You are viewing a single comment's thread from:
RE: The Steem Fork Begins: Meet Steemit Reloaded, New official name: Calibrae
The full RPC witness nodes run with 32GB RAM and uses an SSD RAID array for swap according to @gtg. Is that the magic sauce you were after! ;)
Doesn't seem like a great idea to use SSD like that, but I don't know enough about the MTTF for drives these days - maybe it's not too bad. Also I read about 'NVM Express interface', which make this more reasonable, and since most of the access is reading and not writing, perhaps this is scalable to a greater degree than we thought, you can get 2TB SSD drives now.
Seemingly the reliability is a problem with this approach though, perhaps as the memory is expected to be faster than SSD speeds?
HAHA. SSD raid array? how many? what kind?
Yes, as I said, I have a samsung NVMe drive here. It was slow as shit as well.
I have been saying this right from the beginning: it seems like steemd only is happy in ram. I have a motherboard in one of my miners with two M.2 slots. if I swapped that for my current board and got a second Samsung 256gb...
but if you are following all this, do you realise how fucking expensive this is?
and how much of a LIAR @gtg is.
He said 3 drives and RAID 1, but not sure whether he mean just 2 for the RAID, or 3.
To be absolutely fair, I haven't seen evidence of him lying directly, but there does seem a tendency to obscure the details somewhat ;)
It's not cheap to get remote facilities like these is it? I won't be attempting it on a standard VPS any time soon!
It seems obvious to me that he's living in a university on in a country like this, where throughout the metro area you can get gigabit optical for cheap.
Yeah, maybe. I thought Bulgaria wasn't too bad, but maybe it varies a lot. What do you think the minimum practical bandwidth is? I can't seem to find that anywhere.
I'd guess somewhere around 20mbit.
The issue is mainly about latency, in fact. We know max latency is 3000ms. Min practical is 1500ms. Probably ideal is about 200ms. An easy way to determine this, is to go look at the seed-node list, and get an average ping of each one. The outer bound will tell you a lot about the latency relative to your location, as to the average, that will be the lower and upper bound.
Right, trying a few shows the European ones are <20ms and US <200ms. I have 20mbit, not sure I can afford to tie it all up though!
pretty much as expected, US/EU is the regions where witnesses are.