For every product, Google logs the number of requests it receives in a given time period, along with a guess at the geographic region where the request came from. For each geographic region and product, we create a graph that gives a representation of the ratio of that region's request rate to the worldwide request rate.
For each time period and product, we divide the total worldwide traffic by the traffic for each geographic region, giving a number between 0 and 1. Then we multiply all numbers by a constant, which normalizes but does not change the shape of the graphs.
So the numbers on the Y-axis do not have any meaning except in relation to other numbers in the same graph. In other words, a 20 is not 20 of anything—but it is twice as big as 10. Numbers between different graphs cannot be compared.
Again, keep in mind that what is being graphed is a geographic region's portion of Google's worldwide traffic. If, over a long time, a selected geographic region's graph is decreasing, that doesn't necessarily mean that the amount of traffic we receive from that geographic region is decreasing. It just means that the selected geographic region's traffic is growing slower than the worldwide average.
The graphs show product use over time, so normally you will see a similar pattern every week. For example, as you see in this graph, you'll notice that on weekdays there's usually more traffic than on weekends to Gmail from France. What's interesting is when there's disruption to the normal graph pattern—for example, as you can see in this graph, Google services were largely inaccessible in Egypt when the government halted Internet traffic in January 2011 during a period of political unrest.
We believe that this raw data will give people insight into whether or not our services are accessible in a given geographic region at a given time. Historically, information like this has not been broadly available. We hope this tool will be helpful in studies about service outages and disruptions and that other companies will make similar disclosures.
We add annotations based on the anomalies that we see in the graphs. Since the annotations are implemented manually, they are not updated in real-time.
When a product is annotated as "partially accessible" it refers to the state of the product for users in a given geographic region. For example, if YouTube is "partially accessible" it means that YouTube is only accessible for some users or that traffic to some videos is being disrupted.
Interruptions in our services can have several different causes, ranging from network outages to government-mandated blocks. When the service is inaccessible for an extended period of time (beyond what is standard for a network outage), we investigate and draw conclusions based on the number of users affected and information we receive from local ISPs.
Due to user privacy concerns, some geographic regions' product pairs are not available at this time.
The Traffic graphs show the same information as the Mainland China Service Availability page, but with more granularity and historical context. Also, we're able to annotate these graphs, to provide more granular information about outages.
Users can call +16504194196 or +390662207294 or +442033184514 to leave a voicemail on @speak2tweet. The service will instantly tweet the message, and also add a hashtag indicating which region the message came from if the phone number can be identified. No Internet connection is required. Users can also listen to the messages by dialing the same phone numbers or going to http://twitter.com/speak2tweet.
Transparency Report