Migration at high speeds. What to choose?
The competitiveness of both a separate company and the country's economy as a whole, increasingly depends on information technology. The performance of processors, the density of servers, virtual machines, storage systems capacity is rapidly increasing ... And, of course, the demand for data transfer speeds is growing. And it is not enough to be able to provide only current needs, it is important to look ahead and understand what you should be prepared for tomorrow, outlining the most profitable and flexible migration path to higher speeds. The faster technologies and applications develop, the more difficult it is to solve this problem.
In the center of the digital transformation are data centers (DPCs). Current trends in the development of data centers show that the need for bandwidth of their networks will increase annually by 25-35%. The data center network architecture must ensure processing of large volumes of traffic and, equally important, scaling of server, network and storage resources with the least amount of disconnections and reconfigurations.
The traditional network architecture based on a three-tier topology (access – aggregation – core) does not meet the new requirements. It is not able to provide support for changes in the volume and direction of data transfer and to adapt to the rapid growth of traffic between servers (this traffic is often called "east-west"). In addition, it is far from optimal for supporting modern virtualized applications that require low latency traffic. Therefore, there is a transition to the new leaf-spine architecture, which is often called the “network factory”.
Traditional three-tier architecture (left) and leaf-spine architecture (right)
')
The leaf-spine architecture optimizes the network structure for transmitting large volumes of traffic in the direction of "west-east", which improves the interaction of servers to ensure the operation of virtualized and cloud applications. Each leaf switch connects to all spine switches, thus creating a fault-tolerant “every-everyone” communication structure. The network of fiber-optic connections forms a high-capacity fault-tolerant network factory. Such a factory requires a large number of high-speed optical connections, especially at the inter-switch interworking level. Moreover, the capacity of these connections should be regularly increased, - according to experts, the need to switch to higher data transfer rates will occur every two to three years. No less rapidly growing requirements and bandwidth for connecting servers.
Migration to higher speeds is a complex and multifaceted business. It is necessary to consider a wide range of issues related to the choice of network technology, fiber type, transmission schemes, cable infrastructure configuration, etc. And, of course, consider the cost of the solution. To determine the best option for a particular network, it is necessary to carefully examine various aspects. Below are just a few of them.
The choice of speed (network technology). Until recently, one of the main migration scenarios in high-speed networks was considered to be switching from Ethernet technology 10G to 40G and further to 100G. However, the IEEE Institute, striving not only to continue the “race of speeds”, but also to increase the economic efficiency of new technologies, proposed new speeds that fill the “gaps” between 10G, 40G and 100G. And as the next step in the development of Ethernet networks, the transition to 25 Gbit / s flows has become increasingly common. Constructed on the basis of 25G streams, solutions that allow easy migration to 50G (2x25G) and 100G (4x25G) technologies guarantee a better return on investment than switching to 40G.
Experts believe that the demand for 40G ports has reached its peak, and further their sales will decline, while the demand for ports 25G and 50G will grow. Today, 40G ports are used primarily for connecting servers, and one QSFP 40G port on the switch is usually divided into four 10 Gigabit endpoint connections. But server technologies are rapidly evolving, and new solutions will be able to consume much more traffic than the 10G channel can provide.
According to the forecast, the capacity of one communication line for connecting switches to network factories will double and by 2020 will reach 100 Gbit / s, which will require the organization of even more high-speed communication lines. And IEEE and other organizations are already considering a possible transition to the systems 200G, 400G, 800G, 1.6T and even more speed.
Ethernet development roadmap
It should be understood that with increasing speed increases the complexity of technical solutions. To support high speeds, it is often necessary to move from traditional duplex to parallel circuits. However, with the advent of new technologies, such as SWDM and WB-MMF, it is possible to postpone the transition to parallel optics.
Select the type of fiber. Today there are three main options: single-mode fiber (OMV), multimode (MMV) and the recently appeared broadband multimode (SHP-MMV, or WB-MMF). The greatest speed and range can provide single-mode fiber. But its implementation seriously hampers the high cost of relevant technical solutions. Although the price of single-mode equipment is gradually decreasing, it remains unnecessarily expensive for many projects. In data centers, single-mode fibers are traditionally used only in the area from the point of cable entry from external telecommunication networks to the main distribution point. In addition, OMV is used in mega-data centers with their large distances, which impede the use of IIM.
For most projects, multimode fiber remains the most attractive solution for a set of indicators (performance, density, price). The main problem of IIM is a small distance. Moreover, as data transmission speeds increase, there is a tendency for the maximum range of such transmission to decrease. However, the appearance on the market of higher-quality components and a competent approach to the design of infrastructure makes it possible to provide the necessary speeds on sufficiently long channels and support for new network topologies in data centers.
The newly emerged OM5 class MFW may be the best option for ensuring migration to higher speeds. Presented by CommScope in 2015, ShP-IIM allows the use of SWDM spectral multiplexing technology, while ensuring backward compatibility with OM3 and OM4 solutions, which means support for all “legacy” applications. Multiplexing of four spectral channels in one broadband multimode fiber of class OM5 will allow four times to increase its throughput.
Principles of spectral compaction
Selection of transmission scheme: serial or parallel. As already mentioned, as the requirements for the speed of data transmission increase, the industry tends to switch to parallel optics. Even without BF-IIM with the use of parallel circuits, it is possible to effectively provide 40G, 100G and even 200G / 400G Ethernet connections based on IIM. The use of NBP-IIM and spectral multiplexing technology allows for high transmission speeds using a significantly smaller number of fibers. In general, there is an increase in the number of options for implementing high-speed transmission systems using different numbers of fibers and wavelengths, and the selection of the optimal option is a difficult and very important task.
Choosing an MPO option. Switching to parallel optics means the increasing use of multi-fiber (group) MPO connectors. There are variants of MPO with different numbers of fibers: 8, 12, 24, etc. Systems with 8-fiber MPO connectors effectively support very popular circuits with QSFP transceivers using four lines (channels). They are used primarily in 4X10G and 4X25G solutions for connecting servers and storage systems to the network.
12-fiber MPO systems support duplex and parallel transmission schemes, providing excellent flexibility and range for most data center applications. Systems based on 24-fiber MPO connectors increase the density and capacity of physical infrastructure. They also support duplex and parallel transmission schemes and provide the lowest cost per fiber (compared to 8 and 12 fiber products).
MPO Connector Options with Different Number of Fibers
As you can see, when planning the transition to higher transmission speeds, it is necessary to consider many issues and take into account a lot of nuances. To make life easier for its customers and partners, CommScope has developed a set of recommendations and a full range of solutions for high-speed migration - High-Speed-Migration (HSM). It includes modern copper-fiber and optical cable systems, including high (HD) and ultra-high (UD) density solutions, using the latest types of transmission media (including OM5 wideband multimode fiber).
Optical shelf high density SYSTIMAX HD-4U
For the most rapid and painless modernization of the cable infrastructure, as well as for the rapid deployment of new services, the pre-determined cable systems that are part of the HSM portfolio are important. According to some estimates, plug-and-play deployable solutions can significantly reduce installation time (up to 90%) and maintenance (approximately 50%). As the number of fiber-optic connections in the network increases, the benefit of using such solutions increases.
You should also understand that increasing the speed often means the need to work at the limit of the capabilities of existing transmission media. Allowable losses are significantly reduced, which causes the need for systems with ultra-low losses. CommScope has developed such solutions: ULL - Ultra-Low Loss. In combination with the new OM5 fiber, ULL solutions provide the construction of an optimal optical infrastructure for corporate data centers. These solutions allow for high transmission speeds with a small number of fibers over long distances. For example, even when using six LC connections and four MPO lines in a line, the ULL solution based on LazrSPEED OM5 WideBand fiber provides 100G (100G-SWDM4) systems for a distance of up to 150 m using only two fibers.
ULL Modules
The most important component of the HSM solution portfolio is the imVision automated infrastructure management system. Providing real-time 100% reliable information about connections at the physical level, it provides valuable data necessary for the design and error-free execution of migration scenarios. Also in the HSM portfolio includes network design tools. In particular, they allow you to calculate the maximum line length and the number of connections to support the necessary network technology, taking into account the type of fiber. At the same time, it is possible to take into account the possibility of supporting the technology that is planned to be introduced at the next stage of the network development.
In general, when planning the transition to higher speeds, it is important to choose a partner capable of offering highly scalable solutions that provide the flexibility and adaptability needed to deploy new applications, a painless transition to higher speeds with maximum preservation of the investment made.
In the center of the digital transformation are data centers (DPCs). Current trends in the development of data centers show that the need for bandwidth of their networks will increase annually by 25-35%. The data center network architecture must ensure processing of large volumes of traffic and, equally important, scaling of server, network and storage resources with the least amount of disconnections and reconfigurations.
The traditional network architecture based on a three-tier topology (access – aggregation – core) does not meet the new requirements. It is not able to provide support for changes in the volume and direction of data transfer and to adapt to the rapid growth of traffic between servers (this traffic is often called "east-west"). In addition, it is far from optimal for supporting modern virtualized applications that require low latency traffic. Therefore, there is a transition to the new leaf-spine architecture, which is often called the “network factory”.
Traditional three-tier architecture (left) and leaf-spine architecture (right)
')
The leaf-spine architecture optimizes the network structure for transmitting large volumes of traffic in the direction of "west-east", which improves the interaction of servers to ensure the operation of virtualized and cloud applications. Each leaf switch connects to all spine switches, thus creating a fault-tolerant “every-everyone” communication structure. The network of fiber-optic connections forms a high-capacity fault-tolerant network factory. Such a factory requires a large number of high-speed optical connections, especially at the inter-switch interworking level. Moreover, the capacity of these connections should be regularly increased, - according to experts, the need to switch to higher data transfer rates will occur every two to three years. No less rapidly growing requirements and bandwidth for connecting servers.
Migration to higher speeds is a complex and multifaceted business. It is necessary to consider a wide range of issues related to the choice of network technology, fiber type, transmission schemes, cable infrastructure configuration, etc. And, of course, consider the cost of the solution. To determine the best option for a particular network, it is necessary to carefully examine various aspects. Below are just a few of them.
Choice problems
The choice of speed (network technology). Until recently, one of the main migration scenarios in high-speed networks was considered to be switching from Ethernet technology 10G to 40G and further to 100G. However, the IEEE Institute, striving not only to continue the “race of speeds”, but also to increase the economic efficiency of new technologies, proposed new speeds that fill the “gaps” between 10G, 40G and 100G. And as the next step in the development of Ethernet networks, the transition to 25 Gbit / s flows has become increasingly common. Constructed on the basis of 25G streams, solutions that allow easy migration to 50G (2x25G) and 100G (4x25G) technologies guarantee a better return on investment than switching to 40G.
Experts believe that the demand for 40G ports has reached its peak, and further their sales will decline, while the demand for ports 25G and 50G will grow. Today, 40G ports are used primarily for connecting servers, and one QSFP 40G port on the switch is usually divided into four 10 Gigabit endpoint connections. But server technologies are rapidly evolving, and new solutions will be able to consume much more traffic than the 10G channel can provide.
According to the forecast, the capacity of one communication line for connecting switches to network factories will double and by 2020 will reach 100 Gbit / s, which will require the organization of even more high-speed communication lines. And IEEE and other organizations are already considering a possible transition to the systems 200G, 400G, 800G, 1.6T and even more speed.
Ethernet development roadmap
It should be understood that with increasing speed increases the complexity of technical solutions. To support high speeds, it is often necessary to move from traditional duplex to parallel circuits. However, with the advent of new technologies, such as SWDM and WB-MMF, it is possible to postpone the transition to parallel optics.
Select the type of fiber. Today there are three main options: single-mode fiber (OMV), multimode (MMV) and the recently appeared broadband multimode (SHP-MMV, or WB-MMF). The greatest speed and range can provide single-mode fiber. But its implementation seriously hampers the high cost of relevant technical solutions. Although the price of single-mode equipment is gradually decreasing, it remains unnecessarily expensive for many projects. In data centers, single-mode fibers are traditionally used only in the area from the point of cable entry from external telecommunication networks to the main distribution point. In addition, OMV is used in mega-data centers with their large distances, which impede the use of IIM.
For most projects, multimode fiber remains the most attractive solution for a set of indicators (performance, density, price). The main problem of IIM is a small distance. Moreover, as data transmission speeds increase, there is a tendency for the maximum range of such transmission to decrease. However, the appearance on the market of higher-quality components and a competent approach to the design of infrastructure makes it possible to provide the necessary speeds on sufficiently long channels and support for new network topologies in data centers.
The newly emerged OM5 class MFW may be the best option for ensuring migration to higher speeds. Presented by CommScope in 2015, ShP-IIM allows the use of SWDM spectral multiplexing technology, while ensuring backward compatibility with OM3 and OM4 solutions, which means support for all “legacy” applications. Multiplexing of four spectral channels in one broadband multimode fiber of class OM5 will allow four times to increase its throughput.
Principles of spectral compaction
Selection of transmission scheme: serial or parallel. As already mentioned, as the requirements for the speed of data transmission increase, the industry tends to switch to parallel optics. Even without BF-IIM with the use of parallel circuits, it is possible to effectively provide 40G, 100G and even 200G / 400G Ethernet connections based on IIM. The use of NBP-IIM and spectral multiplexing technology allows for high transmission speeds using a significantly smaller number of fibers. In general, there is an increase in the number of options for implementing high-speed transmission systems using different numbers of fibers and wavelengths, and the selection of the optimal option is a difficult and very important task.
Choosing an MPO option. Switching to parallel optics means the increasing use of multi-fiber (group) MPO connectors. There are variants of MPO with different numbers of fibers: 8, 12, 24, etc. Systems with 8-fiber MPO connectors effectively support very popular circuits with QSFP transceivers using four lines (channels). They are used primarily in 4X10G and 4X25G solutions for connecting servers and storage systems to the network.
12-fiber MPO systems support duplex and parallel transmission schemes, providing excellent flexibility and range for most data center applications. Systems based on 24-fiber MPO connectors increase the density and capacity of physical infrastructure. They also support duplex and parallel transmission schemes and provide the lowest cost per fiber (compared to 8 and 12 fiber products).
MPO Connector Options with Different Number of Fibers
HSM Strategy
As you can see, when planning the transition to higher transmission speeds, it is necessary to consider many issues and take into account a lot of nuances. To make life easier for its customers and partners, CommScope has developed a set of recommendations and a full range of solutions for high-speed migration - High-Speed-Migration (HSM). It includes modern copper-fiber and optical cable systems, including high (HD) and ultra-high (UD) density solutions, using the latest types of transmission media (including OM5 wideband multimode fiber).
Optical shelf high density SYSTIMAX HD-4U
For the most rapid and painless modernization of the cable infrastructure, as well as for the rapid deployment of new services, the pre-determined cable systems that are part of the HSM portfolio are important. According to some estimates, plug-and-play deployable solutions can significantly reduce installation time (up to 90%) and maintenance (approximately 50%). As the number of fiber-optic connections in the network increases, the benefit of using such solutions increases.
You should also understand that increasing the speed often means the need to work at the limit of the capabilities of existing transmission media. Allowable losses are significantly reduced, which causes the need for systems with ultra-low losses. CommScope has developed such solutions: ULL - Ultra-Low Loss. In combination with the new OM5 fiber, ULL solutions provide the construction of an optimal optical infrastructure for corporate data centers. These solutions allow for high transmission speeds with a small number of fibers over long distances. For example, even when using six LC connections and four MPO lines in a line, the ULL solution based on LazrSPEED OM5 WideBand fiber provides 100G (100G-SWDM4) systems for a distance of up to 150 m using only two fibers.
ULL Modules
The most important component of the HSM solution portfolio is the imVision automated infrastructure management system. Providing real-time 100% reliable information about connections at the physical level, it provides valuable data necessary for the design and error-free execution of migration scenarios. Also in the HSM portfolio includes network design tools. In particular, they allow you to calculate the maximum line length and the number of connections to support the necessary network technology, taking into account the type of fiber. At the same time, it is possible to take into account the possibility of supporting the technology that is planned to be introduced at the next stage of the network development.
In general, when planning the transition to higher speeds, it is important to choose a partner capable of offering highly scalable solutions that provide the flexibility and adaptability needed to deploy new applications, a painless transition to higher speeds with maximum preservation of the investment made.
Source: https://habr.com/ru/post/349120/
All Articles