Coinbase, the U.S. cryptocurrency exchange, is using artificial intelligence to help prevent website downtime during busy traffic times.
They’ve rolled out a machine-learning model that predicts when user traffic will spike. This allows them to automatically allocate more server resources before any issues arise.
“Crypto markets can be unpredictable. On our platform, user activity can suddenly surge and then drop off just as quickly,” Coinbase explains. “We manage these changes by scaling up resources when traffic is high and scaling back down once things return to normal.”
Recently, this model proved its worth when the crypto market saw increased volatility. As traffic rose, Coinbase doubled its scale target twice just hours before peak times. The model adjusted continuously to daily patterns until the volatility eased.
Coinbase’s machine-learning model looks at platform load and external signals, like price changes for Bitcoin (BTC) and Ethereum (ETH). This helps determine when to bring in extra server resources. “If cryptocurrency prices are fluctuating a lot and traffic is rising faster than we expected, that’s a sign a spike is likely,” they noted.
In the last cycle, Coinbase faced significant outages after user traffic surged by about 400%. At that time, they committed to finding new solutions to minimize downtime.