still be well within it's guaranteed level of availability.

Below is a table showing the actual allowed downtime values against SLA percentages:

Uptime Guarantee Downtime per Week Downtime per Month Downtime per Year
90% 16.8 hours 72 hours 36.5 days
99% 1.68 hours 7.2 hours 3.65 days
99.9% 10.1 minutes 43.2 minutes 8.76 hours
99.99% 1.01 minutes 4.32 minutes 52.56 minutes
99.999% 6.05 seconds 25.9 seconds 5.26 minutes
99.9999% 0.605 seconds 2.59 seconds 31.5 seconds

Clearly, calculating the cost of this downtime to your business can make a huge difference when choosing a service provider. Five minutes of downtime in the middle of the day will be considerably more financially troublesome to an investment bank than to say a news aggregation website for example.

As systems become increasingly distributed, fueled in part by the adoption of Cloud and 'micro' services, network and hardware fault tolerance is becoming more and more of a consideration for solution design. There are design strategies that we can use to minimize the disruption caused by these, often transient, faults, such as effective use of caches, queues, retry policies, replication and failover, load-balancing, etc., but I'll save these for a future article.

Time frames

When comparing uptime guarantees, it is also useful to compare the reporting periods over which availability is measured, favoring the shorter. A service provider who resets the clock yearly has a much longer period to hide bad spells within the average than one which resets weekly or monthly.

SLA uptimes are composite

My final point for now on SLAs regards the calculation of overall uptime guarantees for a software system. It is important to note, yet less than plainly obvious, that when creating a software system which uses multiple third-party services with independent SLAs, the overall uptime guarantee is a composite.

Let's say for example that we are building a system using three third-party services, each with their own uptime guarantee:

Uptime Guarantee
Service 1 99.9%
Service 2 99.9%
Service 3 99.5%

So what is the maximum downtime that we could expect from this system? It is easy to assume that this would be equal to that offered by the lowest uptime guarantee, in this case 99.5%, or 3.6 hours potential downtime per month, but this is not actually the case.

In this example let's assume that each of the three services are vital to the operation of the system, which may not necessarily always be the case (in reality, the loss of logging, advertising or analytics services for example may not affect the overall system uptime). Each of these three systems may experience outages at different times during the reporting period, so to come up with an overall figure we must multiply these percentages to come up with a composite. In our case this calculation is:

0.999 x 0.999 x 0.995 = 0.993

So our three services combined actually provide an uptime guarantee of 99.3%, or a potential downtime of 5.04 hours per month.


We have seen that, as with all parts of software design, when working with SLAs it is important to pay attention to the detail. Not all SLAs are made equal, and when reporting overall SLA uptimes to customers, simple assumptions can lead to potentially costly mistakes.


Steve Lydford Image

Steve Lydford

Steve is a Software Craftsman and author with over 20 years professional experience. He first discovered his passion for code at the age of seven, when his father brought home a Sinclair Spectrum and a portable black and white television. The countless hours spent typing in listings from Sinclair User magazine sparked an enthusiasm for programming that has stayed with him for life.

During his career, Steve has worked on projects in a wide variety of sectors including retail e-commerce, finance, education, media, government and healthcare, developing large-scale, resilient, distributed systems on an assortment of platforms. He currently specialises in solutions built on the Microsoft .Net stack, with a particular interest in cloud computing using the Microsoft Azure platform.

Steve is a pragmatic problem solver with a focus on the delivery of high-quality, robust and maintainable code. He is a passionate advocate of continuous improvement, both for software and software developers and believes that the future success of the software industry relies on the sharing of knowledge gained through experience.

He is a member of the Microsoft Azure Advisor's Group and the London Software Craftsmanship Community (LSCC).

All author posts
Codurance Logo

Software es nuestra pasión.

Somos Software Craftspeople. Construimos software bien elaborado para nuestros clientes, ayudamos a los desarrolladores a mejorar en su oficio a través de la formación, la orientación y la tutoría y ayudamos a las empresas a mejorar en la distribución de software.

Últimos Blogs


15 Northburgh Street
London EC1V 0JR

Teléfono: +44 207 4902967

2 Mount Street
Manchester M2 5WQ

Teléfono: +44 207 4902967

Carrer Aragó, 208
08011, Barcelona

Teléfono: +34 937 82 28 82

Correo electrónico: