Why isn't TCP good for large data transfer today?
TCP suffers from problems attributable to its congestion control and flow control design decisions. A lot of these decisions were made in the the 70's/80's when wireless didn't really exist and high bandwidth connections weren't readily available.

The main issues with TCP's flow control are head of line blocking and receive buffers that are too small.

The main issue with TCP's congestion control algorithm is its additive increase/ multiplicative decrease (AIMD) scheme heavily penalizes lossy networks. This means in practice even a small amount of packet loss in a network can seriously degrade performance.
What about the network of servers?
To reduce high roundtrip time, we are offering machines closer to your end user. You can think of this as a reverse CDN or a content ingestion network. At the moment we have a few regions but will be building this out as demand increases.

Other Questions?


Read additional information about how Tachyon works

Read docs >

Reach out

Can’t find your answer? Get directly in touch with our team.

Contact us >