Funding Proposal: Public API for Testnet

Backstory

We are, at the moment, running an api for developers to connect to test-net that has 3 nodes. This is publicly available to anyone at https://rpc.pulsar.griptapejs.com & https://lcd.pulsar.griptapejs.com.

Proposal

We propose adding 10 additional nodes to this cluster, and that we build an api gateway that allows devs to signup for an api key to use the API (This is necessary to avoid spam through rate limiting which also insures equal access to all devs.)

The goal of this is to provide an api for anyone who wants to just start building as fast as possible and with as little work as possible.

Budget

Currently the proposal is to lease 10 x Intel E-2286Gs for 1 year, at a price of 150 USD per server per month. In addition, the API gateway software will cost 250 a month. Finally we will build and maintain the service at a cost of $50 an hour. The build phase will be a single 2-week sprint during which 1 dev will work full time (80 hours), this will be followed up with a budget of 5 hours a week for the year.

Rate Qty Time Cost
$150.00 10 Machines 12 Months $18,000.00
$250.00 12 Months $3,000.00
$50.00 40 Hours 2 Weeks $4,000.00
$50.00 5 Hours 52 Weeks $13,000.00
Total $38,000.00

Note

SCRT amount will be calculated at the time the proposal goes on chain with a 10% buffer. Upon receipt of the funds we will sell the SCRT immediately and move to our USD bank account. We will then lock up the servers for 1 year and pay for the service to ensure the price. We will then allocate a developer to build the web page. And commit to 5 hours a week of availability for the year. Any funds not spent will go to reduce the cost of a future proposal if we want to either expand the number of servers, or renew / upgrade them a year from now.

A brief word about leasing.

There are a few reasons not to buy. These are:

  • speed of delivery
  • upgradability
  • reduced overhead and maintenance
  • focus (we want to focus on the API, not procurement, colocation etc etc)

Would love any and all feedback.

Sandy

3 Likes
  1. You can order 2286g mainnet compliant machines and get them delivered within 3 weeks. (I got my first 2 machines within 10 days of ordering) but It is faster to rent.
  2. Maintenance for owning is less expensive than the cost to rent. Roi is achieved in less than a year.

We estimate we save 4k USD per month running our own infra vs the cloud but I understand if you want to focus your efforts elsewhere. Just adding these notes because owning is cheaper.

Questions I have

In regards to the API keys. What are you planning to use to provision the keys?

Nginx plus, kong, something custom, cloud service? Can you specify? Thanks in advance.

3 Likes

The plan is to use Kong. We are testing right now. The 250 a month quote was based on a Kong sub.

I would be open to owning for sure. But at the same time I am nervous about the management as we are def not experts in this area. I think you’re experience allows you to be way more efficient which gets you the savings you mention. For us we would need to start from scratch…

I also worry about the lock in, would it be better to only commit to 1 year, or even 6 months and then be able to upgrade if that make sense?

All that being said, if the community would rather funds go to purchasing we would be up for it. Would
even be open to partnering with someone to handle the colocation side.

3 Likes

Im less worried about lock in because our monthly bill with the data center doesn’t increase as we add more machines to the data center (outside of bandwidth and ip allocation costs). But that is def a byproduct of scale. If the nodes run for at least 10 months, they essentially pay for themselves.

Secret LLC is open to this if you are. :eyes: I would suggest using nginx plus but thats just because of the performance improvements / the fact we have more experience with nginx than kong.


We support other api providers getting funded so don’t take anything i said as not supportive, and we are happy to work with others as we worked with shade for the load balancer they use.

5 Likes

Thanks man, I will look into doing it all with NGINX instead of Kong and of course reduce the prop price if Kong is not needed.

Re: hardware
I do think that the fastest most efficient way for us to get this off the ground quickly is to lease. Maybe I reduce the scope to 6 months though. This would be less of a commitment and during that time we could work on moving beyond leasing. For us, buying hardware just introduces a lot of unknowns. And I don’t think the savings are enough to justify it.

I think the Kong thing is the right place for us to focus on right now. Getting that part of the setup figured out is key and could save us a lot if we can go open source.

Would welcome any and all thoughts from anyone who has done this with NGINX, and will share what we learn as we do.

1 Like

Datahub offer a free api service for testnet. I’d love to see this for production though. Many people have reached out saying that we need competition api providers mainnet.

My feedback is that nginx is open source, similarly priced, has the same or better features than kong (including webgui), and has better performance.

Makes sense. We have NGINX experience too.

I appreciate the feedback. One of the reasons I put this up is that a community member reached out saying that Figments signup was down. It may be that figment has this need totally covered, but I worry that as devs enter the network they are going to be frustrated by only one provider.

For mainnet, I am less interested for a few reasons.

  • Projects going live on mainnet should be in a better position to pay for infrastructure of their own. Maybe improved docs on how to do this would be helpful and I would support anyone working on those.
  • Mainnet infrastructure is mission critical which would require a greater investment of time and resources to live up maintain
  • It is not within the scope of what our business is looking to do long term

I would be interested to hear if Figment does have this covered. I want to do this to help out the network, not just to do it. If it is already covered, than community funds would be better spent elsewhere

Unless things have changed a lot, you need to sign up and get access to their testnet api to do their learn tutorial. I just don’t think there’s any issues interacting with the testnet as it is:

https://secret-pulsar-2--lcd--full.datahub.figment.io/apikey/

However the mainnet, everyone is feeling the last few weeks of issues. I understand if you really don’t want to go down that path, but just thinking, it’d be great if you could get the funding required to run apis for the main net where it is really needed.

1 Like