Imagine a machine learning model that can process a data stream (eg:the live video stream of a product launch, financial earnings reports, a political speech, or general consumer sentiment gauged by processing a real-time Twitter feed) and use the information to execute stock and commodity trades. Such algorithmics (or algos) require a deep pool of computing resources to quickly process and execute upon identified trends. Equinix Metal provides just such computing power through its bare metal cloud. It also provides the network interconnection capabilities to connect to upstream financial service providers.
This repo showcases the use of Equinix Metal to deploy one such algo which communicates to the Alpaca stock trading API. Alpaca provides both a REST (pull) and streaming (push) API (accessible via the programming language of your choice). For this repo, we use a sample algo written in Go. The deployed algo uses a "long/short" to continously monitor a select bucket of stocks and identifies those to go long (purchase) and those to go short (sell). The algorithm runs continously with no human intervention, and closes out all positions before the end of the trading day.
This repository is Experimental meaning that it's based on untested ideas or techniques and not yet established or finalized or involves a radically new and innovative style! This means that support is best effort (at best!) and we strongly encourage you to NOT use this in production.
An account with Equinix Metal and with Alpaca are required to run this algo. Equinix Metal infrastructure is available on a per hour charge and the Alpaca API and trading services are free.
No financial outlay with Alpaca is required to run this demonstration as it uses a "paper trading" account. Paper trading uses live market data and while trades that are submitted are recorded and profits/loss are tracked, the gains (or losses) are purely fictional.
Equinix Metal provides bare metal infrastructure (storage, processing, and networking). An Equinix Metal account can be self-provisioned at https://console.equinix.com/signup.
Alpaca provides a trading API and access to the U.S. financial markets. An Alpaca account can be self-provisioned at https://app.alpaca.markets/signup.
Deploying an Algo requires underling compute infrastructure, Alpaca credentials, and the algorithm itself (source code or binary). The long-short algo is written in Go, run on an Ubuntu server, and deployed from Equinix Metal.
Equinix Metal provides both an API and GUI to deploy infrastructure. This algo requires a single compute instance so the GUI will be used. Via the Equinix Metal console (https://console.equinix.com/), create a new project, and deploy an "on-demand," IAD 2, c3.medium.x86, Ubuntu 20.04 instance. Once it has started, copy the SSH root password (via the verview tab) and take note of the "Public IPv4" address.
The algo has some software dependencies that must be satisfied through software updates.
Start an SSH session, using a client such as Windows PuTTY, and connect to the instance as "root" using the root password saved from the Equinix Metal console.
Run the following commands to satisfy the software depenencies for the algo.
sudo add-apt-repository ppa:ubuntu-lxc/lxd-stable
sudo apt-get update
sudo apt-get -y upgrade
sudo apt-get -y install golang git
Alpaca credentials (API Key & Secret) are required for the algo to talk to the Alpaca APIs. An API key and corresponing key can be generated via the Alpaca console https://app.alpaca.markets/paper/dashboard/overview. The "paper" endpoint will be used for testing (not live) of the algo.
The API key and corresponding secret can be entered as environment variables using the commands below. The algo will read these environment variables when started and use them to authenticate to Alpaca
export APCA_API_KEY_ID="YOUR_KEY_ID_HERE"
export APCA_API_SECRET_KEY="CORRESPONDING_SECRET_HERE"
With all the dependencies in place, the algo can be started up with the following command. The algo will connect to Alpaca, authenticate, and start streaming down trading inforation and identifying trades to execute. If the U.S. stock market is closed, the algo will dutifully wait until it opens and then start trading. Approximately 15 minutes before the U.S markets close (3:45pm ET), the algo will start closing out positions, ending the day with no positions open (long or short).
wget https://raw.githubusercontent.com/alpacahq/alpaca-trade-api-go/master/examples/long-short/long-short.go
go mod init long-short.go
go run long-short.go
Because resources are billed by the hours, the deployed metal should be torn down (released) when no longer needed. Simply navigate back to the Equinix Console and "Delete" the deployed instance. It can be recreated at a later time to re-run the algo.
This is a basic algo run across a single metal instance without any external data analysis. From here, an algorithm can processes other data sources, such as audio, video, press releases, Twitter, weather, etc to build its trading model. This will undoubtably require additional compute power requiring more larger and more complex infrastructure - all of which can be provided via Equinix Metal.