Bitcoind with Mempool.space/Electrum Server

[OmegaStarScream]

I want to run Bitcoind with memepool space on the same machine. Does any one here know which providers should I seek and how much would it cost to do so? Not an overkill, but not something that couldn't handle hundreds of requests.

And also I would like to know how much better is it to run Electrum server as opposed to Memepool space?

This is mainly to make requests to fetch balances and possibly transaction history.

[nc50lc]

Loyce's thread in Electrum board may have a rough estimate of it.
In pages 5 to 6, there's a discussion between Loyce and Direwolf of the possible VPS cost of running a Bitcoin Core and Fulcrum instance.
Mempool space was mentioned that it uses low resources as well.

Here's the link: /index.php?topic=5554840.80

[DireWolfM14]

The kvm server I'm using (and I believe LoyceV as well) is hosted by ServaRica.  It hasn't been 100% problem free, but the issues I've had are minor and quick to resolve themselves.  I'm running a publicly accessible and announced SPV server using Fulcrum as the engine.

I've never ran the production version of mempool.space, but I've ran my own personal use version for years.  It's possible to run mempool.space with only bitcoind as the backend, but it runs faster with both bitcoind and an SPV server to draw transactions from.  From what I understand the production version runs best with blockstream electrs as the SPV server, but I don't know if the other common SPV servers would provide much less performance.  I've used ElectrumX and Romanz Electrs and had good results with both.

I'm sure more information could be had by asking the mempool.space development team.  They tend to be pretty active in the Matrix chat room:
https://matrix.to/#/#mempool.support:bitcoin.kyoto

When I have some time I'll be setting up the personal version in a docker container on the same server I'm hosting Fulcrum.

[ABCbits]

Not an overkill, but not something that couldn't handle hundreds of requests.

per second? per minute?

And also I would like to know how much better is it to run Electrum server as opposed to Memepool space?

One (or maybe more) mempool.space feature require Electrum server.

This step is optional. You can run Mempool without configuring an Electrum Server for it, but address lookups will be disabled.

And just like others, i recommend choosing Fulcrum due to it's performance.

For this reason, Fulcrum emerges as a clear winner in this benchmark. Although it takes slightly longer to index than Electrs, once that indexing is complete the benefits of a scalable and performant server will greatly outweigh this once-off cost, not only in query speed and reliability but in prolonging hardware lifespan.

[OmegaStarScream]

-snip-

Which one did you end up using? I'm guessing the Slim Slice 40 or 48? https://clients.servarica.com/store/v3-kvm-slices

-snip-

-snip-

Per second. I was hoping something in the range of 500req/s if possible? On Mempool's FAQ page, they do mention the different implementations that one could go with:

Mempool uses an Electrum server to do address lookups. There are several implementations of the Electrum server protocol, and Mempool can use any of them, but the implementation you use affects performance:

- romanz/electrs. This is a common choice for its low resource requirements, and most full-node distros use it. But while this implementation works great for basic queries, it will struggle with heavier ones (e.g. looking up addresses with many transactions)—especially when running on low-power hardware like a Raspberry Pi.

- Fulcrum. Fulcrum requires more resources than romanz/electrs but it can still run on a Raspberry Pi, and it handles heavy queries much more efficiently. If you're having issues with romanz/electrs, Fulcrum is worth a try.

- mempool/electrs. If you have stronger hardware, you could consider running mempool/electrs, the backend that powers mempool.space. It's a fork of Blockstream's Esplora, which is in turn a fork of romanz/electrs, intended for maximum performance and larger-scale deployments.

So I guess the choice would be between Fulcrum and memepool's electrs but it's still unclear to me what "larger scale" deployement means in terms of requests per second here...

[DireWolfM14]

-snip-

Which one did you end up using? I'm guessing the Slim Slice 40 or 48? https://clients.servarica.com/store/v3-kvm-slices

No, I went with the V2 FAT Slice 8.  I was more concerned with disk space than the quantity of cores and ram.  Anything more than 8 cores and 32GB of ram is really overkill for my purposes, but if you're looking to host a production environment with lots of traffic, then it may not be enough for you.

I don't know what the target requirements would be for a production mempool.space instance.  So you don't end up costing yourself more money than you need to, and so you don't end up not meeting the requirements you really should ask the development team for their recommendation.

[ABCbits]

Per second. I was hoping something in the range of 500req/s if possible?

Fulcrum may be good enough, if i understand benchmark done by Sparrow Wallet team correctly. Using Raspberry Pi 4 8GB, it can fetch data for wallet with 3K used address with time from 1.4s to 14.1s depending on requested data. Make sure to read their full benchmark though to have better understanding.

So I guess the choice would be between Fulcrum and memepool's electrs but it's still unclear to me what "larger scale" deployement means in terms of requests per second here...

You'll need lots of fast storage if you decide to use mempool's electrs.

The indexes require 1.3TB of storage after running compaction (as of October 2023), but you'll need to have free space of about double that available during the index compaction process. Creating the indexes should take a few hours on a beefy machine with high speed NVMe SSD(s).

For personal or low-volume use, you may set --lightmode to reduce disk storage requirements by roughly 50% at the cost of slower and more expensive lookups.

I don't know the actual number these days, but i wouldn't be surprised it require 2TB or more in March 2026.

[NotATether]

memepool space

 

(Sorry, I just had to point it out)

Your best bet is using a Hetzner reseller. I have a great experience with a few resellers, I can PM them to you if you want.

It's getting harder and harder to find a decent VPS that isn't way overpriced, so I would suggest avoiding VPSes completely.

You'll need lots of fast storage if you decide to use mempool's electrs.

NVMe should be good enough, maybe half a terabyte. Nothing that storage servers can't already provide readily.

[OmegaStarScream]

 

(Sorry, I just had to point it out)

Your best bet is using a Hetzner reseller. I have a great experience with a few resellers, I can PM them to you if you want.

It's getting harder and harder to find a decent VPS that isn't way overpriced, so I would suggest avoiding VPSes completely.

 

Yes, please do. In addition to what have been suggested above, I also found this (not going to be using it for a lightning node): https://contabo.com/en/bitcoin/

But I'm just not sure how good they are.

Fulcrum may be good enough, if i understand benchmark done by Sparrow Wallet team correctly. Using Raspberry Pi 4 8GB, it can fetch data for wallet with 3K used address with time from 1.4s to 14.1s depending on requested data. Make sure to read their full benchmark though to have better understanding.

Thank you! I'll look into it.