Insight and analysis on the data center space from industry thought leaders.
In-Sync: The Crucial Role of Timing in Data Centers
Synchronized timing is paramount for data centers. Discover key considerations for operators as data environments evolve.
April 3, 2024
It is a marvel at how quickly and consistently data centers transmit information to and from locations. Servers in data centers are communicating with each other millions of times per second, processing critical transactions that must be precisely timed. Computers have internal clocks to keep track of timing, but these clocks are constantly drifting in relation to each other. If mechanisms aren’t put in place to continuously synchronize the internal clocks, there is an increased likelihood of data corruption or loss due to these discrepancies.
The importance of synchronized timing has always been prevalent in data centers, but the concern is extrapolated now that computing environments are more dispersed, interconnected, and complex than they have ever been. In addition, increasing cloud usage means more data centers are being built to support it. This demand has led to more facilities in remote locations with challenging connectivity environments. To account for these changes, data center operators need to consider how to improve timing synchronization resiliency and ensure data integrity.
What Is Data Center Timing and How Does It Work?
Network Time Protocol (NTP) and Precision Time Protocol (PTP) are two network-based standards developed to help update and synchronize computer internal clocks with Universal Time Coordinated (UTC), our global timing standard.
However, these timing standards are dependent on packet-based randomly routed Internet connections to communicate with a time server. They are also susceptible to network jitter issues commonly associated with Ethernet transmission.
Network time standards are often satisfactory, but higher precision is often necessary for high-speed communications that are becoming commonplace in modern computing environments. Beyond the network-delivered NTP and PTP just mentioned, other approaches to provide exact timing in a data center include GNSS receivers and ultra-stable clocks. They all have pros and cons, but generally, GNSS is considered the best combination of accuracy and cost-conscious scalability.
Global Navigation Satellite System (GNSS) refers to a constellation of satellites providing signals from space that transmit timing data to GNSS receivers. Global Positioning Systems (GPS) is one example of GNSS. It uses highly precise atomic clocks as a foundation in its architecture, and each satellite in these systems simultaneously transmits a precise time signal. Those signals are used to triangulate the position of a receiver.
The timing info delivered has several advantages over NTP including being available anywhere with a clear view of the sky and isn’t reliant on an internet connection to receive timing data. The largest hurdle for GNSS is the requirement for a clear view of the sky, made worse by data centers built in remote locations near mountains, but fiber-optic technology can help overcome this obstacle.
How Can Fiber Optics Solve the Growing Time Synchronization Problem?
GNSS communicates with the data centers through radio frequencies (RF) between 1.1GHz to 1.6GHz. In most data centers, timing distribution systems use coaxial cables to transmit data from the antenna to its servers. When the distance from the antenna on the data center to the GNSS receiver exceeds 30-50m, the signal attenuates to a level that is unusable.
Leveraging fiber-optic cable improves the resilience of communication as it attenuates several orders of magnitude less than coaxial cable. This makes an RF over fiber (RFoF) network much more efficient and can deliver GPS timing to an entire data center from a small number of antenna sites.
For example, standard RFoF architectures can deliver a usable GPS timing signal to 500+ endpoints from a single antenna. This provides a parallel timing signal to be available to augment and supplement NTP and PTP timing signals from the Internet. Once the RF signal is converted to an optical signal, it can be passively split via a fiber-optic splitter for delivery to multiple receivers.
A second component of ensuring a proper GPS timing distribution solution is deploying double redundancy. Instead of deploying a singular optical receiver with an RF switch, data centers can install an additional backup that picks up the signal should failure occur.
This type of fully redundant GPS-over-fiber architecture provides seamless, reliable, and future-proof integration of analog GPS timing. It also eliminates any single point of failure in the system to ensure that timing standards remain operational even in the event of a hardware failure.
To take it a step further, it is also possible to leverage network management systems across all data centers to monitor all GPS timing solutions in real time from a single remote location.
As the world relies more on precisely timed transactional interchange between computers and data centers, there is a heightened focus on the dependability and reliability of the timing synchronization. It is fundamental for maintaining the reliability, security, and performance of data centers, especially in today's increasingly complex and interconnected computing environments.
Meir Bartur, Ph.D, is the President and CEO of the Optical Zonu Corporation. Dr Bartur has more than 30 years of experience in leadership, product development, and technology innovation. As a Senior Member of the IEEE and recognized leader in the development of low-cost fiber-optic solutions for FTTx, he contributed to the IEEE ITU PON standards. Before founding Optical Zonu, he directed Advanced Product Development and Strategic Technology for access transceivers at MRV Communications, as well as business relations with its major clients.
About the Author
You May Also Like