Unplanned Downtime: Data Center vs. On Premise (part two)

As an organization, you want to experience as little downtime as possible in order to guarantee the continuity of business processes. How can you make sure that the downtime is kept to a minimum? The answer is pretty simple: reduce the risks and make sure you have a plan B.

Considerations
It’s important to consider placing servers in premise (on location) or in a data center. You have to decide if you want to pay monthly costs for colocation in a data center or to save costs while bringing your servers to your own location. But don’t forget to ask the important questions such as: If your servers are down, how much revenue would you lose? What risks are you willing to take? And how much can will that cost? It would be very handy to calculate how much time and money can be saved by preventing downtime. You might need to do investments, but these investments can often be earned back with time.

Data Center vs. On Premise
Critical business applications must be available 24/7. Companies and its customerbase must be able to rely on the availability of data centers for the availability of services and applications at all times. You can choose to install your equipment in your own company or in a data center. A data center might be an investment, but it also brings in many benefits.

Facilities in a data center
A data center is more than a building with servers. Thanks to various security measurements such as: climate control, constant power supply and fire safety measures, there is always an stable and conditioned environment for the hardware. In most data centers these measurements are redundant so that in any case a broken switch can be switched from power path. This means that the servers will never run out of power. Various emergency power solutions have also been installed so that the servers continue to run in the event of a complete power outage. When choosing a data center, it is very important to be able to choose redundant facilities.

The costs of installing a redundant power supply are high that it is almost impossible to achieve this in a regular business building. An emergency unit is also not immediately on site in the event of a power failure. This means that a power failure will shut down the servers with major consequences.

A data center also uses various advanced cooling systems and fire safety measurements that cannot be realized in the same way in a office building. A calamity fire, data centers with special equipment such as gas extinguishing, are being extinguished so that the servers remain intact. Servers at the office will not come out of a fire extinguisher undamaged.

Connectivity
In addition to the above-mentioned security measurements to power supply and cooling, one must also consider the internet connections. These can also go down and most data centers have connections to different Internet Exchanges. If there is a malfunction on the AMS-IX, all providers connected to it will go offline. If a company always wants to be online, the internet connection must also be connected redundantly.

A data center must be chosen where multiple Internet Exchanges can be connected, such as the DE-CIX, NDIX, NL-IX and the R-IX. If a calamity occurs, the data traffic can be routed throughout alternative paths. For example, on 13 May 2015 there was an incident on the AMS-IX, with major consequences. During maintenance work, an error occured by an employee, causing internal routers and switches to become overloaded and certain websites not accessible. This type of malfunction can cost companies millions of euros.

100% uptime doesn’t exist
Uptime stands for the amount of time that a service is available and operational online. Is it possible to guarantee 100% uptime? In theory this should be possible, but this is not achievable. A guarantee of 100% uptime will create conflict between technical issues and the expectations of customers.

No organization can provide 100% uptime and that’s including data centers. Data centers depend on external parties when it comes to providing internet, infrastructure, network and power facilities. These facilities are implemented redundantly within a data center, but nobody has an impact on external factors. Therefore it’s important to think about the implications of downtime and how to minimize this.

Uptime and SLA
A Service Level Agreement (SLA) is a contract between a hosting provider and its customer. This document defines the solutions offered by the hosting provider with what requirements. It also states what the consequences are for the service provider if the agreements from the SLA are not complied. It’s very important to check which uptime is guaranteed in the SLA. 

Uptime in a SLA is measured with ninesIf there are more ninesthe uptime is much better. The nines also determine how much downtime there will in a period of time. 

What do the nines mean? 

Uptime SLA Table 

Downtime/ Year  Downtime/ Month  Downtime/ Week  Downtime/ Day  Uptime 
3.65 days  7.20 hours  1.68 hours  14.4 minutes  99% (“two nines”) 
8.76 hours  43.8 minutes  10.1 minutes  1.44 minutes  99.9% (“three nines”) 
52.56 minutes  4.38 minutes  1.01 minutes  8.64 seconds  99.99% (“four nines”) 
5.26 minutes  25.9 seconds  6.05 seconds  864.3 ms  99.999% (“five nines”) 
31.5 seconds  2.59 seconds  604.8 ms  86.4 ms  99.9999% (“six nines”) 
3.15 seconds  262.97 ms  60.48 ms  8.64 ms  99.99999% (“seven nines”) 
315.569 ms  26.297 ms  6.048 ms  0.864 ms  99.999999% (“eight nines”) 
31.5569 ms  2.6297 ms  0.6048 ms  0.0864 ms  99.9999999% (“nine nines”) 

(https://www.nefiber.com/blog/five-nines-uptime-sla-mean/) 

In e-commerce a uptime of 99,999% is recommended as a minimum for the systems. This means less than 6 minutes downtime a yearIf a data center has more redunancythere will be more nines behind the dot. More nines equals a lower downtime. 

Make sure you have a plan B
In addition to redundancy, having a extra location can provide extra security. Many data centers currently cluster around Amsterdam. As a result, many companies depend on a relatively small area. A disaster in this region can have major economic consequences. To prevent damage in the event of an emergency, large companies often put their data storage away in several data centers in different regions. This serves as a base for always having a backup.

Everyone in the IT industry will agree that downtime has a bad impact on business operations. It causes financial damage and must be contested with the use of the correct available resources. This can involve redundant facilities within a data center, SLAs, or possibly by using different data centers.

As conclusion, it’s important for companies to first make a thorough evaluation process of failure risks with an estimate of how much it costs per hour (or minute) in the case of downtime. After this evaluation it’s good to make a plan to be stand-by in a case of downtime.