If ever a week has felt like a marathon race filled with interval training, it is this one. Let's start with some release news: In preparation for the release of Elemental Chat for Hosts, we've pushed the latest upgrade for HPOS - the HoloPort operating system out to alpha test production. Hosts can log into the Holo Forum to read details about how to prepare their HoloPorts.We've run live and automated testing repeatedly this week for Elemental Chat. We're working to have our scaling tests accurate and especially reflective of what the use patterns will be when we open it up to hosts. The results have continued to be quite good and users report better and better experiences in the live tests. According to one tester, Friday's live test was the "fastest yet experienced" and others said the back and forth chat experience feels like your typical instant message chat. That's pretty amazing for a proof of concept P2P chat app.Tonight, we have run load tests that we believe model what the experience will be for when we release Elemental Chat to hosts. The trick in setting up the test is that we have to consider the baseline activity of nodes gossiping on the network, and then model over top that -- the likely peak messaging activity between hosts.In a recent test with 440 ports on the network, and 150 of them active, 135 messages were sent in the space of a minute. What isn't obvious about this scenario is that sending 135 messages in Elemental Chat is really the equivalent of sending 20,115 P2P signals. Because with Elemental chat, we don't want the messages to simply post to the DHT and get gossiped out to nodes eventually, we want the nodes that are active to receive the messages really fast so that they get a real-time chat experience. Tonight the test results are showing that of those 135 messages / 20,115 direct to peer signals sent in the space of one minute: - 92.3% are arriving within 30 seconds - 98.6% are arriving within 60 seconds - 99.9% have been been received by peers within 3 minutes. - 100% of the messages arrive shortly after that The live testing demonstrated that the messages arriving under 30 seconds tend to be rather instantaneous.We are still adding in the aggregation of latency logs to assess how quickly the remaining nodes on the network who have not been active recently receive the messages via gossip. What is quite promising is that as we increased nodes on the network, the it took to begin the test increased but overall speed of signals does not appear to degrade, in fact we've seen small improvements.In parallel, in the later part of this week, the Holochain team has continued to prioritise a few improvements to networking, especially pertaining to the way that the proxy server is tuned for the receiving side of the pipeline for concurrent processing. We've not finalised or done integration testing but these changes are extremely promising for immediate optimisation.What's great is that during the testing, we've learned more how individual ports with high latency effect the network and after our upcoming release to hosts we'll continue to optimise and improve -- throughout the alpha test period. Further Holochain networking optimisations will also come after the dB refactor and sharding. The tweaks to networking this week are merely the first small step towards the full scalability promises of Holochain.