Google Cloud customers will get faster network speeds as part of an update the company is making across the board to support new technology that reduces network congestion. That tech is what helped accelerate speeds for YouTube, google.com, and other services the company runs.
The algorithm driving the change is called Bottleneck Bandwidth and Round-trip propagation time, or BBR for short. It accelerates networking speeds by more intelligently deciding to slow down or speed up data transfer. Past algorithms relied on tracking packet loss to see when they needed to slow down transfer, which isn’t the best choice for modern networks.
Google tackled that issue from the ground up with BBR, which builds a model based on the maximum bandwidth along a network connection, as well as the recent round-trip time. With that information, the algorithm decides how quickly it can send information, which should be faster than previous approaches.
Customers accessing Google services like Cloud BigTable, Cloud Spanner, or Cloud Storage will see traffic to and from their computers accelerated by the new algorithm. In addition, cloud customers’ applications that use Google Cloud Load Balancing or Google Cloud CDN will also benefit. That means end users of a customer’s application will see their performance improve as well, without any extra work on the part of developers.
Google claims that BBR can transfer data up to 2,700 times faster than previous versions of TCP on a 10Gbps network link with 100ms latency and 1 percent packet loss.
The benefits can be fairly significant, if Google’s experience is any guide. When YouTube implemented BBR, that site saw a 4 percent improvement in network throughput and a 33 percent reduction of round-trip time. While those may not seem like massive gains, YouTube’s network performance was already highly optimized, so even small improvements are significant for that service at this point.
The audio problem: Learn how new cloud-based API solutions are solving imperfect, frustrating audio in video conferences. Access here