How can you handle latency and jitter in industrial communication networks?

Start with an honest assessment of the application's requirements related to timing e.g. jitter, bandwidth, desired cycle, etc.) and use a protocol that makes sense based on that. Not all applications required stringent real-time communications and that can open up choices of suitable protocols.

Any network is susceptible to cabling issues. Retries may silently solve your problems, at the expense of less available bandwidth and more jitter, up to the moment there's so many retries (may take years to develop) that maximum response times are exceeded and applications start failing or break.

Projetar uma arquitetura confiável e robusta requer muita aten??o nos detalhes dos produtos aplicados, pois um simples device que n?o possui todos os requisitos necessários pode prejudicar a performance desejada da sua rede em em casos extremos tornar a ideia inicial inviável. Para redes entre PLC e remotas de I/O, sempre que possível utilizo o Profinet em meus projetos, apesar de ser um protocolo muito flexível, pontos sensíveis devem ser observados com cuidado: quantidade de devices, limite de devices do controlador da rede, suporte a MRP, velocidade dos dispositivos (Fast ou Gigabit Ethernet), possibilidade de inclus?o de switches como manager do anel principal, inclus?o de sub anéis para aumentar a disponibilidade do sistema e etc.

You don't choose the protocol, you choose the controls OEM and the controls come with it. Latency or jitter can be due to comms issues or to controls CPU issues I.e. Comms will suffer if: ??Network topology is star and loops arise ??New devices are added to the network and broadcast is not well solved. Special attention to new devices with IP address out of segment range ??Changes happen in part of devices addresses and others start to drop packets ??New services are incorporated and bandwidth is saturated ??Isolated networks are connected and segment address ranges overlap CPU in control devices can cause issues if: ?? Encryption is incorporated but hardware is not designed for it ?? More routines added that the available for one cycle

Protocol selection is important factor. But there isn't the best network , every network Protocol will have pros and cons. For low latency and real-time performance always select industrial Protocols.

Software is often a big spoiler w.r.t. latency. The network may be fast, but the software speed is varying, within the same device and even more so between different vendors. Why bother about nsec's when software counts in msec's.

Adding network security devices or countermeasures such as IPS, IDS , firewalls , WAF can cause latency , especially if there is deep inspection, choosing network security devices with high throughput and low latency is the best option

Optimizing industrial communication network design involves prioritizing critical data, implementing redundancy for continuous communication, and isolating traffic types. This approach ensures efficient and reliable real-time performance by managing bandwidth, establishing predictable routing, and addressing scalability and security considerations. A well-designed network minimizes interference and delays, enhancing overall system efficiency in industrial automation.

All valid actions to mitigate the risks associated. Ethernet networks above 80 percent utilisation can suffer from seemingly exponential latency. The other element not mentioned is the control design itself. Ensuring time critical control and safety applications have logic closer to the I/O mitigates more risks associated with network connectivity and availability.

Even beter select the right medium , fiber is faster then satellite, copper is faster then microwave radio. Try and select a faster connection types where available.

QoS only comes into play if your link is congested….. if that’s the case upgrading is much beter then trying to use end to end QoS to reduce jitter. Latency can’t be effectively reduced using QoS.

Ethernet TSN (time-switched networking) is coming your way. But progress is very slow. Wonder whether the market is really waiting for it, or whether this is a new technology waiting for a problem.

Upgrading bandwidth can mitigate link congestion but it's not always viable. QoS can play a crucial role, especially in environments where upgrading infrastructure is not feasible due to cost or other constraints. QoS can effectively prioritize critical traffic, ensuring key applications perform reliably even under congestion. OT processes depend on real-time data & QoS helps minimize latency & jitter for time-sensitive traffic, crucial to the integrity of these systems. QoS won’t reduce latency inherent to the network, it can manage latency caused by congestion, which is important in real-time applications where jitter reduction is crucial. While bandwidth upgrades are beneficial, QoS remains a vital tool in optimizing network performance.

Routers have a quality of service (QoS) setting, which lets you prioritize packets over other types of traffic. Therefore, implementing it can be lifesaving. However, use packet prioritization only if you want to prioritize internet calls over other traffic. It lowers jitter, allowing calls without interference.

This section underscores the importance of Quality of Service (QoS) mechanisms in managing industrial communication network traffic effectively. QoS techniques prioritize messages based on importance and urgency, thereby reducing latency and jitter for critical messages. Examples include traffic shaping, scheduling, policing, and filtering, each addressing specific aspects like preventing overload, assigning priorities, enforcing policies, and controlling traffic based on criteria. Choosing and implementing suitable QoS mechanisms aligned with protocol and network device capabilities can significantly improve overall network performance and reliability.

Never, ever plan on buffering. Buffering kills every chance of properly analyzing issues and carefully planning the process. Switches that buffer transmissions should be configured to avoid that. Buffering only makes sense, where information leaves the industrial network or it is received. But again, it needs to be carefully planned and not dependent on running the process. TSN is the future, but adaption is very slow. So far it may solve most issues, but it will not solve poorly planned networks

This can be a costly affair, Since you are going to need array of managed switches. Layer 2 and layer 3 switches with TSN networks will definitely work, but look at their prices.

And good luck finding people to help with integration ?? I still remain unconvinced that carting such a heavy stack of software and highly specialized silicon is the best way forward. Do you really need one stack to be compatible across the entire IOT space, or are you creating a solution in search of a problem? I guess time will tell, and so far it’s telling us:”meh”!

Distributed system design can greatly improve network performance. Local traffic to be consumed locally and not to overload highway traffic. This is the key I have implemented in DC BMS designs. Mapping of sources and destination of data can greatly improve system performance.

Understand the application, BEFORE you start planning the network. Do we really need those 800μs cycle time or will 2ms suffice, because we are using analog io that does not provide the information any faster? Which components will be used and what is their calculated or determined delay/latch time? Which application part will use which type of information transfer? Synchron vs. Asynchron? Do we need to distribute data and if yes, can we aggregate it?

Modern standards based deterministic networks such as Ethernet TSN (time sensitive network), Wi-Fi with TSN and private 5G networks with TSN grade synchronization and determinism should be considered for your network infrastructure.

While it may seem simple it can be easily overlooked. By updating firmware on routers, switches, and servers can be crucial for network stability. Firmware updates can provide enhanced features and performance improvements. Manufacturers continuously optimize their software to improve the speed, reliability, and functionality of their devices. Updates can resolve bugs or issues that were present in previous versions. These could be causing device instability or other operational problems. Therefore, keeping your firmware up-to-date is an essential part of maintaining a secure and efficient network.

Latency can be improved by: In profinet or network based in Ethernet: 1. Load balancing of the data between multiple servers. 2. Optimizing the data size. 3. Using the latest network drivers with highest baud rate. 4. Subnet the networks to small sizes. In profibus/modbus networks: 1. Less using of the repeaters. 2. Using the min length of physical layers. 3. Optimizing the data sizes. jitter can be improved by: 1. Reducing the interference or noise to avoid re-requesting of the corrupted data. 2. Reducing the acyclic data transmission either by class 1 or 2 masters. In IRT applications, you may need a reserved bandwidth for these data, so you can use the devices that supports IRT in profinet or DP V2 in profibus.