You think that it's a problem that can be solved, but it's actually quite difficult to get to a level of accuracy that you want. It is an axiom in computer science that you should never build a system based on clock time. The $21 million Series A funding round announced by Clockwork.io will change this with sync accuracy as low as 5 seconds with hardware timestamps and hundreds of seconds with software timestamps.
The company is launching its first product today, which will give its users extremely fine-grained latency data in their cloud, on-premises and hybrid environments, which they can then use to find bottlenecks and tune their networks. The company's customers include the likes of Wells Fargo.
The image is called Clockwork.
The startup was founded by Yilong Geng, a computer science professor at the University of California, Berkeley. The core research behind Clockwork's system is based on academic research done by the team.
The Network Time Synchronization Protocol is the standard format that most computers use, but it is not very accurate. The Clockwork team promises far greater accuracy, despite some work on improving that, with Facebook contributing a hardware solution to the Open Compute Project last year.
I couldn't get them to agree on a second. The base station and my phone agree on the second. Then you get down to the microseconds. That is very difficult. It is hard for two clocks to know what time it is. It's not good enough to sync these clocks once. You have to keep them on track. If you want to put a high-accuracy clock into a server, you have to pay more for it than the server itself.
The image is called Clockwork.
The team built a system and machine learning model that can accurately measure the time it takes for a timestamp to arrive at a server. The team takes this a step further by looking at a variety of timestamps and then getting both the offset of the clock and the relative frequencies. This feeds into the machine learning model. The team built the system so the different clocks can communicate when they are not synchronized.
In the absence of trustworthy timestamps, distributed systems have had to rely on clockless designs, which adds an extra level of complexity to building complex systems. The Clockwork team hopes that its work will allow researchers to experiment with new time-based algorithms across a number of problem areas like database consistency, event ordering, consensus protocols and ledgers.
If you could trust the clocks in a distributed system, the original research was all about what you could do.
Currently, nobody uses time except for someone doing database things, or for someone doing something else. We figured out how to do time sync. We asked if this was part of a trend where we're going to start programming these systems differently. Researchers were excited about the possibility of us being able to pull this off.
The Clockwork team is looking to build products on top of the issues that have been solved. The team is working on a project that will make it easier to detect congestion in data centers. He noted that the data center uses a lot of TCP, but it is great for wide-area networks. If you know more about the network and its latencies, you can give the protocol better hints about how to best route packets inside the data center.
The company's Series A round was led by NEA, with participation from well-known angel investors, including MIPS co-founder John Hennessey.
All Rights Reserved © 2024
