Bitcoin Forum

I've spent the last 6 months working on a new open source block chain explorer/browser written in Go: https://btcplex.com.
It's an early release (still not production ready), but I would love to hear any feedback!

Some features:

• Browse blocks and transactions.
• Check address balance history.
• Search for blocks, transactions, and addresses.

It also provides a full-featured JSON API, a plain-text query API and few server-sent events endpoint to access data pragmatically.

GitHub repository: https://github.com/tsileo/btcplex
Demo website: https://btcplex.com
Hosted documentation: http://docs.btcplex.com
Donation address: 19gzwTuuZDec8JZEddQUZH9kwzqkBfFtDa

Let me know if you have any suggestion/idea/question !

Hey,

Nice Work !

Is there are reason you are storing block json again in leveldb ? Shouldn't just maintaining the indexes suffice ? You can always get the raw block from bitcoind

I would say you need to do some optimization, waiting a week for the database to be built seems unreasonable.

@btc_enigma  thanks! I'm storing block data in SSDB (which is backed by LevelDB internally) for performance, by design the webapp never call directly bitcoind.

@9inety7even, I spent a lot of time working on optimization, it takes one week on a small server (1.2Ghz dual core/6GB RAM).
On a more high-end server (i5/16GB RAM, the server btcplex.com is currently hosted), it took 48hours, and I store a lot of data (since disk is cheap, the database is 100+GB), but request are fast, even for querying the balance of an address with thousands of transactions.

What seems to be the bottleneck? Do you think you could run it in parallel, or is IO speed the problem?

It's pretty neat and functional. Me like.

@9inety7even You're right, the real bottleneck seems to be IO, I think running BTCplex on a server with a SSD should really makes a difference.

I already run everything I can in parallel, but I must parse blocks and transactions chronologically since I need to fetch data about previous TxOuts.
I think 2 days to load the entire block chain is not that bad (when comparing the time it take to start a new bitcoind node), and given the fact that BTCplex keeps track of every known addresses balance.

@SomeoneInNeed Thanks!

Hi,

Can you throw some insights/figures on how this is faster than directly calling bitcoind. This is because I am also writing a similar block explorer and I am thinking about design issues. I was trying to  avoid exporting out the whole blocks seeing the huge amount of time its taking.   For example , I did


This completed in 0.013 sec which gives me around 71 requests/second . As far as I know, even bitcoind  would just read its leveldb database stored in ~/.bitcoin and give you the output. So it should give similar performance. Actually I see that c++ code is able to read leveldb faster than python.

Yeah I would definitely grab a DO vps and see how much of a difference it makes. Chances are you'll see fairly huge speedups.

@btc_enigma Calling bitcoind on itself is fast, but when you add you own server call:

bitcoind <-> your app <-> client
your app <-> client

Even if you can perform multiple RPC calls at once to prevent HTTP round-trip time, I prefer maintaining my own database, when I request a block, I'll want to also fetch every related transactions (sometimes 1000+ txs), that's why BTCplex save the block with every belonging transactions in a single key, so I think it's faster in my use case, but depending on what you want to accomplish, you may achieve better performance calling bitcoind RPC API rather than maintaining your own database.

@9inety7even I would love to but sadly I can't afford a 160USD/month server (database is 120+GB), maybe I'll give it a try if I get enough donation.

You could buy it for a few hours just to test the speed; I think the lowest of the high-end plans is like 30 cents an hour.

Bump!