Is website downtime a barrier to open data?
Open Data Watch tested 180 websites to find out.
by Caleb Rudow
with the ODW Team
Open Data Watch grapples with website downtime every year when assessors evaluate the openness and coverage of the 180 national statistical office (NSO) websites included in the Open Data Inventory (ODIN). Assessors must often return to an NSO site to find a window when the site is online to perform the assessment. And some countries were not evaluated due to a long or constant period of downtime during the assessment. What is just a frustration for assessors is a more serious problem for data users who can’t access critical datasets when these sites are offline.
Estimates from Gartner show that website downtimes can cost an average of $5,600 a minute and up to $300,000 an hour. Though access to data on NSO websites is not valued in the same way as industry website outages, downtime can still have significant impacts on trust in the NSO, not to mention access to data.
How do NSOs compare to businesses which often take elaborate measures to ensure constant uptime (or time a website is online)? And, what lessons can they learn from industry to improve their uptime, site usability, and search engine optimization so that more people can access and use their data?
To investigate these questions, ODW tested the uptime for the NSO websites it assessed in ODIN for nine months from July 2018 to March 2019. Our results show that open data for official statistics has a long way to go to catch up with industry.
There is not a clear industry standard for uptime of a website. High availability – or an uptime percentage of 99 percent or more – is one measure. (This may sound like a high standard, but even a site with 99 percent uptime is down for 87 hours and 36 minutes a year.) And while NSO websites don’t face the same loss of profit as businesses do, downtime can cause losses in productivity when users can’t find their data; loss of resources as IT administrators struggle to get the system back online and a loss of faith in the NSO as a reliable data source.
Although ODIN assessors are instructed to check back frequently on an unavailable NSO website to see if it is online, how many data users would consistently return to a website after they have checked a few times and found it down? And, more importantly, what use are open data if they are on a site that is frequently offline? Access comes before openness.
To evaluate the average uptime for NSO websites, @OpenDataWatch used a website monitoring service to ping or check NSO servers every 30 minutes to see if their websites were online.
– https://bit.ly/2VJFvZn
To evaluate the average uptime for NSO websites, ODW used a website monitoring service to ping or check NSO servers every 30 minutes to see if their websites were online. These pings require a very small amount of bandwidth from the server and are typically used by businesses and developers to track their websites’ uptime. Although ODIN 2018/19 included only 178 countries – two countries were dropped due website accessibility errors and extremely high downtimes – we included all 180 countries in this study. Data were available for 171 countries out of the 180, due to technical issues with uptime checks for nine websites. Our findings show the global distribution of uptime for NSO websites and which regions are falling behind.
The majority of sites tested met the industry standard for high availability, but many countries lag behind.
Twenty six percent of countries had less than 99 percent uptime with 23 countries having less than 95 percent. While most countries are meeting or exceeding the high availability standard, a significant number are far below. Data on these sites are frequently unavailable.
| Uptime percentages | Countries |
| Greater than 99% | 126 |
| 95-99% | 22 |
| 80-95% | 16 |
| 40-80% | 4 |
| 20-40% | 3 |
Extended periods of downtime are more likely for users in certain regions.
There is a wide range of average uptimes for regions from 78.9 percent in Middle Africa to 100 percent in Australia and New Zealand. The lower uptime in the underperforming regions is especially problematic as these regions have less reliable internet access 1 . Users living there may not be able to return to a down website to check for data as the internet may be only accessed from certain locations or it may be too costly. NSO officials in these regions may also experience challenges accessing their data and performing other research tasks that require internet access.
Uptime percentages by UN sub regions show a clear correlation between income level and downtime.
Lower–income countries experience 92.2 percent uptime; high–income countries have 99.8 percent uptime. This difference, when converted into hours, means that NSO websites in high-income countries are online for 28 days a year more than those in low-income countries.
NSO websites in Low-income countries experience 92.2% uptime; high-income countries have 99.8% uptime. This means that NSO websites in high-income countries are online for 28 days a year more than those in low-income countries. – https://bit.ly/2VJFvZn via @OpenDataWatch
| Income level | Uptime |
| Lower | 92.2% |
| Lower middle | 94.2% |
| Upper middle | 96.8% |
| High | 99.8% |
NSO websites’ uptime for certain regions are lagging: How do we bring them up and what are the next steps?
Low rates of uptime may be due to human error, security, hardware- and software-related problems, interoperability and migration issues, natural disasters, and power outages. While countries need to evaluate their own unique uptime issues, one place to start is to investigate whether a server based in a cloud network would provide better service than one maintained in the country. Major cloud networks have a higher uptime and require less technical expertise to set up, making them an effective and less technically challenging option for website hosting.
The servers for a sample of countries in our dataset were geolocated, revealing whether they are hosted in–country (likely a domestic setup) or out–of–country (likely in the cloud). The results show that there is a clear correlation between the location of servers and website uptime/downtime.
NSO websites’ uptime for certain regions are lagging: How do we bring them up and what are the next steps to ensure #opendata is accessible? – https://bit.ly/2VJFvZn via @OpenDataWatch
| Server location | Average uptime |
| In–country | 95% |
| Out–of–country | 99% |
Further research could be done on server configurations and other possible causes of downtime as well as research on how NSO sites could utilize other industry standards and practices to increase the usability of their sites. To evaluate whether NSO websites are using industry best practices, ODW is working with PARIS21 on a study of NSO websites and how they are implementing the UNSD principles and guidelines for data portals, using ODW’s data site evaluation toolkit. This landscaping exercise will provide further information on the state of NSO websites and areas for improvement. Stay tuned.
Annex: Uptime by UN sub-regions
| Region | Uptime |
| Middle Africa | 78.9% |
| Central Asia | 84.9% |
| Western Africa | 92.0% |
| Southern Africa | 92.0% |
| Caribbean | 92.4% |
| Polynesia | 93.3% |
| Western Asia | 94.9% |
| Central America | 97.2% |
| South Eastern Asia | 97.9% |
| Eastern Europe | 98.0% |
| Southern Asia | 98.2% |
| Eastern Africa | 98.3% |
| Melanesia | 98.4% |
| Northern Africa | 98.8% |
| South America | 99.3% |
| North America | 99.4% |
| Southern Europe | 99.4% |
| Northern Europe | 99.9% |
| Eastern Asia | 99.9% |
| Western Europe | 99.9% |
| Micronesia | 99.9% |
| Australia and New Zealand | 100.00% |
Endnotes:
↑
1. Additional information on Internet usage by categories may be found here.
