IBM z17 Time Synchronization Resiliency Enhancements

Oct 29, 2025

Steve Guendert, PhD, (aka “Dr. Steve”) is the chief architect for IBM Z and LinuxONE time synchronization technology. He represents IBM in several time synchronization standards development organizations and projects, including the ITU, IEEE, and IETF. He has 48 patents in time synchronization and IBM Z I/O technology. He has served on both the SHARE and CMG Boards of Directors. 

Most likely, you have heard the generic acronym RAS, which typically stands for Resiliency, Availability, and Serviceability. However, in the world of time synchronization at IBM, we changed RAS to mean Resiliency, Accuracy, and Security.

From RAS to IBMz17

Timing, timekeeping, time synchronization, and, more specifically, accurate synchronization are key requirements for modern IT systems. This is especially true for industries involved in transaction processing, such as the financial sector.

This need for accuracy is why the IBM Z sysplex relies on highly precise timing and synchronization technology to ensure data integrity and enable database reconstruction from logs. To achieve this, IBM Z uses the best oven-controlled crystal oscillators (OCXOs) in the industry.

But in 2025, it’s not enough. We also need tremendous resiliency and security to maintain those levels of accuracy.

IBM z17 introduced several important time synchronization enhancements that improve the security and resiliency of a parallel sysplex environment.

Enter IBM z17. The IBM z17 introduced several important time synchronization enhancements that improve the security and resiliency of a parallel sysplex environment. These updates help end users maintain the accuracy required to comply with government and industry regulations.

Background: The Evolution of IBM Time Synchronization

Here is a brief overview of IBM’s Z time synchronization evolution.

For the past two decades, time synchronization in IBM Z has centered around Server Time Protocol (STP). STP is IBM’s proprietary, message-based protocol that allows a collection of connected mainframes (a parallel sysplex) to maintain synchronized time known as Coordinated Server Time (CST).

This network of synchronized IBM Z machines is called a Coordinated Timing Network (CTN).

However, STP does not synchronize a sysplex with the outside environment. That function relies on a different protocol. 

From 2007 to 2019, that protocol was the Network Time Protocol (NTP). Starting with the IBM z15 in 2019, the Precision Time Protocol (PTP) (IEEE 1588) became a second option. Now in 2025, there’s a new option.

New on IBM z17: Enhanced Time Synchronization

NTP resiliency refers to the network’s ability to maintain accurate time synchronization despite network issues or failures. To improve overall resiliency, IBM z17 introduced two new components:

  1. Support for NTPv4/Chrony – Improves accuracy and stability by leveraging the full suite of NTP algorithms through Chrony.
  2. Mixed Mode Operation (NTP + PTP) – Increases resiliency and stability by allowing STP to use up to five external reference sources (three NTP and two PTP) simultaneously.

The rest of this article focuses on the NTPv4/Chrony support.

NTPv4 and Chrony: A Smarter, More Accurate Approach

NTP has existed since 1985 and remains one of the most common Internet protocols. Even with the z15’s PTP support, NTP continues to serve as the preferred external time reference for many IBM Z customers.

Therefore, IBM continues to enhance its NTP implementation, most recently with support for NTPv4/Chrony on the z17.

  • NTPv4 is defined by IETF standard RFC 5905 and is backward compatible with NTPv3 (RFC 1305).
  • It adds IPv6 compatibility and algorithmic improvements that can achieve accuracy within tens of microseconds.
  • Chrony, a newer NTPv4 implementation, performs well in congested networks—achieving millisecond accuracy over the Internet and tens of microseconds on a LAN.

Chrony achieves this by using hardware timestamping, similar to PTP, rather than the software timestamping of standard NTPv4.

Chrony gives IBM Z systems the best of both worlds:
NTP reliability and modern precision.

In short, Chrony gives IBM Z systems the best of both worlds: NTP reliability and modern precision.

How IBM z17 Improves NTP Resiliency

For IBM z17 and later, you can configure up to three NTP servers per IBM Z system in an STP-only CTN (z16 and earlier were limited to two).

Key Definitions

  • Truechimer: A clock that maintains time accuracy according to a trusted standard such as UTC.
  • Falseticker: A clock that fails to maintain that accuracy due to error, fault, or malicious interference.
  • Candidate: An association that has valid peer variables.

Each NTP server operates through two processes:

  1. Peer Process – Receives and validates each packet. Invalid packets are discarded; valid ones are passed to the Clock Filter algorithm. Optionally, an Access Control List (ACL) can verify IP address entries for added security.
  2. Poll Process – Sends packets at programmed intervals.

These processes, together with their peer state variables, form what’s known as an association. As packets arrive, NTP compares the server’s time to the system clock, calculates offsets, and applies corrections through the NTP discipline process.

Inside the Algorithms: How It All Works

There are four algorithms at work.

Clock Filter Algorithm

  • It selects the best sample data and rejects noise caused by collisions or congestion.
  • Filters out transient “popcorn spikes.”

Select Algorithm

It determines which sources are truechimers and which are falsetickers.

  • Uses multiple redundant servers and network paths.
  • Checks NTP stratum level, root distance, and source reachability.
  • Identifies truechimers via correctness intervals, then passes them to the Cluster algorithm.

Cluster Algorithm

  • It ranks truechimers by evaluating peer jitter and selecting jitter to determine which sources deliver the highest overall accuracy.
  • Produces a list of “survivors” from most to least favored.

Combine Algorithm

  • Produces a weighted average of offset and jitter from surviving sources.
  • Weights are based on the reciprocal of each source’s root distance.
  • These normalized weights sum to one.
  • The combined offset synchronizes the NTP server’s system clock.

Putting It All Together: More Resilient Timekeeping

IBM z17 introduced multiple significant enhancements to its NTP implementation. When configured correctly, these changes enhance IBM z17’s synchronization accuracy to UTC and create a more resilient implementation than was possible on prior generations.

Learn More

For more details, refer to the following sources or contact the author directly at steve.guendert[at]ibm.com.

References

0 Comments

Submit a Comment

Your email address will not be published. Required fields are marked *

Sign up to receive the latest mainframe information

This field is for validation purposes and should be left unchanged.

Read More

Inventor, Mother, Creator: Grace Hopper

Inventor, Mother, Creator: Grace Hopper

Grace Hopper is often touted as the inventor of COBOL, and for good reason. In 1955, she created the definition of a data-processing compiler and many of the rules that still govern COBOL today. She also created the first working data-processing compiler, FLOW-MATIC....

Observability and Resiliency Trivia

Observability and Resiliency Trivia

At Planet Mainframe, September is the month that we pay special attention to two critical pillars of modern mainframe operations — observability and resiliency.  Observability goes beyond traditional monitoring by providing deep insights into applications,...