@poga 可以说是非常吃资源（由于 full mesh 的网络结构和他的 consistency 模型）。不过 Synapse 吃的资源是和房间复杂度成正比，所以如果大房间数量一定的话，有更多用户反而不一定会吃更多资源 —— 因此用户少的时候反而会看起来更吃资源。
People: radiation is bad, 5G will kill us all
Also people: *puts on literal radioactive underwear*
@loosy Mastodon is probably less efficient than Pleroma, but at least it scales linearly and is not a total database hog. With libjemalloc nowadays its base resource usage is also much lower than before. But Synapse.... to be fair, it's also better than before, and Matrix's full mesh & consistency model does require a lot of work, but it is still quite heavy.
@loosy No, my Mastodon instance only eats up <1GB of memory, and it doesn't need a whole lot of database performance (so less cache needed). The real problem is Matrix, in particular, Synapse -- it eats up about 2 GB to begin with and grows rapidly when you join any sort of large rooms, and it hates the database and needs a bunch of cache to work.
Of course our Mastodon or Matrix server is not running slow right now. It's just that we barely have any headroom and I think I should keep an eye on Hetzner auctions for a good server deal with bigger ECC memory
Unfortunately Hetzner auction servers do not normally come with ECC.
Android (LineageOS) Developer / Linux User / Otaku / Webmaster @ angry.im
The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!