Drone
Drone is my custom implementation of "Jussi" which is a caching, reverse proxy app used by every node operator. However, with its python implementation and age over time, it hasn't been "kept up to date" per se. Some of its features are lacking or are not working as expected. It is also heavily bloated with features that are not being utilized at all. (Such as AWS S3 support) While I do not mean to "reimplement" every feature of Jussi and replace it, I wanted to improve upon some of its features.
So I have rewritten the entire reverse proxy/caching layer in Rust. Per it's explanation on GitHub:
Drone is an API caching layer application for the Hive blockchain. It is built using Rust with Actix Web, and its primary purpose is to cache and serve API requests for a specific set of methods. While Drone is not meant to be a Jussi replacement, it aims to improve API node performance.
Features
- Written in Rust for optimal performance and reliability.
- Actix Web for high-performance, asynchronous HTTP handling.
- LRU cache with time-based expiration to store API responses.
- Multiple API endpoints support for seamless request handling with HAF apps.
- Caching support for select Hive API methods to reduce strain on API nodes.
Drone is currently in beta and is operational on all api.deathwing.me clusters around the world.
The code will be fully open source once it is fully production ready.
DHF Proposal Updates
As you guys may know, for a while now, I have been running a DHF proposal. In the proposal itself, I have explained that the funds will also be used not only to get started with the expansion but also to gather enough funds for upgrades and additional servers and services over time.
So this post is another update for my DHF proposal, so I can tell you guys more about what's been going on behind the scenes.
I am happy to announce, with the increased overall activity on North America cluster, there is now a new "test" server, custom built 1U InWin RA100 chassis with AMD Ryzen 9 7950X, 128 GB DDR5 ECC RAM and 2x Samsung 980 PRO 2TB drives, and most interesting part, it's completely liquid cooled.
This node will add more into the overall capacity of api.deathwing.me while greatly improving latency and capacity of the North America cluster especially.
For the record, with the addition of this node, the total number of nodes in the cluster has hit 7. Yes, there are 7 nodes running api.deathwing.me every day (And a backup).
Here is a YABS benchmark. The North America cluster is hosted at Hurricane Electric's Fremont, CA datacenter, colocated under Ryamer. For the record, the network CAN be upgraded to 10 Gbps at a later date, should that ever be the bottleneck.
Future
There's still quite a lot of time until the proposal ends, but I was thinking of potentially renewing it. Of course, there will be some differences in funding purposes and I will also have to account for the cost of hosting a node since it grows over time. api.deathwing.me cluster runs on HAF entirely. With its current apps, it takes about at least 3TB of storage space. This requirement will only increase over time. High capacity and performant NVMe drives are quite costly.
I would love to know what your thoughts are for this, are you supporting the proposal now? Will you support it in the future? If yes, thanks! If no, is there anything I can do to convince you?