Syncing a mainnet eth node takes a while

syncing a mainnet eth node takes a while.
Thinking of buying a server-grade machine with about 1 tb of RAM and sync it using a RAM-based filesystem
Would it speed up the process significantly?

Attached: 1531325141096.png (400x560, 304K)

Other urls found in this thread:

reddit.com/r/ethereum/comments/bmlp2u/nethermind_099_45gb_mainnet_fast_sync_in_5_hours/
github.com/martingeorg/tmpfs-mysql
twitter.com/NSFWRedditImage

Order a fully synced ETH node on a 2TB SSD from Vitalik. I think you only have to pay $200 for the SSD and a $25 processing fee

That is a nice option, thank you. Nevertheless, still thinking if the ram-fs approach would work well

Synching is mostly CPU and internet-bandwidth limited unless you're colocated with 100 mbit or more.

Also RAM FS would mean you need to sync again after every power outage. I wouldn't do it.

Given the whole process is write intensive, it promises a large cut in syncing time - in theory. After the sync, the plan is to write it into durable media.
100 mbit connection is doable

The writing happens in the background and is not the bottleneck if you have a decent OS and filesystem. Even older cheap HDD can easily sustain 50-100 MB/s which is a factor 10 faster than your internet connection.

The OS will use the RAM for caching writes anyway, fucking around with it manually is not going to help as much as you seem to think.

Interesting, thanks

>wasting tons of money on 1 TB of RAM you'll never use just to make a one-time process faster

>waiting weeks for a fresh sync

you'll get no more than a 5% speed up from it, which you'll waste later by having to copy everything from RAM to the HDD, plus all the headaches from setting it up in a way that hasn't been necessary or helpful since the 90's

Why do you need a full ETH node in such a rush that you'd spend a premium of several hundred dollars to over a grand just to get it more quickly? Also, you're not making the CPU crunch numbers any faster, just delivering data to it faster, and it would bottleneck. You're better off spending that money on SSD storage you literal brainlet.

fast sync takes few hours, full sync over a day
you can try nethermind
reddit.com/r/ethereum/comments/bmlp2u/nethermind_099_45gb_mainnet_fast_sync_in_5_hours/

what is the advantage or gain in running an eth node? complete noob honestly asking

>Synching is mostly CPU and internet-bandwidth limited
no, syncing on eth is limited by random io latency.
That's why it's impossible on hdd.

seems to work for mysql
>github.com/martingeorg/tmpfs-mysql

This

verifying transactions

A full sync takes like 2 days? Why the rush

right so SSD then, still cheaper than RAM

Given geth backend is leveldb, why won't the devs add support for all the databases that satisfy the LevelUp interface? It's very similar in nature to what LevelDb interface is by design.
This would allow one to host the Eth node backend on a cluster-based database making shit fast and fail-proof (on a single-node level)

ssd is enough, yes, but latency is 1000x worse than ram.
Validation would be actually cpu bound if everything was in ram