Author Topic: Witness pay framework  (Read 685 times)

0 Members and 1 Guest are viewing this topic.

Offline santaclause102

  • Hero Member
  • *****
  • Posts: 2487
    • View Profile
Witness pay framework
« on: April 29, 2017, 07:36:53 pm »
There seems to be consensus that discussing witness pay every two month is no long term solution. The only way out is a general framework that automatically adjusts witness pay according to BTS market cap changes. So here is my suggestion. I am looking forward to your feedback on it.

The building blocks (with my rough estimates):
A) Amount of work in hours per month and witness: 5
B) Hourly rate for witness work: 60 USD
C) Hardware and network expenses per month and witness: 100 USD
-> A * B + C = 5*70+100 = 400 USD / month and witness

-> Overpaying witnesses 2x to increase their loyalty -> 400*2 = 800 USD / month and witness. Rather than going lower here I would rather have less witnesses. As fav pointed out the quality (responsiveness in case of a fork for example, quality price feeds, not missing blocks) of witnesses is much more important than their number.

Adaptation algorithm: Delta witness pay = Delta BTS market cap * 0.1. This means that if BTS market cap for example increases 100% witness pay increases 10%.

If we reach Bitcoin market cap (~20 bn) that would be 45,000 USD per month per witness and 900,000 USD of security costs for the whole BTS network per month if there are 20 witnesses.

If you are a witness please report how many hours of dedicated witness work you put in monthly and for what specifically. Also, as a witness, what hourly rate do you think is adequate? What hardware/network costs do you have and for what specs? 
« Last Edit: April 29, 2017, 07:39:16 pm by santaclause102 »

Offline Thom

Re: Witness pay framework
« Reply #1 on: April 29, 2017, 11:11:43 pm »
Your server cost estimate is too low as is your hourly time to maintain and upgrade witness servers.

You might be able to run two 8GB shared VPS servers for $100 / month, IF you look REAL HARD. However, that would leave NOTHING to pay for a witness's time. Plus, keep in mind 8GB is not enough memory for reliable and stable service at the current rate of network use. I am moving to dedicated servers with at least 16GB of RAM for all future node installations. Seed nodes can get by right now with only 8GB as long as adequate swap is configured.

I have also rented a 32GB server for use on the testnet.

And as for the amount of time per month I devote to witness work: It has been far more than 5 hours a month since February. Here is my estimate of hours spent this month alone (April 2017):

Hrs         Task
-----         -------
3             Hosting Service Research (locations, cost, owned by, complaints / community opinion)
8             Community involvement (Friday mumble, reading forum, Telegram etc)
10           Server upgrades & installation
4             Software updates and feed tweaking

To provide reasonable compensation for server costs 3 servers should be factored in MINIMUM: 1 Witness node, 1 Seed node and 1 node for the testnet. The testnet doesn't have to be rented continuously, it can be spun up for testing and destroyed when testing is not underway.

It's there for all to see, very few witnesses run seed nodes in addition to their witness node. A seed node can serve as a hot backup if configured with adequate RAM.

Another factor not taken into account is how often feeds are published. Verbaltech2 holds the record for the most operations performed which is due to the number of assets I post feeds for as well as their publication frequency.

I will also say this again, that few people are considering the architecture, reliability and decentralization aspects of the network like Wackou has. Nobody seems interested in looking at potential attack vectors and how to protect witness nodes from being discovered and DDoS'ed. Wackou's backbone scheme is the ONLY work I've ever seen that addresses such issues. Where is any of that in this discussion? Security of the blockchain is obviously extrememly important, so why is the job of witnesses treated with such little respect?

B/c many witnesses do not live up to their responsibilities and many shareholders don't bother to vote them out.

The fact that the number of witnesses does NOT impact the cost of security to the network and shareholders is not intuitive. People think the expense is Pay per block * number of witnesses == cost of network's security. That is FALSE, it actually is:

(Pay per block * ( ( number of seconds in maintenance interval ) / 3 ) ) / number of active witnesses. The witness pay pool is set entirely by the pay per block (currently 3 BTS) and the length of the maintenance interval. Increasing the number of witnesses can only reduce each witnesses' pay and does not increase the direct cost of securing the network.
Injustice anywhere is a threat to justice everywhere - MLK |  Verbaltech2 Witness Reports: https://bitsharestalk.org/index.php/topic,23902.0.html

Offline 天籁

  • Hero Member
  • *****
  • Posts: 643
    • View Profile
Re: Witness pay framework
« Reply #2 on: April 29, 2017, 11:43:54 pm »
+5%

Offline sahkan

  • Full Member
  • ***
  • Posts: 185
    • View Profile
    • BitShares DEX
Re: Witness pay framework
« Reply #3 on: April 30, 2017, 01:36:21 am »
Thom has some very good points. I will add a couple of things from my point of view:

Bare minimum:
The truth is that all you need is a home computer with 8GB RAM and some 8GB swap file and you can run a witness. You don't have to do anything else except for being voted in as active <-- very cheap and very crappy at the same time

What a witness should have (VPS/server running costs):
- a server can cost anywhere from around $1K to $3K
- Co-location costs around $80 to $200/month if you want something decent
- Dedicated Server $80 - $150
- VPS  about $60 & up/month if you want decent RAM

In my opinion an active witness should have at least 3 (I have 3 and will be adding more - one in May)

It costs BTS to do price feeds,  and here again the number of assets that witnesses post prices for vary because there is no set list that we should go off.

The more servers we have (I think Thom has 6) the more time it takes to maintain them. Some witnesses provide additional work for the community  and maintain other products.

I think it's important to build a capable network of Witness Node servers and the motivation should not be the witness pay. We just need more active witnesses capable of providing the backbone infrastructure for DEX. For example, in addition to the voted in witnesses I would like to see a Network select witness that's selected during the maintenance window based on his/her Server capabilities (connection, RAM etc.) - if there is a way to measure that properly. Because few proxy voters have not changed their votes for months some active witnesses feel very secure and provide very little feedback on their activities.

As far as the hours and monthly pay, lately I have been putting in a lot more than 5 hours but I am ok with your $400 as pay was not my motivation (and I suspect there are few others that share that opinion). If we can focus on improving DEX and adding traders, it will help us to generate higher income for DEX. At the same time based on the BTS reserves we should have a plan to decrease witness and worker pay to generate an overall profit and maybe create a dividend for all share holders. With the rising price of BTS we might be able to get there quicker.

Offline twitter

  • Full Member
  • ***
  • Posts: 176
    • View Profile
    • open a online trading account
Re: Witness pay framework
« Reply #4 on: May 06, 2017, 04:06:21 am »
 +5%
open a trading account today ....  https://bitshares.openledger.info/?r=bue

Offline abit

  • Committee member
  • Hero Member
  • *
  • Posts: 3297
    • View Profile
    • Steemit Blog
  • BitShares: abit
  • GitHub: abitmore
Re: Witness pay framework
« Reply #5 on: May 08, 2017, 04:48:11 pm »
bump.
BTS account: abit
BTS committee member: abit
BTS witness: in.abit

Offline Customminer

  • Sr. Member
  • ****
  • Posts: 492
  • Bitshares FTW!
    • View Profile
    • Gridcoin.US
  • GitHub: grctest
Re: Witness pay framework
« Reply #6 on: May 08, 2017, 06:52:16 pm »
I think paying a fixed bitUSD rate would be more sensible than paying in terms of BTS, because the committee is having to reevaluate witness pay at increasing frequencies.
Hertz, Beyond Bitshares, Gridcoin!

Offline tehdos

Re: Witness pay framework
« Reply #7 on: May 09, 2017, 01:12:52 am »
I think paying a fixed bitUSD rate would be more sensible than paying in terms of BTS, because the committee is having to reevaluate witness pay at increasing frequencies.

Offline BunkerChainLabs-com

Re: Witness pay framework
« Reply #8 on: May 09, 2017, 01:50:16 am »
In terms of server resources I would suggest the standard should be at this point for bitshares:

2X Dedicated Server minimum specs 1 Quad Core CPU, 32GB RAM, 64GB SSD drive (preferably RAID1), 100mb uplink
2 X Virtual Server Seed Nodes of whatever is required at that time with open API connections for the public to use.
1 Fully qualified domain name in your control not using a free service domain.
Located in a minimum Tier 2 uptime standard data center, or the equivalent.

If we can agree on this being the standard by which witnesses should operate, then it could help with the baseline in regards to the framework of network resources for each node.

I think the 5hr a month baseline is decent as a retainer at this stage.

All that said, I think a baseline that comes to $900 a month is a good place to be with all this in mind.
+-+-+-+-+-+-+-+-+-+-+
www.Peerplays.com | World's First Decentralized Tournament Platform Built Entirely on the Blockchain!
+-+-+-+-+-+-+-+-+-+-+