The Future Network: Unlocking the Potential of Training Super-Large-Scale AI Models

From Transformers to the widespread adoption of ChatGPT in 2023, people have come to realize that increasing model parameters can enhance performance, aligning with the scaling law of parameters and performance. Particularly, when the parameter scale exceeds trillions, the language comprehension, logical reasoning, and problem-solving capabilities of large AI models improve rapidly.

To meet the demands of efficient distributed computing in large-scale training clusters, the training process of AI models typically involves various parallel computing modes such as data parallelism, pipeline parallelism, and tensor parallelism. In these parallel modes, collective communication operations among multiple computing devices become crucial. Therefore, designing efficient cluster networking schemes in large-scale training clusters of AI models is key to reducing communication overhead and improving the effective computation-to-communication time ratio of GPUs.

Challenges in Scaling GPU Networks for Efficient Training of Ultra-Large AI Models

The computing demands of artificial intelligence applications are experiencing exponential growth, with model sizes continuously expanding, necessitating significant computational power and high memory requirements. Appropriate parallelization methods such as data, pipeline, and tensor parallelism have become key to improving training efficiency. Training extra-large models requires clusters containing thousands of GPUs, utilizing high-performance GPUs and RDMA protocols to achieve throughputs of 100 to 400 Gbps. Specifically, achieving high-performance interconnection among thousands of GPUs poses several challenges in terms of network scalability:

  • Challenges encountered in large-scale RDMA networks, such as head-of-line blocking and PFC deadlock storms.
  • Network performance optimization, including more effective congestion control and load balancing techniques.
  • Issues with NIC connectivity, as individual hosts are subject to hardware performance limitations. Addressing how to establish thousands of RDMA QP connections.
  • Selection of network topology, considering whether to adopt traditional Fat Tree structures or reference high-performance computing network topologies like Torus or Dragonfly.

Optimizing GPU Communication for Efficient AI Model Training Across Machines

In AI large-scale model training, GPU communication within and across machines generates significant data. With billions of model parameters, communication from parallelism can reach hundreds of GB. Efficient completion relies on GPU communication bandwidth within machines. GPUs should support high-speed protocols to reduce CPU memory copying. PCIe bus bandwidth determines if network card bandwidth is fully utilized. For example, with PCIe 3.0 (16 lanes = 16GB/s), if inter-machine communication has 200Gbps bandwidth, network performance may not be fully utilized.

Crucial Factors in AI Large-Scale Model Training Efficiency

In data communication, network latency comprises two components: static latency and dynamic latency. Static latency includes data serialization, device forwarding, and electro-optical transmission delays, determined by the forwarding chip’s capacity and transmission distance, representing a constant value when network topology and data volume are fixed. In contrast, dynamic latency significantly affects network performance, including queuing delays within switches and delays caused by packet loss and retransmission typically due to network congestion. Besides latency, network fluctuations introduce latency jitter, affecting training efficiency.

Critical for Computational Power in Large-Scale AI Model Training

Cluster computing power is crucial for AI model training speed. Network system reliability forms the foundation of cluster stability. Network failures disrupt computing node connections, impairing overall computing capability. Performance fluctuations may decrease resource utilization. Fault-tolerant replacement or elastic expansion may be necessary to address failed nodes during training tasks. Additionally, unexpected network failures can lead to communication library timeouts, severely impacting efficiency. Therefore, obtaining detailed throughput, packet loss, and other information is vital for fault detection.

The Role of Automated Deployment and Fault Detection in Large-Scale AI Clusters

The establishment of intelligent lossless networks often relies on RDMA protocols and congestion control mechanisms, accompanied by a variety of complex configurations. Any misconfiguration of these parameters can potentially impact network performance and lead to unforeseen issues. Therefore, efficient and automated deployment can effectively enhance the reliability and efficiency of large-scale model cluster systems.

Similarly, in complex architectural and configuration scenarios, timely and accurate fault localization during business operations is crucial for ensuring overall business efficiency. Automated fault detection aids in quickly identifying issues, notifying management accurately, and reducing costs associated with issue identification. It can swiftly identify root causes and provide corresponding solutions.

Large-scale AI models have specific requirements in terms of scale, bandwidth, stability, latency/jitter, and automation capabilities. However, there still exists a technological gap in current data center network configurations to fully meet these requirements.

Al Intelligent Computing Center Network Architecture Design Practice

Traditional cloud data center networks prioritize north-south traffic, leading to congestion, high latency, and bandwidth constraints for east-west traffic. For intelligent computing scenarios, it’s recommended to build dedicated high-performance networks to accommodate workloads, meeting high-bandwidth, low-latency, and lossless requirements.

Based on current mature commercial switches, it is recommended to consider different models of InfiniBand/RoCE switches and the supported GPU scale to set the following specifications for physical network architecture:

Standard: Based on InfiniBand HDR switches, a dual-layer Fat-Tree network architecture supports up to 800 GPU cards per cluster.

Large-scale: Based on 128-port 100G Ethernet switches, a RoCE dual-layer Fat-Tree network architecture supports up to 8192 GPU cards per cluster.

Extra-large: Based on InfiniBand HDR switches, an InfiniBand three-layer Fat-Tree network architecture supports up to 16000 GPU cards per cluster.

Extra-extra-large: Based on InfiniBand Quantum-2 switches or equivalent Ethernet data center switches, adopting a three-layer Fat-Tree network architecture supports up to 100000 GPU cards per cluster.

In addition, high-speed network connections are crucial for ensuring efficient data transmission and processing.

How FS Can Help

FS provides high-quality connectivity products to meet the demands of AI model network deployment. The FS product portfolio includes (200G, 400G) InfiniBand switches, data center switches (10G, 40G, 100G, 400G) network cards, and (10/25G, 40G, 50/56G, 100G) optical modules, accelerating AI model training and inference processes. Optical modules offer high bandwidth, low latency, and low error rates, enhancing data center network capabilities for faster and more efficient AI computing. For more information, please visit the FS website.

Switch Definition In Networking Explained

When it comes to networking technology, we won’t miss Ethernet switches. Because it is an essential part of networking communication. Connecting devices, such as computers, routers, and servers, together on a network, it enables current to be turned on and off and selects a channel for data transmission. Then you may ask, what is the switch definition in networking? What are the types of switches in networking? How to choose a switch for my network? Now, this passage will give you answers and suggestions.

What Is the Switch Definition In Networking

There is one question that confuses many people: what is a switch in networking? A switch, in the switch definition of networking, is high-speed network equipment used to connect devices together on a network and enable the data transmission between different devices. It receives incoming data packets and redirects them to their destination on a local area network (LAN).

In a local area network (LAN) using Ethernet, a network switch determines where to send each incoming message frame according to the physical device address. This kind of address is also known as the Media Access Control address or MAC address. If a switch needs to forward a frame to a MAC address that is unknown by the switch, then the frame is flooded to all ports in the switching domain. Generally speaking, a data switch can create an electronic tunnel between the source and the destination ports that no other traffic can enter for a short time.

Switch Definition In Networking: Types of Switches In Networking

The Ethernet switch is an essential part of any network. Generally speaking, the Ethernet switch can be classified into two categories: the modular switch and the fixed switch.

Modular Switch

The modular switch has expansion ability and high flexibility. Modular switch makes it possible for you to add expansion modules as needed into the switches. It is much more complex than fixed switch, so it costs more than fixed switch.

Fixed Switch

The fixed switch isn’t expandable and has a fixed number of ports. Although it has less flexibility, it offers a lower entry cost. There are mainly three types of fixed switches in networking. They are the unmanaged switch, the smart switch, and the managed switch.

Unmanaged Switch

The unmanaged switch is often used in home networks, small business offices or shops. It can’t be managed, so we can’t enable or disable interfaces of it. Although it doesn’t provide security features, it can offer enough support if you use it in a small network of fewer than 5-10 computers.

Smart Switch

The smart switch is mainly used for business applications such as smaller networks and VoIP. It is suitable for small VLANs, VoIP phones, and labs. Smart switch can let you configure ports and set up virtual networks but doesn’t have the ability to allow troubleshooting, monitoring, remote-accessing to manage network issues.

Managed Switch

The managed switch is widely used in data centers and enterprise networks. It provides control, high-levels of network security, and management. It’s ideal for remote-access control capabilities and off-site round-the-clock monitoring. The managed switches can improve a network’s resource utilization and speed. Although it costs the most, it worth the investment for a long run.

Switch Definition In Networking: types of fixed switches

How to Choose a Switch For Your Network?

When you choose a switch for your network, you need to consider several factors at the same time. These factors include the number of ports, transmission speed, and stackable vs standalone.

Number of Ports

Most of the switches on the market have 4 to 48ports. You need to consider the number of ports you’ll need according to the number of users and devices and devices your network supports. The larger your organization is, the more ports you’ll need. Considering the possible expansion of your network and the possible increase of your user amount, you need to prepare extra ports for a long term plan.

Speed

There are various switches with different speeds, such as Gigabit Ethernet switch and 10GbE switch used at the edge of the network, as well as 40GbE switch and 100GbE switch used in the network core layers. When you determine the speed, the key factor to consider is the need for your network users and future growth. Such as how large are the volumes of the transferring data and whether do you require a faster link.

Stackable vs Standalone

Will your network grow larger? If your answer is yes, then you may choose a stackable switch. Standalone switches need to be configured individually, and troubleshooting also needs to be handled on an individual basis. While stackable switches allow for multiple switches to be configured as one entity. With this advantage, you can save time and energy when you manage on the stackable switches. Here I want to recommend you FS.COM S3900 switches, which are stackable switches. The following video is a tutorial about how to stack switches using S3900 switches.

Summary

In the above passage, we’ve explained how people define switch in networking and analyze the types of switches. Besides, this article offers some suggestions about how to choose a switch for your network. I believe that you have got a general idea about switch definitions in networking. If you need a little more help and advice with switch definition in networking, then please do not hesitate to let us know. For purchasing a high-quality switch with a low cost or for more products’ information, please contact us at sales@fs.com.

Ethernet Switch: How Much Do You Know It?

Today, all plants are virtually networked via Ethernet. High requirements are placed on the network infrastructure and network components. Ethernet switch is the integral piece of IT infrastructure, capable of receiving, processing and transmitting data between two devices connected by a physical layer. Due to the increasing application of big data analytics and cloud-based services in various end-user segments, data centers are envisaged to fuel the adoption of Ethernet switch. The augmented global demand for data centers is the key driver for the growth of Ethernet switch market. To satisfy the large and ever-increasing market for Ethernet switch, there are many varieties of switches offered different purposes. This article will help you get a deep understanding of the different types of Ethernet switch.

What is an Ethernet Switch?

A Ethernet switch is a tool for connections between the systems and equipment to forward data selectively to one or more connected devices on the same network. These connections are generally created through the use of structured cabling that links both the station side and the device that you are trying to share data with, such as a server or another computer. In this way, Ethernet switch can control the flow of traffic passing through a network, maximizing the network’s efficiency and security. More advanced Ethernet switch, called managed switch, are also capable of providing additional functions, such as network load balancing, address translation or data encryption and decryption.

FS Ethernet switch

How Dose an Ethernet Switch Work?

Ethernet switch links Ethernet devices together by relaying Ethernet frames between the devices connected to the switches. By moving Ethernet frames between the switch ports, a switch links the traffic carried by the individual network connections into a larger Ethernet network. Ethernet switches perform their linking function by bridging Ethernet frames between Ethernet segments. To do this, they copy Ethernet frames from one switch port to another, based on the Media Access Control (MAC) addresses in the Ethernet frames. Ethernet bridging was initially defined in the 802.1D IEEE Standard for Local and Metropolitan Area Networks: Media Access Control (MAC) Bridges. The standardization of bridging operations in switches makes it possible to buy switches from different vendors that will work together when combined in a network design. That’s the result of lots of hard work on the part of the standards engineers to define a set of standards that vendors could agree upon and implement in their switch designs.

diagram of Ethernet switches connections

Different Types of Ethernet Switch

Ethernet switch are broadly categorized into two main categories – modular switches and fixed switches. Modular switches allow you to add expansion modules into the switches as needed, thereby delivering the best flexibility to address changing networks. Fixed switches are switches with a fixed number of ports and are typically not expandable. This category can be broken down even further into unmanaged, lightly managed, and fully managed.

Unmanaged Switch

An unmanaged switch is mostly used in home networks and small companies or businesses, as it is the most cost effective for deployment scenarios that require only basic layer 2 switching and connectivity. The unmanaged switch is not configurable and have all of their programming built in. It is ready to work straight out of the box. And it is the easiest and simplest installation, because of its small cable connections. An unmanaged switch is perfect in this situation since it requires the least amount of investment with regards to both expense and time.

Smart Switch / Lightly Managed Switch

A smart switch is the middle ground between the unmanaged and fully managed switches. These smart switches offer limited customization, but do possess the granular control abilities that a fully managed switch has. In addition, smart switches offer certain levels of management, quality-of-service (QoS), security, but they are lighter in capabilities and less scalable than the managed switches. Smart switches tend to have a management interface that is more simplified than what managed switches offer. They also offer the capability to set up options like Quality of Service (QoS) and VLANs, which can be helpful if your organization has VoIP phones, or if you want to segment your network into work groups. Therefore, smart switches are the cost-effective alternative to managed switches. They are still valid choices for the regular consumer, as they are generally easy to use and you can glean a bit more information off of them on how your network is configured compared to unmanaged switches.

Fully Managed Switch / Enterprise Managed Switch

Managed Layer 2 Switch: A modern managed switch provides all the functionality of an unmanaged switch. In addition, it can control and configure the behavior of the device. This typically introduces the ability to support virtual LANs (VLANs), which is why almost all organizations deploy managed switches versus their cheaper alternatives.

Managed Layer 3 Switch (Multilayer Switch): This type of switch provides a mix of functionality between that of a managed Layer 2 switch and a router. The amount of router function overlap is highly dependent on the switch model. At the highest level, a multilayer switch provides better performance for LAN routing than almost any standard router on the market, because these switches are designed to offload a lot of this functionality to hardware.

data-center-network-architecture

Managed switches are designed to deliver the most comprehensive set of features to provide the best application experience, the highest levels of security, the most precise control and management of the network, and offer the greatest scalability in the fixed configuration category of switches. As a result, they are usually deployed as aggregation/access switches in very large networks or as core switches in relatively smaller networks. Managed switches should support both L2 switching and L3 IP routing, though you’ll find some with only L2 switching support.

Conclusion

The Ethernet switch plays an integral role in most modern Ethernet local area networks (LANs). Mid-to-large sized LANs contain a number of linked managed switches. Small office/home office (SOHO) applications typically use a single unmanaged switch. This article has introduced different types of switches. Depending on the number of devices you have and the number of people using the network, you have to choose the right kind of switch that fits your space. FS.COM has provided a comprehensive set of Ethernet switches. If you have any requirements, welcome to visit our website for more detailed information.

Understanding PoE & PoE Switch

Enterprises are quickly evolving with new network devices to improve communication and security. Power over Ethernet (PoE), a way to deliver electrical power over LAN cables to network devices, has been widely deployed to provide power to various endpoints in enterprise environments. If you want to upgrade your network to PoE, one way is to deploy a PoE switch. This paper will provide an overview of the PoE technology and PoE switches.

What Is a PoE in Networking?

Power over Ethernet, also known as PoE, is a networking feature defined by the IEEE 802.3af and 802.3at standards. PoE can combine the two connections into one Ethernet cable so that a single network cable will transmit both data and 25W of electricity. In this way, it can minimise the number of wires when installing the network, which realises the lower cost, less downtime, easier maintenance, and greater installation flexibility in networking.

POE-working-principle

Why Use PoE?

Because PoE is allowed to use one cable for both power and data transmission, PoE can save money on purchasing and running cable for networking equipment. It can bring many advantages to the network as follows.

  • Time and cost savings

Network cables do not require a qualified electrician to install them, and can be located anywhere, so PoE eliminates the time and cost of hiring professional electrical installers.

  • Flexibility

Network administrators can deploy devices (eg: IP cameras and wireless access points) wherever they are needed most, and redeploy easily if required.

  • Safety

Because PoE utilises a relatively low voltage, it presents low risks of electrical hazards.

  • Scalability

PoE makes it simple to add new equipment to a network.

What is a PoE Switch Used for?

A POE switch is a network switch that has a built-in PoE injection. It can connect to other network devices as normal, and the switch will detect whether they are PoE-compatible and enable power automatically. PoE switches are available to suit all applications, ranging from low-cost unmanaged edge switches with a few ports, to complex multi-port rack-mounted units with sophisticated management. They can run PoE up to 100 metres from the switch or hub to the NIC, regardless of where the power is injected. The limitation is not the power, it’s the Ethernet cabling standards that limit the total length of cabling to 100 metres.

FS PoE Switch Solution?

FS provides fully managed PoE switches, which are available with 8, 24 or 48 PoE Gigabit Ethernet ports of auto-sensing IEEE 802.3af/at. The PoE Switches are ideal for small business networks that need to inexpensively use PoE to deploy wireless access points and IP-based network surveillance cameras. They deliver robust performance and intelligent switching for growing networks, so PoE switches will be the best choice to install and manage your devices. The model details of FS’s PoE switches are listed below.

How to Ensure Successful PoE Deployments?
1. Provide Sufficient Power to the Remote Powered Device

According to the IEEE 802.3af standard, the powered remote device can draw up to 12.95 watts of power. Considering the loss of the cable length, the power sourcing equipment (PSE) must have the ability to provide 15.4 watts of power to each port. For example, a 24-port Ethernet switch needs approximately 370 watts of power to supply the necessary power to each port. The PoE switches should have more than 370 watts available given the size of the power supply used in each device. It depends on how much power their switching functions require.

2. Connect the Power Source to Uninterruptible and Redundant Power

Connect the critical power-sourcing devices to an uninterruptible power supply, and use devices with dual redundant power supplies to ensure that your critical devices never lose power.

3. Deploy Only IEEE 802.3af-Compliant Devices

Carefully read the technical documentation and contact the technical support number to determine compatibility. Failure to do so will leave you frustrated and will cost you time and money.

4. Pay Attention to Cabling-performance Specifications

Pay close attention to the manufacturer’s specifications and look for Cat5e and Cat6a compliance. Also, you should remember per TIA standards, only four connectors can exist between the switch or hub and the network interface card (NIC). A midspan device should be counted and treated as one of these connection points.

5. Use the Most Cost-effective PoE Method for Your Network

The business motivation behind deploying IP-based technologies like WiFi and VoIP is to decrease networking costs. A significant benefit of PoE is that it runs on your existing infrastructure.

Conclusion

PoE is a recently developed technology, and it simplifies enterprise deployment with lower operating expenses, higher availability, and faster deployment. FS has provided PoE switches in a variety of specifications, such as 8 port Poe switches, 24 port PoE switches and 48 port PoE switches, which may make your trip as comfortable as possible. For more information, please welcome to www.fs.com.

Related Article: Power over Ethernet Analysis

Whitebox Switch: Are You Ready for Deploy It?

With the development of Cloud services and networking, 40G/100G whitebox switch become very popular in the market. These switches will provide Web scale organizations and service providers more control and flexibility in their data center networks. So what is whitebox switch? Whitebox switches refers to the ability to use ‘generic,’ off-the-shelf switching (or whitebox switching) and routing hardware, in the forwarding plane of a software-defined network (SDN). Moreover, whitebox switches rely on an operating system (OS), which may come already installed or can be purchased from a software vendor and loaded separately, and then integrate with the deploying organization’s Layer 2/Layer 3 topology and support a set of basic networking features. On the whole, OS is an integral part of whitebox switches, and the rise of SDN has brought whitebox switches into the public eye. Next, let’s take a closer look at whitebox switches.

Whitebox Switch

OS Defines Whitebox Switch

Whitebox switches are useless without software, because every switch needs an operating system. The OS needs to seamlessly integrate with existing L2/L3 topology and support a basic set of features. A common operating system for whitebox switches is Linux-based, because many open and free Linux tools are available, which can help administrators customize the switches to their needs. Typically, a whitebox switch may come pre-loaded with minimal software or it may be sold as a bare metal device. The advantage of the whitebox switches is that switches can be customized to meet an organization’s specific business and networking needs.

Network OS

However, how to put the OS on the whitebox switches? Some vendors sell a complete solution with the OS that is already installed on the whitebox, while others set up distributors to provide the bare metal devices that the OS is directly brought from the software vendor. Both of these two approaches are feasible, depending on the scale of the deployment and the desire for the network.

SDN & Whitebox Switch

Beyond the operating system, whitebox switches are more valuable if they interact with SDN controllers. And the widespread implementation of SDN has boosted the use of whitebox switches. SDN is an approach to design, build and manage networks, which can separate the network’s control and forwarding planes. In result, the network control will become directly programmable and the underlying infrastructure will be abstracted for applications and network services. The goal of SDN is to enable cloud and network engineers and administrators to respond quickly to changing business requirements via a centralized control console. At the same time, the switches in SDN environment rely on software-based network function virtualization (NFV), which offers great convenience for the users of whitebox switches. Because whitebox switch allows its customers to choose the best suitable operating system for themselves. And in the future, most whitebox switches will function in an SDN environment in which the SDN controller is making forwarding and control-plane decisions from a centralized point for all switches in the network.

sdn

The Growing Market for Whitebox Switch

In general, the data center Ethernet switch market has seen tremendous growth and investment over the past years. The Layer 2-3 Ethernet switch market is expected to exceed $25 billion in 2019, according to Dell’Oro Group. What’s more, some high-end users are tired of vendor lock-in switches, and they might be ready to try a whitebox switch to get what they want. Whitebox switches can customize the system to limit unneeded processes and concentrate the processing power of the switch on the important features, so it leads to a customized switch platform that provides perfect performance for a narrow range of uses. Customers with highly unique support needs will also benefit from whitebox switches. Through the separation of software and hardware, customers can obtain different support levels for hardware and software.

FS.COM 40G/100G Whitebox Switch Solutions

FS.COM 40G/100G whitebox switches are based on IPinfusion’s ZebOS with integration of Layer 2 to Layer 4 packet processing engine, traffic management and fabric interface. The aim is to achieve flexibility, scalability, efficiency and cost effectiveness in data center networks. Furthermore, the operating systems of these switches are developed on the basis of Linux. Last but not least, all the 40G/100G whitebox switches in FS.COM support SDN function which can make networks more affordable and easier to manage.

fs-40g-100g-white-box-switches

All in all, the whitebox switches support current and future data center requirements, which is ideally suited for data center environments in either Leaf or Spine deployments. They provide superior low latency and power efficiency in a clean PHYless design, while offering high reliability features such as redundant and hot swappable power supplies and fans in forward and reverse airflow configurations. And they provide QSFP+ ports, which enable flexible choices of port speed providing unparalleled flexibility and the ability to seamlessly transition data centers to the next generation of Ethernet performance.

Summary

Whitebox switches can be deployed either in the data center or in the access network, which can reduce capital expenditures and leverage open SDN tools to improve time to deployment and automation. If you want to deploy whitebox switches with lower cost and great flexibility, welcome to contact us via www.fs.com.

Optical Switches Overview

An optical switch is a device that can selectively switch light signals that run through in optical fibers or integrated optical circuits from one circuit to another. That is to say, optical switches can transfer light signals between different channels in communication networks. As the growing popularity of Internet and telephone, greater quantities of data managed by communication networks also expanded. Optical switching technology provides a perfect solution to fully exploit capacity of optical systems. The main focus of this post is to introduce basics of optical switches in optical communication.

optical-switches

Working Principles & Functions of Optical Switches

As we all know, when a light signal runs through from one computer to another in fiber optic networks, it may be required to move the signal between different fiber paths. To accomplish this, a switch is required to transfer the signal with a minimum loss. Optical switch is a technology needed. The optical switch we often see is operated by mechanical method which just moves fiber or other bulk optic components. But they can offer unprecedented high stability and unmatched low cost performance.

Optical switches are mainly deployed in establishing the light path. They feature scalability and highly reliable switching capacity. Following are the major functions that optical switches bear in optical cross networks.

  • Protection. Sometimes a failure of some single point can cause the whole network breaking down. And the protection switching is to protect the transmission data, which can avoid network fault before finding the failure causes.
  • Optical add/drop multiplexing. Optical switches must be equipped with the capability that can add or delete the wave channels without any electronic processing. This kind of optical switches is also called wavelength selective switches.
  • Optical spectral monitoring. Optical spectral monitoring is a network management operations. In this process, operators receive a small portion of optically tapped signal for monitoring power level, wavelength accuracy and optical cross talk.
Common Types of Optical Switches

As data requirements grow, the traditional electrical switches no longer meet people’s demand. There are two major types of optical switches on the market: opto-mechanical optical switches and MEMS (Micro-electromechanical Systems) optical switches.

Opto-Mechanical Optical Switches

Opto-mechanical optical switch is an old type of switches but the most widely used one. It can produce different optical path selections out of a plurality of optical path sections that are oriented in different spatial directions. Hence opto-mechanical optical switches can be used in multi-channel optical power monitoring, optical local area networks, switching multiple laser sources or optical receivers in Ethernet networks. They are also very useful in optical fiber, components or systems testing and measurement, as well as applications in multi-point fiber sensor systems. Generally, according to the number of redirecting signals, opto-mechanical optical switches have different configurations such as 1×1, 1×2, 1×4, 1×16, etc. In simple terms, the 1×8 opto-mechanical optical switch module connects optical channels by redirecting an 1 incoming optical signal into a selected signal from 8 output fibers. This kind of optical switches can achieve excellent reliability, insertion loss, and cross talk.

opto-mechanical-optical-switches

MEMS Optical Switches

MEMS optical switches use a micro-mirror to reflect a light beam. And the direction that the light beam is reflected can be changed by adjusting the angle of the mirror, which allows the input light to be connected to any out port. It is a compact optical switch which connects optical channels by redirecting incoming optical signals into the selected output fibers. And the switching state is highly stable against environmental variations of temperature and vibration due to its unique design. In some degree MEMS optical switch can be considered as a subcategory of opto-mechanical switches. But it is distinguished from opto-mechanical switches in many aspects such as the characteristics, performance and reliability. The most obvious is the opto-mechanical switch has more bulk compared to other alternatives, but the MEMS switch overcomes this. Besides, MEMS optical switches also have different configurations such as 1×8, 1×12, 1×16, etc.

1x8-mems-optical-switch

Summary

As the increasing growth of high speed transmission demand for networks, optical networks have become the most cost-effective solution. Optical switches play a vital role in today’s optical network system. They can offer users significant power, space and cost savings. Now different optical switches are available on the market, so you can choose a suitable one based on your requirements.

Custom Fiber Optic Swich Systems

The custom fiber optic system is conceived as a collection of building blocks which can be chosen by the costumer to fulfill his needs. The heart of the systems are the MEMS Optical Switch, varialbe optical attenuators, couplers, splitters together with kilometers of fiber divided into sections. All integrated in the same instrument to achieve costumer’s needs concerning measurements, simulations and testing.

The inputs are basically the fiber optical inputs and the control inputs. For the moment a serial input (RS232) and an USB control are available. The outputs available for measurement are the optical fibers themselves toghether with outputs to measure the energy, s/n ration, and BERT.

Applications

Some of the applications for Custom Fiber Optic System are the following:

Real Network Simulation: This system allows to simulate real network conditions by switching reels of fiber into the optical path. Due to its flexibility and its configurable character, it can simulate different network configurations scenarios in order to study the behavior in each case.

Improve Optical Testing and Measurements: Thanks to the electronic control of the optical parts, the user can choose easily which components are active simply by sending the required information to the serial/USB port.

Automatic Testing: The user can build his own solution integrating system together with a computer, other measurement instruments and a bus.

fiber optic swiches

Fiberstore optical switch uses a moving prism between fixed collimator pairs, which allows bi-directional operation independent of data rate and signal format. 9/125 and 62.5/125 µm, 1 m long, jacketed fiber pigtails are available on all models. In addition, optional ST, FC and SC fiber connector termination options can be added. We supply optical switches based on Opto-mechanical Switch, MEMS, Solid-State technology with proven reliability and the configurations are available as 1 x 1, 1 x 2, 2 x 2, etc.. Moreover, we offer non-latching, latching, single-mode and multimode versions. Our optical switches are all with high quality and ready for the FTTx applications.

What is an Optical Switch

Optic Switch is a device that enables signals in optical fibers or integrated optical circuits to be selectively switched from one circuit to another. In telecommunication, an optical switch is the unit that actually switches light between fibers, and a photonic switch is one that does this by exploiting nonlinear material properties to steer light (i.e., to switch wavelengths or signals within a given fiber). Hence a certain portion of the optical switch market is made up of photonic switches. These will contain within them an optical switch, which will, in some cases, be a photonic switch.

An optical switch may operate by mechanical means, such as physically shifting an optical fiber to drive one or more alternative fibers, or by electro-optic effects, magneto-optic effects, or other methods. Slow optical switches, such as those using moving fibers, may be used for alternate routing of an optical switch transmission path, such as routing around a fault. Fast optical switches, such as those using electro-optic or magneto-optic effects, may be used to perform logic operations; also included in this category are semiconductor optical amplifiers, which are optoelectronic devices that can be sued as optical switches and be integrated with discrete or integrated ,microelectronic circuits.

Types of optical switches

Optical switches can be divided into mechanical and non-mechanical ones according to the driving methods.

Mechanical optical switch relies on the movement of optical fiber or optical elements to convert the optical path, such as a mobile optical fiber type, moving the sleeve to move the lens (including mirrors, prisms and self-focusing lens) types. The biggest advantage of this kind of optical switch is a low insertion loss and low crosstalk. Its disadvantage is slow and easy to wear, easy to vibration, impact shocks.

Non-mechanical optical switch relies electro-optic, magneto-optic, thermo-optic and other effects to change the refractive index of the optical waveguide, the optical path changes, such as electro-optic switch, magneto-optic switch, and thermo-optic switch. This kind of optical switch has good repeatability, fast switching speed, high reliability, long life and other advantages, and small size, can be monolithically integrated. The disadvantage is that the insertion loss and crosstalk performance is not ideal, which should be improved.

There are three common types of optical switches: Opto-Mechanical Switch, Thermo-Optical Switch, Electro-Optical Switch.

Opto-Mechanical Switch

Opto-mechanical switch is the oldest type of optical switch and the most widely deployed at the time. These devices achieve switching by moving fiber or other bulk optic elements by means of stepper motors or relay arms. This causes them to be relatively slow with switching times in the 10-100 ms range. They can achieve excellent reliability, insertion loss, and crosstalk. Usually, opto-mechanical optical switches collimate the optical beam from each input and output fiber and move these collimated beams around inside the device. This allows for low optical loss, and allows distance between the input and output fiber without deleterious effects. These devices have more bulk compared to other alternatives, although new micro-mechanical devices overcome this.

Thermo-Optic Switch

Thermo-optic switches are normally based on waveguides made in polymers or silica. For operation, they rely on the change of refractive index with temperature created by a resistive heater placed above the waveguide. Their slowness does not limit them in current applications.

Electro-Optic Switch

These are typically semiconductor-based, and their operation depends on the change of refractive index with electric field. This characteristic makes them intrinsically high-speed devices with low power consumption. However, neither the electro-optic nor thermo-optic optical switches can yet match the insertion loss, backreflection, and long-term stability of opto-mechanical optical switches. The latest technology incorporates all-optical switches that can cross-connect fibers without translating the signal into the electrical domain. This greatly increases switching speed, allowing today’s telcos and networks to increase data rates. However, this technology is only now in development, and deployed systems cost much more than systems that use traditional opto-mechanical switches.

We supply optical switches based on Opto-Mechanical technology with proven reliability and the configurations are available as 1 x 1, 1 x 2, 2 x 2, etc.. Moreover, we offer non-latching, latching, single-mode and multimode versions. What’s more, we also suppy multi mode om3 armored fiber, if you would like to know more information about our products, please send your requirements to sales@fiberstore.com or call us.