 So, hi! We as ProBlab had a look at the Hydra boosters in the last couple of days and yeah, in this quick demo I want to show or present some of the results that we got from our measurements. So for everyone who's not that familiar with Hydra boosters, Hydra boosters attempt to cover the whole hash space in the DHT so that every time you provide something to the network, you hit one Hydra booster and so that if anyone else tries to retrieve that CID will also hit a Hydra booster and gets this content or the provider I got much faster than if it was without those Hydras. And so as I said, the Hydras attempt to cover the whole hash space and so we just, yeah, at first we wanted to verify this proposition here. The first thing that we took a look at is if there's actually a uniform distribution of hash, sorry, Hydra heads and in this graph we can see that's the case so it should be a straight line and yeah, well that's the case. And the other thing is Hydra head actually in the proximity of the 20 closest peers for every every peer we can find in the network and for that we took full network rods from Nebula then put all of those peer IDs in a binary try and calculated for each peer ID in the network the 20 closest peers and checked whether a Hydra head is actually inside the proximity of these 20 closest peers and the results show that it's actually the case so in this particular example we had around 16,200 peers in the DHT and for 15,700 there was actually a Hydra head close by which makes up more than 97% coverage of the whole of the whole hash space and so this gives us an excellent advantage into the network. Yeah, just a reminder the provider that consists of the CID TTL and the provider multi-hash and also those Hydra users have peer records so in memory and what we can do now is we can take all the provider records that the Hydras know of and correlate the providers with their multi-addresses and in turn the geolocation from the IP addresses so we can actually tell where in the world the CIDs actually reside. Since I'm short on time I think I will skip the architecture and so maybe just some general information and the Hydra users know of around one billion CIDs each day one billion unique CIDs so this is on the X axis in the days of the last week and what we can see here is if we take the set intersection between two days we see that only around 500 million CIDs actually intersect here which means that in each day around 50% of all CIDs churn and leave the network and if we assume that a CID covers around 256 kilobytes worth of data this means every day 120 terabytes leave the network but also join the network again and so this is just the CID churn graph it's just another representation of exactly that and what we can do is check which are the top providers so here we can see which peer IDs actually provide how many CIDs and if we just take a look at the top provider here this is just one peer in the network and this one peer provides around 13% of all CIDs of the whole network and this goes down this the next one is around 9% 7% and so on and so what we wanted to do now is actually find out who those peers are and for that well we I thought these are maybe gateways or large pinning services and so on and so we developed a tiny tool that's called Antares that you can see here which is just a tool that sits there it's a lippy to peer host it provides content to the network and then requests that content through a gateway or through a pinning service and then just tracks which peer ID actually requested this content and I forgot to say this content is random and so no one else should know about it and so if others request that content we can track which peer IDs belong to which services and well I'm running out of time otherwise I would have shown you that but it turns out none of the well I checked it with Infura and with Pinata none of them correspond to these large pinning services and also it's not no gateways or so on and so but I'm leaving it running maybe I will discover some of them and maybe just one last thing if we take it as I said we can correlate CIDs or provider record with provider records with peer records and then in turn with the geolocation we can have this country distribution that I told you about and we see that more than 50 percent of all CIDs can be associated with the US and then the Netherlands and France so these are also quite interesting results in my opinion and yeah we are also looking at the dependence for content retrievals and content publications and right now we are running experiments where we exclude hydras from content retrievals and content publications and just check how the how the performance differs there and yes so these will be the next steps