Nxtbase node sponsorship expires 3/31
Issue #27
resolved
When I raised funds to do 100+ nodes in 10 data centers, the terms of the deal were that I would deploy, manage, and pay for the nodes through the end of March. The end of march is a few weeks away.
Looking for ideas from the committee.
Comments (60)
-
-
reporter
@marcus03 - Payment is for hosting, bandwidth, and personal time (gotta pay the bills). Servers are at Amazon, Rackspace, and Linode for both geographic and provider diversity. A node is run as a single virtual server.
-
Thanks. Can you make a small list of planned expenses for a month for one node?
Hosting: ... NXT per month per NRS node Bandwidth: ... NXT per month per NRS node Personal time: ... NXT per month per NRS node Anything else...
-
reporter
1-2gb node with normal bandwidth charges is $25-35USD per month depending on location.
You can get much cheaper VPS options but you won't have the nice features like image cloning, international data centers, etc.
It's probably easier if I put together a proposal and submit it. I'll run the numbers and think how I could improve things and incorporate some of these brainstorming issues.
-
@ferment Sounds good for the start.
I would like to add that we definitely need long-term strategy for self-sustaining service providers. Any ideas on that?
-
reporter
@ChuckOne One idea that I had was if you could integrate some kind of profitable PoW (distributed computing problem) then one could use the mostly idle nature of nodes to do other stuff.
For example, you could have a node do image conversions from one format to another and the node owner would get paid in NXT per unit of work completed. The work execution framework would run as an extension to NRS.
Building CPU mining pools on top of nxt nodes would be another example.
-
reporter
@EvilDave wrote:
I think it's an obvious decision, we need to have those nodes up and running, IMHO The only controversial part is whether Ferment gets a bounty for his own work on this. F; Have u recieved funding/bounty for your own workng time up til now? Do u want or expect payment for your worked hours on the VPSes?
Yes, I received sponsorships to pay for the nodes through the end of march. I've received bounties from BCNext/CFB as well for various contributions.
I'd happily run a few to support the community, but running 100+ across 10 data centers requires both time and financial commitment.
I think there are 3 approaches that could be applied to me or anyone else:
-
Salary Bounty + Costs = Qualified individual gets paid on some kind of time frame to manage up to a certain amount of nodes. Committee pays hosting costs for these nodes directly. This is my least favorite, but it is what is implied when people ask about costs and time. It's highly centralized and requires trust.
-
Bounties are paid per node per period of time directly to maintainer. How much time/expense is not the concern of the funding committee. The committee is only interested in the end result - a fast and resilient network. Nodes should receive bounties if they meet a clearly specified SLA regarding uptime, performance, updates, etc. This is the most common approach (sponsorships, peer explorer). In the absence of #3 below, I prefer this model, but it rewards investment in massive automation and the ability to grow on demand.
-
Decentralized service providers are built on top of the nxt network. Nodes are rewarded through some kind of PoW that supports the business of the service provider. This is a long-term sustainable strategy until foraging become viable. Work processing code is distributed as "plugins" to NRS and run on nodes. Think Seti@home or CPU mining.
I'm sure there are other interesting models.
-
-
We need professionally run nodes... I think we have over 500 nodes now on average, ok they are not all like Ferment's VPS.
I think we need the 10 public ones and some more peer nodes but do we still need 100+?
Also are they being used for forging - and if so by whom?
Also as a funding model - look at my BCT PM for a bounty based model which is a bounty divided up by people earning Node Tokens via a modified time-limited redeemable asset (node coin 2.0)
-
Guys: I like approaches 2 (short term) and 3 (longer term). However, I feel we should do at least some of our brainstorming on requests in public, particualry if it is a request for funding from an INF-COM member. I've reposted most of teh above on: https://bitcointalk.org/index.php?topic=506757.new#new
and I'm going to wander over to the main thread and say hello in a bit. maybe we can get some extra input from the community.... Has anyone posted an introduction from the INF-COM on the main firehose thread yet?
-
nope - you are the spokesperson.... :)
-
reporter
@EvilDave I think you can link the issues directly. I believe they are public.
-
Been avoiding the firehose thread....it's a monster time sink. Anyhow, intro posted. Ian: did Damelon get it right that u will be attending the BTC A'dam conference in any case ?
And yeah, we need to work up some sort of performance metrics and audit for the entire NXTwork. I'm feeling a bit in the dark about what we have and what we actually need. But I do feel that Nodecoin is essential......
-
reporter
Link for Nodecoin: https://bitcointalk.org/index.php?topic=345619.msg5666282#msg5666282
-
reporter
rickyjames's comments from BTT PM:
Those nodes got set up in December primarily as a response to a DDoS attack. To me, question one is are they still needed to deal with a DDoS attack - is their original rationale still valid? This has two parts - is the DDoS attack still underway, and have enough other nodes come online in the past few months to take up the slack if these nodes go offline?
I think one of the first thing the infrastructure committee should do is a census or inventory of which major NXT players have what servers, and what were their monthly network /bandwidth costs. This is the database required to have facts to evaluate further spending on infrastructure. A secondary point to learn from this exercise is to identify who found and deployed the cheapest/most cost-effective servers - these are the ones to duplicate going forward.
I think it is very important from an ethics standpoint that all proposals for funding, even by committee members for something like upcoming server fees, go through the same identical public application procedure. What you don't want to have happen is to have people think you went behind closed doors and started voting yourselves NXT.
I think the most important long-range task for InfCom is 1000TPS. But the most important short-range immediate task is to assess the current health of the existing NXT network and see if it needs help. Maybe we can take some servers offline with no ill effects. Maybe we need to add more hallmarked nodes. Maybe we need to dole out some NXT subsidies to support current ops, maybe not.
-
reporter
@chanc3r:
I think we need the 10 public ones and some more peer nodes but do we still need 100+? Also are they being used for forging - and if so by whom?
I have no idea if we still need 100 to be honest. They are not used for foraging. One is hallmarked with panda's account (
#1node on peer explorer) and the other 100 are hallmarked with the account that I held the sponsorship NXT in which I think incidently has lessened their impact based on the way that traffic is weighted.It seems like weighting traffic based on hallmark account balance might actually do strange things to our attempt to spec network performance. Might need to revisit with JLP or CFB.
-
My take on this:
We don't know the status of the network yet and can't make an educated descison. We need to get the data as soon as possible, but this will not happen until 03/31.
We should fund the existing nodes for another 2 or 3 months and then reevaluate.
@ferment: Make a proposal which states how much you need peer node or in total per month.
-
btw - my problem with node coins - the Nodecoin Whale!!
https://bitcointalk.org/index.php?topic=345619.msg5674729#msg5674729
-
reporter
@marcus03: I'll figure out how to reduce costs and maybe do a few less nodes. It'd be nice to know how important they really are to the network. I'll see if we can find providers that take bitcoins.
-
From 'james' on how to use node coins... part of a much longer series of PM's but creating a demand for node coins could be seen as creating a demand for 'nodes' - this is james suggestion below... Thoughts?
If we make it public that nodecoins are set to be worth .01 NXT by a floor bid from inf-com and also a ceiling price of .02 NXT by promise of inf-com to sell at that price, then that would remove all speculation. If you put the ceiling at .05 NXT it could create some speculation, but the advantage is that it could become self funding as inf-com would make the spread of .04 NXT to subsidize future purchases.
-
Thoughts? Wrong issue you've posted this to... :-)
-
oops - possibly not... I was thinking a trial.... how about we fund some nodes directly so we are secure... for the others... we calculate the cost of funding a node... we know how many node coins a nodes would generate we can calculate the unit value of a node coin we place daily or weekly buy orders in the AE at a unit cost that might stimulate the running of the right number of nodes see if it works?
-
reporter
Create a new brainstorm issue for Nodecoin discussion?
-
I'd say create a new issue for nodecoin. This one is specifically for ferments nxtbase nodes expiring on 31/03.
-
@marcus03 Agreed. Let us focus here on the nxtbase node issue.
-
OK, leaving nodecoin out of the equation.
We are still more than a little bit in the dark about wtf a healthy NXTwork looks like, and also how the individual components contribute to NXTwork performance.
I vote to fund Ferment immediately for 50 of the current 100 VPS. The end of the month is coming up and losing 100 NRS nodes will probably have a negative impact on the NXTwork.
We really need to measure the NXTworks performance, though, both now and in the future. TPS would seem to be the most important metric, is there a simple method to benchmark the NXTwork with a large tx volume ?
Once we have a performance metric defined and measured, we will be in a much better position to make decisions on VPS needs.
-
Ferment we need a price for 50 VPS per month and then we can vote an amount to fund these... I can't vote at the moment because I don't know how much NTX I am voting for..
-
reporter
Working on it. I'm running numbers on moving everything to EC2.
-
Ferment: Some numbers for u: just had a look at the billing history for my EC2 instance (Win2008, no hallmark, no forging, totally public, NRS 7.5) and it was $2.38 for January, $0.36 for February and nothing so far on March billing. I'm going to upgrade to 8.x soon (hopefully Sunday) and add a hallmark, see how that effects traffic and costs. I'm also thinking (if costs remain so damn low with the 8.x) of setting up a few more nodes outside the continental US, preferably in Europe. I know geography isn't that important, but checking PeerExplorer shows a heavy concentration of US nodes.
-
@EvilDave: I've seen these prices mentioned before, but don't understand them. The cheapest server seems to be the t1.micro for $0.020 per hour, which would sum up to US$14.40 per month (plus storage). Which product exactly are you using to pay so little?
-
On phone, forgive mistakes. I'm using the Amazon EC2 instance, I think BaiMangal made a guide for setup, and Pinarello rd-posted it a couple of times. If u can find the guide...... I'm nowhere near a PC at moment, be at least a couple of hours b4 I can check it out properly.
-
Is $10-$20 per month a realistic price tag for a server? Complete maintaince?
-
reporter
@ChuckOne $15-30 depending on location and provider. That's assuming 1gb-2gb memory.
The problem with all of this is that hallmarking a node has a much larger influence on a nodes impact on the network then the resources of the node. For example, the nxtcrypto nodes are beefy with ram and cpu - much bigger than 1gb ram average nxtnode. However, if you look at the ul/dl stats for panda.nxtnode.com, it's almost the same as nxtcrypto. This is due to the resulting weight of his hallmark. So, if panda.nxtbase.com was beefed up would it out pace nxtcrypto servers? Maybe, but given that it's at 90% of the impact but at 1/8th of the size (much lower cost) is pretty curious.
Hallmarks have a much greater impact on the network than node size or quantity.
Since we have a little time before the end of the month, I want to find a nice model for node resources, deployment, and hallmarks so we can offer the same deal to other folks to run nodes as nxtnode. The question is what combination of attributes creates the greatest impact (bang for our nxt).
-
I would suggest to fund 50 nodes for another 2 months.
2 months * US$20 * 50 servers = US$2000 (currently around 43.000 NXT).
Ok, I really think that Amazon is a bit on the expensive side and I am not sure if there is a huge benefit from paying the extra. I somehow think that we rather need more low-tech nodes instead of less high-tech nodes with high availabilty.
In two month, we should be have an idea about alternative VPS hosters with different locations for their servers and ideally a way for operators to easily manage lots of NRS instances. (The above would cost 1/4th with a hoster like budgetVM, which also offers 4 different datacenters.)
We also still need to get informed about if we actually need these servers, but that might take longer than 2 months.
-
reporter
Crazy IDEA Alert!!!
Can @Come-from-Beyond still test the network robustness? If so, let's turn them all off on April 1 and see what happens. Then we can add them back as needed.
It'll be a great decentralization test.
-
@marcus: i'm running a T1 micro instance, admittedly with only 0.5gb. Going to set up a couple of extra instances in the next few days,, see how that effects my billing, which is currently almost nothing. Lets see what happens when I start scaling up a little. I'd also like to know exactly how much my VPS contributes to the NXTwork, but I'm pretty sure that more nodes can't be a bad thing.
@Ferment : like the idea. Could be exciting....
@everyone:
Based on all of our discussions so far, I would like to propose that we fund Ferment for 50,000 NXT. This is based on 50 servers for 2 months, + a reasonable chunk of NXT for Ferments own pocket. I can't be absolutely certain about this, without a NXTwork measurement tool, but I think that these VPS are going to be useful.
BUT: first I want to see a proper proposal document with lots of numbers ! No finance without proper paperwork.
EDIT: Ignore this bit, see Ferment is already there: I also request that Ferment, as part of his proposal, checks out the Amazon T1 micro instances, and maybe use the funding to spread VPSes around different providers, to give us a better idea of the actual costs/benefits of VPS ownership. Stop ignoring.
i'm also a fan of lots of small nodes, rather than a smaller number of heavy hitters, but again I can't back that up with any real evidence, other than that a larger number of small nodes will be more difficult to DDOS or attack.
-
@ferment: I don't think the monitoring is still available. As far as I understand, it only worked for detecting the level of DOS attacks. But you would need to ask CfB for this.
-
@ferment :
We are looking at one week left to get this sorted out, could you get a proposal together for voting ?
-
reporter
Working on it now. I'm providing a list of different providers and costs with a setup fee and monthly maintence.
I'm going to stay with 1-2gb memory nodes since nothing has shown me that resources trump hallmark balance.
-
reporter
Here's wiki page of cloud providers with a cost per node (USD) per month and the data center location. https://bitbucket.org/nxtinfrastructure/committee/wiki/Cloud%20Server%20Pricing
If others want to add more providers or options, please feel free. I went through these because they have nice tools for provisioning, imaging, etc which will cut down on management and setup time.
-
reporter
I should also note that current 100 nxtbase are spread among 3 providers (AWS, Rackspace, and Linode). In addition to monthly costs, they average $1-5/month in additional bandwidth.
-
reporter
Here's my proposal. It's simple so I'll just put it here.
- Committee decides how many nodes, provider, location, and length of term. See the Pricing Chart for details. Nodes will be prepaid monthly or whole term to Ferment in NXT or BTC with a 10% handling fee. Exchange rates will be agreed upon at time of funding based on DGEX order book and Coinbase sell price.
- Ferment will setup nodes for 1000 NXT per node.
- Ferment will maintain nodes for 350 NXT per month. Nodes will be upgraded within 1 business day of JLP post.
- Nodes can be setup/managed as production or testnet.
- Nodes can assigned be assigned any domain name, platform, known peers, public API status, or hallmark as specified by the committee.
-
OK, I'm happy with the above, though maybe the initial set-up price of 1000 NXT/node is a bit on the high side, IMHO.....that'll cost 50,000 NXT just in set-up costs for 50 nodes. F: how many hours setup time do you need per node, assuming that u image as much as possible ?
-
reporter
@EvilDave If you use existing nodes, then no setup fee.
Setup time varies depending on provider, but about 10 minutes including provisioning, booting from image, updating image packages, generating node-spefic cfg files, deploying NRS, DNS, testing, monitoring, etc. Node setup fees were slightly padded to cover initial development time if the committee choose different providers than what's in use now.
Setup time could eventually get down to just a few minutes and be completely "brainless" with enough automation of provisioning, DNS, and image initialization (would require several hours of coding however and commitment to specific providers/APIs).
-
aha ! That makes sense, withdrawing objection on setup costs.
In the absence of a network performance metric or performance targets, this is all gonig to be guesswork and intuition, so:
Keep the 40 best nodes from our existing NXTwork, use your own judgement on 'best'. Set-up 10 new nodes from other providers, just to see how they perform.
-
Node drop off
-
ignoring nodecoin - we can see public hallmarked nodes which are the most important. so we can script the uptime we know the owning account from the hallmark we could send them nxt for doing this. probably need to check a few public nodes to make sure we get all the hall mark nodes then just poll them for uptime...
we can create a bounty and divide it between the hall mark nodes based on the stats we can retrieve.
thoughts?
-
Lets ask Ferment to concentrate on setting up heavy hitter nodes with big hallmarks on his remaining VPSes. If i ever find klee, i'll get his permission to use our funds hallmark.
I've started encouraging people to set up AWS freebies for themselves, which will help stabilty and blockchain propagation.
Maybe we should finance peerexplorer for their caches ? That seemed to be a good idea that just fizzled out slowly....
-
reporter
Given the current NXT->BTC->USD rates, I'm starting to lean toward not spending commitee NXT on hosting fees.
I'd much rather the committee funds were used for bounties on infrastructure related tools or funding peerexplorer.
I guess I'm allowed to vote against myself? Or crap, I can't vote. LOL.
-
From phone: OK, but do you have the costs covered for VPS hosting for a while ? Be back online in 8 hours or so......
-
reporter
@EvilDave no, I need to start shutting them down today. I'll do it over a few days though.
-
reporter
I've left the 30 at Linode online since they're cheap. I'll let them roll one more month.
-
@ferment I'm with u on the not liking to spend committee NXT when the exchange rate is so damn low.....but if it's really needed, we may have to bite the bullet, and use them. Do you have any back-up for your decision ? I don't want u to get in the situation of being blamed for any NXTwork issues that may appear to be a result of your decision
Still, keeping 30 up will be useful. Can we give these heavy hallmarks (from the InfCom fund, for example) to take up the slack from losing the other 70 ?
-
reporter
@EvilDave you guys can vote to fund whatever amount of nodes that you want and hosted wherever. I reviewed the billing and some of the exotic nodes with high bandwidth were running over $40/month. Given the drop in NXT to fiat, I'll be upside down if I kept going with them past this month. So, they needed to go anyway.
What would make sense to me is to bring up 40 at Digital Ocean and see how things go. At the current prices at this moment, it would be 400 NXT per node (Linode 800 NXT/month). That could change so I'll need to reevaluate the conversion rate every month.
I'll stick to my 1000 NXT per node setup and 300/month management fee for a few months.
-
Crunching numbers:
400 x 40 = 16,000 hosting 300 x 40 = 12,000 management 1000 x 40 = 40,000 set-up costs
Gives a total of 68,000 NXT
Lets see if we can hustle up some more InfCom members and get some more input on this, and also keep an eye on the NXTwork, see if any issues/problems turn up.
-
Where are we with this now? Does ferment need money? Are we surviving without the nodes? I'm confused....
-
It looks as if the dropping out of many of the hosted nodes has had very little effect on the overall NXTwork.....i'd like to hear Feremnts further input on this, and any other thoughts people may be having.
-
Keep an eye out... if we see / hear any issues then we create some quick bounties for people to run nodes for a while... takes what, about an hour to set up a VPS for NXT.
-
reporter
I took the rest of them down yesterday. Let's watch for problems. However I think it will be totally fine.
Vote to close this issue?
-
Close it.
-
reporter
- changed status to resolved
no new funding required!
- Log in to comment
How much do you need per node/VPS per months and for what is it spent? VPS hosting costs only or something else? Where do you host? Number of nodes=number of VPS?