# LN Project evaluation >How does the operation of the Lightning Network change if the selected node goes offline for 2 days? Yes, this kind of test could be good to measure the centralization of the network! We can just simulate it and we can not test it in the real life because require stopping big node operator. lnmetrics can do it just for luck event that happens on mainet. > Simulate how the network operates This could be very good because it is difficult today to simulate some operations, there is some attempt to do it with some projects https://github.com/bottlepay/pathfinding-benchmark that shows how it is difficult to integrate different nodes >How would this event impact transaction fees during the outage? Which nodes would be most affected? How many failed transactions might occur during this outage? Mh regarding "transaction" (are payment) failure affected I do not know if this will be real because today there are a lot of payment failures due to some probing, a real example can be my node https://bruce.lnmetrics.info/metrics imho the simulation tool can be good only to try to estimate how the routing algorithm will work, but we can not estimate the failure because we can not have without cheating a lot on the model how many payments will not fail without an outage. >And lastly, how would all these statistics be different if that outage was 30 minutes instead of an hour? What about a day instead of an hour? this is the point to have lnmetrics that work in the mainet, the statistic on fake networks are just for fake networks, on mainet you have problem like network latency, connection with node over tor, the real balance of the channels for a node. >All these things could be simulated and then analyzed. Charts and reports could be generated so that the user can make decisions based on this data. I like the idea to have a simulator to simulate different situation and I think it is super useful if this simulator support different node because with the current status is painful to put this configuration together, but I'm also worried that the statistic will also be taken by the user as true and not just that need to be confirmed on mainet.