Unlike many other websites on the Internet, the Foundation runs its own servers and edge distributions to distribute our projects to the world. We do this for several reasons, but our strong stand on privacy is important. By avoiding third-party services, we maintain the technical and policy controls necessary to implement the strict user data handling that we require. Secrecy is necessary for Wikimedia’s vision to share all in the sum of all human knowledge.
Prior to this launch, our existing server network offered (and still has!)
User-facing services from four locations. Three are in the United States, located in San Francisco, California; Carolon, Texas; And Ashburn, Virginia, while the fourth is in the Netherlands. While our site locations are all well connected, large parts of the world may face significant delays when accessing Wikimedia projects based solely on their distance from our sites.
Jakarta is about 14,000 kilometers from San Francisco. Bangalore is about 8,000 kilometers from the Netherlands. Although light travels incredibly quickly (300,000 kilometers every second), loading a web page requires multiple trips back and forth from server to client,
This new torrent site in Singapore fills the large gaps between the Netherlands and San Francisco, providing lower latency and improved user experience in the region.
Under the map, public domain. Data center indicator added by Wikimedia Foundation.
Returns one of several different IP addresses to suit different edge sites, depending on the user’s location. The closer the edge site is to the user, the less latency the user will have for a secure connection to Wikimedia, resulting in faster page load times.
Shore sites also host content caches that keep copies of our most popular URLs close to the user. The content cache on the edge must be saved by other caches in the US, and in turn saved by our actual application layers and databases. Each step in this process is slower than the previous, so we try to have more and more content in Edge Cache servers as “hot” that users access first.
We started experimentally staying with
user service from Singapore from 22 March, and we are slowly expanding into the region as long as we validate the performance of the site. The process of optimizing our regional network and bringing in the last few countries will continue for the next few weeks to months, but at the moment most of Asia and Oceania are being served by the Singapore site and this is about 17% of our global operations. Used to be. transportation.
We have seen some significant improvements in page load times in the region. In that context, preliminary performance results suggest that this has been a significant advance.
On 22 March, we started directing traffic to the new site from Singapore itself.
It took several hours for the cash hit rates to increase, as most of the traffic requested was never delivered to that location before. Once we did this, we found that the time required to deliver a page was reduced by more than 40 percent. Prior to the change, the median page load was taken between 1.5 seconds and 1.7 seconds. After the change, the average page load occurred between 0.8 and 0.95 seconds. (In the image below, the arrow indicates the point at which traffic was directed to the new data center.
Of course, Singapore is the easiest possible case. Every customer is within a few kilometers of the actual data center. However, we are pleased to see similar results in other countries. Some examples are given below.
On 26 March, we enabled traffic from Indonesia, and as a result we saw a reduction in page load time from about 3 seconds to 2 seconds.
On 28 March, we enabled traffic from Bangladesh, and as a result saw a reduction in medial load time from 1.8 to 2.1 seconds between 3.0 and 3.2 seconds. (This graph is very spikier because the rate of traffic from Bangladesh is very low, and as a result the set of performance samples is very small.