Hey everyone!
As you may know, a couple of weeks ago I created a DHF proposal to help fund the global expansion of api.deathwing.me both to increase overall capacity while reducing latency and improving responsiveness for Hive users worldwide that utilize this API on their day-to-day browsing experience.
This is going to be a relatively short post, I just wanted to let everyone know that the GeoDNS balancing is now active and operational :)
So first, pictures!
Here are some of the racking process and overall glam shots from the good ol' server rack.
The servers are located in Hurricane Electric's Fremont 2 datacenter in Fremont, California -- so users on West Coast, rejoice!
In case anyone wants to "evade" the GeoDNS for testing or other purposes, just contact me and I can get you regional access.
Second! Node and Cluster Details
Most of the nodes are equipped with ASRockRack X570D4U Motherboard housing a Ryzen 9 5950X, 128GB DDR4 RAM and 2x2TB Samsung 980 PRO (PCIe Gen4) NVMe drives. (Though I did pick out an alternative drive that's roughly equivalent in price but hopefully faster performance for one of the servers for testing purposes)
All the nodes are setup using HAF and use HAF structure. (HAF itself, HAFAH etc.)
As HAF is under heavy active development, things are changing almost every day, however, as soon as everything is more or less stabilized, I am planning to write a shell script to speed up the overall deployment of what we call a "full" node, so by simply running the script, you can also have a full-fledged HAF node :)
But what is GeoDNS?
Previously, all nodes under api.deathwing.me were located in the EU. So no matter where you are, you would be connecting to the servers in Europe. It did not matter if you had 500ms latency to the server. Now, with GeoDNS activated, depending on wherever you are in the world, you will connect to the server closest to you, as well as has the lowest latency. So users in the United States have up to 10 times faster connection latency to Hive through my nodes :)
This also means that overall load is shared across continents and clusters, meaning less overall load and more capacity for the clusters.
I see, how about the other side of the world, Asia?
I am happy to also announce that the servers arrived in Singapore, where the nodes will be hosted, just waiting for them to clear customs for the time being.
A note
Thank you all so much for supporting the proposal and making this possible. On top of enabling faster response times and lower latency, this distribution also saves us from many outside factors (such as Solana losing a substantial amount of their nodes because Hetzner decided to close them down, or a fire breaking out in OVH's Datacenter) by having multiple servers in 3 different continents, in 3 different providers, in 3 different countries.
If you haven't checked out the proposal yet, you can check it out here on Peakd.
That's it! Let me know if you have any questions.