Nostr Web Client

That's interesting do you have some stats on that?

It would be interesting to compare the volume that goes through REST and the volume that goes through websockets, and also number of users serviced

And also compare the resource needed

Scaling to 1 billion users is still a big challenge for nostr, given that it's mainly grass roots

⚡️🌱🌙 3y ago

No unfortunately I don’t have data I can share, but on big market platforms the clients want to know market liquidity every 5ms. High Frequency Trading demands information about the liquidity (not transactions) of every consolidated limited order book (every ticker) at a ferocious rate, in order to execute arbitrage.

Every HFT client wants every limit order of every ticker available on every trading terminal, every 5ms. People pay a lot of money for these feeds and the feeds are all built on websockets.

Pro Tip: clients don’t poll the exchange. The solution to the traffic and scalability problem lies in the choice of database framework.

Financial markets are orders of magnitude more traffic than all of social media, equivalent to millions of Tweets/second. These are the people renting all the antenna slots at data centres and dropping $20m on superconducting lasers to get insane bandwidth at insane latency.

I might be able to dig something up to share next week.

Reply to this note

Please Login to reply.

Discussion

chowcollection 3y ago

I’m pretty sure these were the kinds of stories that finally got me to look at Bitcoin:

https://youtu.be/WyPPaiZEYO4

cody 3y ago

HFT's current stage is a point on a really interesting through-line of how systems align themselves to push information exchange to its limits. We've gone from shouting hawkers bartering in open-air markets to skyscrapers filled with floors of traders shouting into phones to 5 quants doing math in a basement server room.

Melvin Carvalho 3y ago

Got it. Im still unsure it compares to the whole REST surface

A friend of mine does market monitoring for the biggest exchange in europe. It's a few terabytes a month

Archive.org which is one site has 10,000 terabytes

REST has a proven track record to scale to billions of people with low resource intensity. Websockets are a newer technology, let's see how they get on.

What's interesting is most older web systems started with HTTP then added websockets for realtime. Nostr starts with realtime, and we can explore how to scale that with supporting NIPS.

Fun times! :)

⚡️🌱🌙 3y ago

The exchange is a few terabytes a month, because it is architected correctly.

You don’t have every user constantly polling the entire exchange for a refresh, that simply doesn’t scale. Instead they broadcast change to clients.

It’s a feed, not a query. Again the solution is the choice of DB engine. Does that make sense?