October 2nd Test Network

[clayop]

What kind of transactions were you making, clayop?

IIRC, transfers will cost approx $0.20 for non-members and $0.04 otherwise.. at 4 cents, 1000 TPS for 15 mins is $36,000.

Sell assets only costs 5 CORE instead of 20 CORE of transfer.

[roadscape]

What kind of transactions were you making, clayop?

IIRC, transfers will cost approx $0.20 for non-members and $0.04 otherwise.. at 4 cents, 1000 TPS for 15 mins is $36,000.

Well, that's not really expensive... 3600 USD will give you:

1 hour of 100TPS attack or... 30 seconds of 12000 TPS attack. Is that not enough to bring down the network?

True, good point.. perhaps eventually the network will automatically scales up fees.

[clayop]

You may be right clayop, but do we need to concern ourselves with that right now? IMO it's more important to address the functionality and feature set that will be available at launch.

Out of the gate we probably won't require 1000 TPS. I do recognize that is only 10% of our claims, and if we DID have a massive response on the launch and we couldn't handle it, it would be very bad indeed.

I'm not sure where the best balance is and what our testing goals should be, but I think we need to get serious very quickly and figure that out, devise a plan to reach those goals and coordinate our efforts to achieve them.

Exactly.. and if each tx costs $0.01, 100 TPS "attacks" cost $1/second, or $3,600/hr. :)

Well, that's not really expensive... 3600 USD will give you:

1 hour of 100TPS attack or... 30 seconds of 12000 TPS attack. Is that not enough to bring down the network?

I killed the network with 15 minutes of 1000 TPS.

[clayop]

You may be right clayop, but do we need to concern ourselves with that right now? IMO it's more important to address the functionality and feature set that will be available at launch.

Out of the gate we probably won't require 1000 TPS. I do recognize that is only 10% of our claims, and if we DID have a massive response on the launch and we couldn't handle it, it would be very bad indeed.

I'm not sure where the best balance is and what our testing goals should be, but I think we need to get serious very quickly and figure that out, devise a plan to reach those goals and coordinate our efforts to achieve them.

I agree with the point that we don't need 1000 TPS at the beginning. But it shouldn't be an attacking vector against witness nodes, so we have to find out a way (e.g. capping maximum TPS).


FYI, based on the current BTS price, the 'attack' requires $18,000 and costs about $3,700 only.

[lakerta06]

You may be right clayop, but do we need to concern ourselves with that right now? IMO it's more important to address the functionality and feature set that will be available at launch.

Out of the gate we probably won't require 1000 TPS. I do recognize that is only 10% of our claims, and if we DID have a massive response on the launch and we couldn't handle it, it would be very bad indeed.

I'm not sure where the best balance is and what our testing goals should be, but I think we need to get serious very quickly and figure that out, devise a plan to reach those goals and coordinate our efforts to achieve them.

Exactly.. and if each tx costs $0.01, 100 TPS "attacks" cost $1/second, or $3,600/hr. :)

Well, that's not really expensive... 3600 USD will give you:

1 hour of 100TPS attack or... 30 seconds of 12000 TPS attack. Is that not enough to bring down the network?

We can thank them for the money and resurrect the network from the last block :D

[Spectral]

You may be right clayop, but do we need to concern ourselves with that right now? IMO it's more important to address the functionality and feature set that will be available at launch.

Out of the gate we probably won't require 1000 TPS. I do recognize that is only 10% of our claims, and if we DID have a massive response on the launch and we couldn't handle it, it would be very bad indeed.

I'm not sure where the best balance is and what our testing goals should be, but I think we need to get serious very quickly and figure that out, devise a plan to reach those goals and coordinate our efforts to achieve them.

Exactly.. and if each tx costs $0.01, 100 TPS "attacks" cost $1/second, or $3,600/hr. :)

Well, that's not really expensive... 3600 USD will give you:

1 hour of 100TPS attack or... 30 seconds of 12000 TPS attack. Is that not enough to bring down the network?

[roadscape]

You may be right clayop, but do we need to concern ourselves with that right now? IMO it's more important to address the functionality and feature set that will be available at launch.

Out of the gate we probably won't require 1000 TPS. I do recognize that is only 10% of our claims, and if we DID have a massive response on the launch and we couldn't handle it, it would be very bad indeed.

I'm not sure where the best balance is and what our testing goals should be, but I think we need to get serious very quickly and figure that out, devise a plan to reach those goals and coordinate our efforts to achieve them.

Exactly.. and if each tx costs $0.01, 100 TPS "attacks" cost $1/second, or $3,600/hr. :)

[Spectral]

I'm not sure where the best balance is and what our testing goals should be, but I think we need to get serious very quickly and figure that out, devise a plan to reach those goals and coordinate our efforts to achieve them.

+5%

That sounds about right. A launch target TPS could be specified, the minimum requirement to meet that target would be identified, so the recommended witness minimum specification would be ready at launch.

Maybe there should even be a max TPS cap (possibly a votable parameter), to protect the network from overload. Is spam an attack vector into BitShares?

[Thom]

You may be right clayop, but do we need to concern ourselves with that right now? IMO it's more important to address the functionality and feature set that will be available at launch.

Out of the gate we probably won't require 1000 TPS. I do recognize that is only 10% of our claims, and if we DID have a massive response on the launch and we couldn't handle it, it would be very bad indeed.

I'm not sure where the best balance is and what our testing goals should be, but I think we need to get serious very quickly and figure that out, devise a plan to reach those goals and coordinate our efforts to achieve them.

[clayop]

What exactly are you guys asking? (iHashFury, clayop)

Are you asking about config.ini values, RAM/CPU/Network/OpSys or what exactly?

I'm sure you have something in mind, some hypothesis. Care to share?

I'm thinking that minimum requirements for 1000 tps may be higher than BM expected.

[Thom]

Yes, we need to describe these physical specs and weigh their performance under various load conditions.

Wackou and I have 7 VPS servers so far, all on the low end of the spectrum, relatively speaking. The lowest being 1GB RAM. Wackou will have to give you the specs for the VPSs he has setup, which the most recent ones were on Vultr with 1GB RAM.

I have 4:
Vultr with a 4GB RAM / Dual core
Crown Cloud 1 2 GB RAM Dual core
Crown Cloud 2 4 GB RAM Dual Core
Bithost 2 GB RAM / Dual core

The above is from memory, I can be more specific when I get back to my workstation.

BM doesn't seem to think our VPS specs should be difficult to meet. Wackou chose 1GB b/c of the low RAM use he observed in testing.

These specs are much more important while testing, primarily the spam testing. This is why the requirements for heavy load testing are different than if we're trying to functionally test transaction variations. It's all necessary testing, it's a matter of coordination. IMO we should first look to verify functionality then look at performance. If something fails it might become difficult to determine if it's a feature logic issue or if it's due to timing / network / VPS limitations.

[iHashFury]

minimum VPS RAM/SWAP/CPUs/NETWORK?

so witnesses know what min hardware spec is required

[puppies]

OK so are we ready to coordinate and establish a new checkpoint?

Has that already been done?

If not I think we should set a time identify whose block will be the first one in the new checkpoint, and then we can setup new seed nodes.

I am trying to reset locally.  If I can I will post directions.  Although if we leave allowed missed blocks at 1000 the network wont last long.

will try to catch up if there is a way

I can't get it to work without enabling stale block production.  We will have to wait for someone else to figure it out. I have to go to work.  I'll try to keep an eye on this thread, but will probably be pretty busy.

last signs of life .....

I'm wondering what these witnesses' VPS specs are.

My witness is currently running on a dedicated server Intel Xeon quad core at 3.2Ghz, and 16GB ram.  I still miss random blocks.  Perhaps its due to networking, and or storage, as the server has a mechanical hdd. 

[Thom]

What exactly are you guys asking? (iHashFury, clayop)

Are you asking about config.ini values, RAM/CPU/Network/OpSys or what exactly?

I'm sure you have something in mind, some hypothesis. Care to share?

[iHashFury]

last signs of life .....

I'm wondering what these witnesses' VPS specs are.

What is the required minimum initial VPS spec?