While a precise time synchronization infrastructure provides critical support for a variety of applications within a data center, it cannot deliver reliable service without robust monitoring and reporting tools. These tools ensure that the precise timing infrastructure is working per the designed specifications and meeting the Service Level Agreement (SLA) as it relates to synchronization accuracy and availability. Equinix has implemented extensive monitoring and reporting tools to support our Equinix Precision Time™ (EPT) service, ensuring that the accuracy and availability obligations are met. A new independent technical study confirms that we are exceeding our SLAs, and we have just released a new feature that both increases availability and supports virtual connections to EPT.
Financial institutions that are required to comply with FINRA regulatory compliance rules need to synchronize their entire infrastructure to within 50 milliseconds. Organizations that are required to meet MiFID II compliance requirements have an even more stringent synchronization accuracy requirement of 100 microseconds. Equinix has deployed extensive monitoring and reporting tools across all installations of our EPT hardware timing system infrastructure worldwide to monitor, report and generate alerts on the timing accuracy levels of the system as well as any failures in real time. In addition, we monitor synchronization accuracy delivered to the various metros across all the regions through our extensive SLA monitoring infrastructure using Equinix Fabric, a highly secure private interconnection network.
Deliver secure, reliable and precise time synchronization
Equinix Precision Time provides a global, highly accurate, and reliable Time-as-a-Service application for distributed systems over Equinix Fabric.Read More
Monitoring tools in over 14 metros worldwide
Our extensive global footprint with SLA monitoring tools ensures that the customers using our Precision Time service in various metros worldwide get access to precise, secure, and reliable time synchronization in the data center where they are operating. These SLA monitoring tools are deployed in over 18 metros supported by more than 40 clients and we are constantly expanding the deployment to other metros to ensure the reliable time service that our customers count on. The example below depicts a typical SLA monitoring set-up between New York and Silicon Valley.
Independent study validates that EPT exceeds SLA
In a recent study, we contracted an independent company specializing in precise time and frequency synchronization to analyze the SLA monitoring data from 18 different metros across all regions. We asked them to report how well the Precision Time service is performing against the accuracy levels in our SLA for PTP service.
We continuously monitored and logged the time synchronization data at various locations using two types of PTP clients: a standard compute server with a PTP-enabled smart network interface card and a purpose-built PTP client with highly optimized servos for clock recovery. While both types of client devices performed well, we observed that the purpose-built client device with the optimized servo performed better. There was no need to apply a post filtering with a 0.1Hz filter to remove the noise unlike the compute server operating with a PTP-enabled NIC card, which needed the filter to remove high-frequency noise. We also analyzed the Network Time Protocol (NTP) synchronization accuracy as part of this study.
Our extensive global footprint with SLA monitoring tools ensures that the customers using our Precision Time service in various metros worldwide get access to precise, secure, and reliable time synchronization in the data center where they are operating.”
The analysis calculated max time error for all metros to assess the synchronization accuracy from SLA monitoring data. The New York to Silicon Valley SLA was by far the most challenging due to the long distance involved between the timing source and the client. We observed a max time error of 3 microseconds at the client located in New York connecting locally to the time source in New York. The client in Silicon Valley synchronizing to the time source in New York had a maximum time error of 9.2 microseconds. We also observed that the probability distribution of the error had a “long tail.” Specifically, less than 2% of the monitored data was greater than 75% of the max time error calculated, which means the clients in both New York and Silicon Valley were performing much better than the max time error observed for 98% of the time. The observation from this study points to the fact that the Equinix Precision Time Service accuracy levels for PTP perform much better than the 50-microsecond accuracy level guaranteed in our SLA. Our NTP SLA monitoring data observed a max time error in the range of 30-50 microseconds across all metros, which is also well within the acceptable range of a typical NTP deployment. The robust timing architecture and a low latency/low jitter Equinix Fabric interconnection make these accuracy levels for the Precision Time Service possible.
New York to Silicon Valley PTP SLA Monitoring Results
New York to New York (Within Metro) PTP SLA Monitoring Results
Qulsar CTO validates EPT approach for time synchronization
The purpose of synchronization over the network is to reproduce the time reference, necessitating the availability of the time reference. This begs the question of if the reference is available, then why is there the need to synchronize the clock over the network? Dr. Kishan Shenoi, CTO of Qulsar, Inc. observes: “The practical solution to this conundrum is to have a good PTP client with a clock recovery servo algorithm that provides reliable estimates of clock error excursions such as peak-to-peak error or maximum relative time error. A PTP client cannot estimate absolute error because it is blind to constant time error arising from network asymmetry. However, a well-designed, well monitored, well managed network lends itself to requiring just routine calibration of asymmetry. Continuous monitoring and a good servo serve to fill in the gaps with a high level of confidence. This is the approach taken by Equinix Precision Time and it is a good one.”
This study points to the fact that the Equinix Precision Time Service accuracy levels for PTP perform much better than the 50-microsecond accuracy level guaranteed in our SLA.”
Dr. Shenoi also acknowledged the important role that Equinix Fabric plays in EPT. “Even though [Equinix Fabric] is PTP-unaware, it supports PTP-based synchronization with peak-to-peak accuracy of single-digit microseconds,” he said. “Control of the routing ensures that changes in asymmetry, as routes are changed, is small. With such a well-behaved network, measurement of asymmetry across the wider network provides reliable upper bounds on synchronization quality, including constant time error, over sub-networks. EPT has geared for that by establishing Primary Reference Time Clocks in suitable locations.”
New features increase availability and support virtual connections
We are constantly expanding and enhancing our Precision Time Services across the globe. In addition to beefing up our monitoring and reporting, we continue to expand our geographic coverage of the service. We have just released a Fabric port redundancy service option to improve the availability SLA to 99.999% and also added the capability to connect to the Precision Time service through Network Edge, which is a NFV platform for virtual network services. You can learn more about some of these enhancements in this brief spotlight video. Equinix Precision Time is an excellent choice for your mission-critical applications in the data center or cloud deployment. Take advantage of a Free Trial to try EPT and see for yourself.
A well-designed, well monitored, well managed network lends itself to requiring just routine calibration of asymmetry. … This is the approach taken by “Equinix Precision Time” and it is a good one.”