Globally Synchronized Time via Datacenter Networks

Abstract

Synchronized time is critical to distributed systems and network applications in a datacenter network. Unfortunately, many clock synchronization protocols in datacenter networks such as NTP and PTP are fundamentally limited by the characteristics of packet-switched networks. In particular, network jitter, packet buffering and scheduling in switches, and network stack overheads add non-deterministic variances to the round trip time, which must be accurately measured to synchronize clocks precisely. We present the Datacenter Time Protocol (DTP), a clock synchronization protocol that does not use packets at all, but is able to achieve nanosecond precision. In essence, the DTP uses the physical layer of network devices to implement a decentralized clock synchronization protocol. By doing so, the DTP eliminates most non-deterministic elements in clock synchronization protocols and has virtually zero protocol overhead since it does not add load at layer-2 or higher at all. It does require replacing network devices, which can be done incrementally and with very small amount of hardware resource consumption. We demonstrate that the precision provided by DTP in hardware is bounded by 4TD where D is the longest distance between any two nodes in a network in terms of number of hops and T is the period of the fastest clock. The precision can be further improved by combining DTP with frequency synchronization. By contrast, the precision of the state-of-theart protocol (PTP) is not bounded: The precision is hundreds of nanoseconds in an idle network and can decrease to hundreds of microseconds in a heavily congested network.

Revisiting Time, Clocks, and Synchronization

“…DTP [22] and Sundial [11] modified the physical layer hardware to send timestamp packets more frequently than PTP. However, they cannot surpass PTP with SyncE enabled, which synchronizes the quartz crystal oscillators at the Ethernet carrier frequency, and can further disciplines them with a more stable clock.…”

“…Recently, there is work to get the accurate time for each computer by distributing the GNSS signal [23]. There are also some works claimed to achieve high-precision synchronization by the SVM machine learning algorithm (Huygens [8]), by the highperformance network (FARMv2 [20]), or by modifying the physical layer of the Ethernet (DTP [22], and Sundial [11]), even without any atomic clock or GNSS clock.…”

Phase-Stabilized Side-Branch RoF Link for Extensible Frequency Dissemination in Distributed Systems

“…REQUENCY synchronization is the key enabling technology in modern large-scale distributed information systems and applications like time-sensitive networking for 5G fronthaul and backhaul [1][2][3][4], indoor/outdoor positioning [5,6], carrier recovery in optical communications [7,8], heterodyne receivers in coherently netted radars and radio telescopes [9][10][11][12], and data center interconnections [13,14], etc., where the disseminated and recovered frequency signals are usually essential either for carrier (local oscillator, LO) synchronization [6][7][8][9][10][11] or for precise and fine-grained time control [2,13]. The required synchronization accuracy can be at sub-nanosecond level or beyond [1,5], for decimeter/centimeter positioning, etc.…”