Can we prove 100k transactions per second when public testnet launches?

[clout]

@bytemaster can you please get rid of the reference to nasdaq transaction processing speed. even a cursory investigation demonstrates that the information you provided is false. it's rather concerning that you are creating a securities exchange without knowing much about the industry...

Could you please provide more accurate information I can quote.

Sorry didn't see your reply. I provided this link in a previous thread so I figured you already had the information http://www.nasdaqtrader.com/Trader.aspx?id=Latencystats

[liondani]

What we will be testing with the first test network is 1000 transactions per second and 1 second block intervals.   This should require about 100KB /sec of network bandwidth and be viable on the simple VPS systems that we are using today.

Sounds promising. BTW how many transactions Ethereum can process per second?

Good question. I would love to know that too.

[xeroc]

Is the market open 24h a day? You assume messages are evenly distributed to 24 hours.

good point .. ehm .. selling point

I found this:

Code: [Select]

Pre-Market Trading Hours from 4:00 a.m. to 9:30 a.m.
Market Hours from 9:30 a.m. to 4:00 p.m.
After-Market Hours from 4:00 p.m. to 8:00 p.m.

Quote and order-entry from 4:00 a.m. to 8:00 p.m.
Quotes are open and firm from 4:00 a.m. to 8:00 p.m.

[lakerta06]

@bytemaster can you please get rid of the reference to nasdaq transaction processing speed. even a cursory investigation demonstrates that the information you provided is false. it's rather concerning that you are creating a securities exchange without knowing much about the industry...

Could you please provide more accurate information I can quote.

Press release from 2013:

"Approximately 25 billion messages pass through NASDAQ OMX's U.S. equities and options systems on an active day, which amounts to over 1 million transactions per second in a trading day. During volatile periods, peak message traffic can be double."

Here is some earlier discussion about this.

Assuming a 'message' is a transaction then 25 billion tx/ day mean 290k tx/s.

Is the market open 24h a day? You assume messages are evenly distributed to 24 hours.

[fran2k]

@bytemaster can you please get rid of the reference to nasdaq transaction processing speed. even a cursory investigation demonstrates that the information you provided is false. it's rather concerning that you are creating a securities exchange without knowing much about the industry...

Could you please provide more accurate information I can quote.

Press release from 2013:

"Approximately 25 billion messages pass through NASDAQ OMX's U.S. equities and options systems on an active day, which amounts to over 1 million transactions per second in a trading day. During volatile periods, peak message traffic can be double."

Here is some earlier discussion about this.

Assuming a 'message' is a transaction then 25 billion tx/ day mean 290k tx/s.

[Samupaha]

@bytemaster can you please get rid of the reference to nasdaq transaction processing speed. even a cursory investigation demonstrates that the information you provided is false. it's rather concerning that you are creating a securities exchange without knowing much about the industry...

Could you please provide more accurate information I can quote.

Press release from 2013:

"Approximately 25 billion messages pass through NASDAQ OMX's U.S. equities and options systems on an active day, which amounts to over 1 million transactions per second in a trading day. During volatile periods, peak message traffic can be double."

Here is some earlier discussion about this.

[bytemaster]

@bytemaster can you please get rid of the reference to nasdaq transaction processing speed. even a cursory investigation demonstrates that the information you provided is false. it's rather concerning that you are creating a securities exchange without knowing much about the industry...

Could you please provide more accurate information I can quote. 

[clout]

@bytemaster can you please get rid of the reference to nasdaq transaction processing speed. even a cursory investigation demonstrates that the information you provided is false. it's rather concerning that you are creating a securities exchange without knowing much about the industry...

[xeroc]

https://www.nginx.com/blog/thread-pools-boost-performance-9x/
 maybe this tech can strengthen our performance.

asynchronous processes are the contrary to atomic processes ..

[chono]

https://www.nginx.com/blog/thread-pools-boost-performance-9x/
 maybe this tech can strengthen our performance.

[xeroc]

So, according to that doc, and bytemaster's comments above, this claim is only substantiated for blockchain sync, it remains untested on a live network.

Yes. He claims that ECC signature verification can be paralellized easily by a cluster of nodes and the bottleneck of the network is the "order matching" engine which has to be single-threaded (e.g. atomic) for avoid exploits .. The results from LMAX basically state how to separated authorization (via ECC signature) and execution of orders efficiently .. and that is what graphene is going to do

[monsterer]

https://bitshares.org/blog/2015/06/08/measuring-performance/
I like it that you are getting more and more critical lately .. !

There are a lot of wild claims flying about - I am trying to get clarity on the facts. So, according to that doc, and bytemaster's comments above, this claim is only substantiated for blockchain sync, it remains untested on a live network.

[xeroc]

I've asked for this clarification before, but lets re-state the question for clarity: the 100K TPS figure that gets quoted, is this for actual generated transactions, or just for blockchain sync?

https://bitshares.org/blog/2015/06/08/measuring-performance/
I like it that you are getting more and more critical lately .. !

[monsterer]

What you will be able to verify is that we can "reindex" the blockchain at over 100K TPS because we get to skip signature verification and the total data size is much less. 

I've asked for this clarification before, but lets re-state the question for clarity: the 100K TPS figure that gets quoted, is this for actual generated transactions, or just for blockchain sync?

[clayop]

What we will be testing with the first test network is 1000 transactions per second and 1 second block intervals.   This should require about 100KB /sec of network bandwidth and be viable on the simple VPS systems that we are using today.

Sounds promising. BTW how many transactions Ethereum can process per second?

[arhag]

100K TPS requires over 10 MB/sec of network bandwidth

I'm guessing that capital B wasn't a typo. Do we have any idea what sort of infrastructure requirements Visa has, or perhaps Verizon for tracking cellphone data/minutes usage? I'd imagine it is comparable.

Considering 100K TPS at 100 bytes per transaction would require more than 10 Megabytes per second and just one signature alone requires 65 bytes... yeah the capital B is not a typo. And if most of the transactions are confidential transactions, we are looking at an order of magnitude or more increase in network bandwidth (in that case, at 100K TPS, 1 Gbps would likely not be enough).

[Riverhead]

100K TPS requires over 10 MB/sec of network bandwidth

I'm guessing that capital B wasn't a typo. Do we have any idea what sort of infrastructure requirements Visa has, or perhaps Verizon for tracking cellphone data/minutes usage? I'd imagine it is comparable.

It will be a good problem to have.

Yes.. clearly the standard for being a witness is going to be raised significantly. 

As long as the ramp up proceeds at a steady pace; giving us time to sign contracts with data centers, etc.

[BunkerChainLabs-DataSecurityNode]

100K TPS requires over 10 MB/sec of network bandwidth

I'm guessing that capital B wasn't a typo. Do we have any idea what sort of infrastructure requirements Visa has, or perhaps Verizon for tracking cellphone data/minutes usage? I'd imagine it is comparable.

It will be a good problem to have.

Yes.. clearly the standard for being a witness is going to be raised significantly. 

[Riverhead]

100K TPS requires over 10 MB/sec of network bandwidth

I'm guessing that capital B wasn't a typo. Do we have any idea what sort of infrastructure requirements Visa has, or perhaps Verizon for tracking cellphone data/minutes usage? I'd imagine it is comparable.

It will be a good problem to have.

[sudo]

100K TPS requires over 10 MB/sec of network bandwidth and servers with 32 cores and 256 GB of RAM.    The transaction history will grow at 1000 GB per day.  None of these issues are problems when we have network income of 1 billion dollars per day (assuming 100K per sec $0.20).

What you will be able to verify is that we can "reindex" the blockchain at over 100K TPS because we get to skip signature verification and the total data size is much less. 

What we will be testing with the first test network is 1000 transactions per second and 1 second block intervals.   This should require about 100KB /sec of network bandwidth and be viable on the simple VPS systems that we are using today.

awesome  can‘t wait to see

then how the devshares snapshot?  from bts2.0 or  from devshare0.9.x?

What you will be able to verify is that we can "reindex" the blockchain at over 100K TPS because we get to skip signature verification and the total data size is much less. 

What we will be testing with the first test network is 1000 transactions per second and 1 second block intervals.   This should require about 100KB /sec of network bandwidth and be viable on the simple VPS systems that we are using today.

[bytemaster]

100K TPS requires over 10 MB/sec of network bandwidth and servers with 32 cores and 256 GB of RAM.    The transaction history will grow at 1000 GB per day.  None of these issues are problems when we have network income of 1 billion dollars per day (assuming 100K per sec $0.20).

What you will be able to verify is that we can "reindex" the blockchain at over 100K TPS because we get to skip signature verification and the total data size is much less. 

What we will be testing with the first test network is 1000 transactions per second and 1 second block intervals.   This should require about 100KB /sec of network bandwidth and be viable on the simple VPS systems that we are using today.

[Ander]

Cant wait for the public testnet, and for people to confirm that 100k tps is possible.

[fav]

just wait for the testnet (soonTM) and everyone can proof it

[clayop]

People are still dubious about BTS 2.0's 100k throughput capability. This is maybe due to that we claimed and tested this but haven't proved it yet publicly.

I personally tested it in my VM with flood_network command but I only can produce 120 txs blocks.

If we have the block with 200k txs (since block time is 2 second), we can just screenshot and show it for convincing non-believers.

Any ideas?