Skip to content

Latest commit

 

History

History
98 lines (71 loc) · 3.27 KB

README.md

File metadata and controls

98 lines (71 loc) · 3.27 KB
Raw

TODO: some details might need updates; until then see ../run.sh for intended usage

analyze historical blocks

In this case, the "Tobalaba" chain of the energywebfoundation.

prep

All happens in virtualenv (see ../quorum.md#virtualenv).

Start virtualenv, install dependencies:

source env/bin/activate
pip3 install web3 pandas jupyter ipykernel matplotlib 
ipython kernel install --user --name="Python.3.py3eth"

step 0: sync

Sync the chain. Possibly with archive flag on:

cd energywebfoundation_energyweb-client
./target/release/parity --pruning=archive --geth --chain tobalaba --rpcapi "web3,eth,personal" --db-compaction=ssd --cache-size=2048

It took about ~3 hours for syncing the >3 million transactions in the >4 million blocks of the "Tobalaba" chain.

step 1: create DB

Parity is slow to query.

So we first extract all block information data, and dump it into an SQLite database.

python3 blocksDB_create.py

TODO: Is the parity RPC server really single-threaded? Would perhaps the IPC connection be faster? But there is no ~/.local/share/io.ethereum.parity/jsonrpc.ipc. How to enable IPC in Tobalaba-parity?

The 4.4 million blocks took 2.5 hours to extract:

4429200 blocks took 9431.95 seconds
execute & commit 4429200 SQL statements into DB took 40.71 seconds
TABLE blocks has 4429200 rows
MIN(blocknumber), MAX(blocknumber) = [(0, 4429199)]

But now we have them in our sqlite3 database, blazing fast...

step 2: analyze & visualize

Python Jupyter is a nice graphics enabled IDE to show tables, diagrams, etc inline that are created with Python pandas, numpy, matplotlib.

jupyter notebook --ip=127.0.0.1

then execute all cells in blocksDB_analyze.ipynb

Actually, the gitlab rendering of that ^ file is not bad, so (even though it blows up the filesize), I kept the results, incl all tables and diagrams, in that file.
(Scroll down to approx the middle of the file, to see results).

results

blocknumber_max               4392279
blocksize_max                 118,576 bytes
txcount_max                   1,179 transactions in block 1,210,825
txcount_sum                   3,313,470 transactions in total
txcount_average               0.754 transactions per block
blocks_nonempty_count         1,951,767
average txcount per nonempty blocks =  1.698

That biggest block is block 1,210,825 - see block explorer.

TPS = transactions per second:

img/TPS_allBlocks.png

For many more such diagrams, see blocksDB_analyze.ipynb.
SCROLL DOWN TO THE MIDDLE OF THAT FILE.

some images are also in img/

notes

are the transactions actually executed successfully (or are they failing)

open a JSRE console to one of the nodes:

geth_quorum attach http://localhost:22001

then query example transaction (here tx 11 in block 55):

blockheight=55; index=10; 
txid=eth.getBlock(blockheight)["transactions"][index]; 
console.log(eth.getTransactionReceipt(txid)["gasUsed"], eth.getTransaction(txid)["gas"])

26691 90000

If gasUsed != gas then the transaction got executed.