Glados continuously queries the Portal network for random 4444s headers, blocks, and receipts. Statistical results of these audits are displayed here and can be viewed in more granular detail on the Audit Dashboard.

4444s stats

Period New content Total audits Total audit passes Total audit failures Pass rate (%) Failure rate (%) Audits per minute
Last hour 897 2163 2161 2 99.9% 0.1% 36
Last day 21510 51781 51752 29 99.9% 0.1% 35
Last week 146628 353472 353342 130 100.0% 0.0% 35
Every 15 minutes, Glados runs a census of all reachable Portal network nodes. The results of the most recent census are displayed here, and in-depth historical data can be viewed on the Network Explorer.

Client Count

The Y-axis shows the radius as a percentage of the content keyspace. The X-axis shows the node ID of each node found during the most recent census of the network, which runs every 15 minutes. The individual clients are color coded: blue for Fluffy, purple for Trin, red for Shisui, and orange for Ultralight.
This graph shows a week's worth of success rate statistics for the various audits that Glados runs, as well as for the types of content that it audits.
This graph shows the redundancy at each given band of Portal network keyspace during the most recent network census. Each node on the network advertises a radius that describes the band of data that it is responsible for. The graph shows the number of nodes that store each part of the keyspace.

The darker color represents the nodes that are responsible for every key in that band, while the lighter color represents the nodes that are only responsible for a portion of the keys in that band. So the darker color shows the guaranteed redundancy of properly functioning nodes, while the lighter color shows the average redundancy. Three nodes that cover half the keyspace in a band would show up as an average coverage of 1.5 in the lighter color. On average, every piece of content is stored 1.5 times, but in the worst case, half the keys would be stored 3 times, and the other half 0 times. A chart that is mostly lighter color would be suspect until you split the bands into narrower slices, to see if the redundancy is evenly distributed.