Consolidation of LAN and SAN data center networks based on DCB and FCoE protocols
The purpose of this article is to give the reader a basic understanding of the technologies that made it possible to combine two Ethernet and Fiber Channel (FC) networks. For many years these networks were built in parallel and were accompanied independently. The Data Center Bridging (DCB) and Fiber Channel over Ethernet (FCoE) protocols allow you to combine the functions of both on a single set of equipment, which saves capital and operational costs for the infrastructure of the Data Center.
Ethernet is undoubtedly the most popular networking technology today. During the existence of Ethernet, the maximum data transfer rate increased from 10 Mbit / s to 100 Gbit / s. Ethernet has pushed many other protocols out of the market due to its simplicity, low cost and broad support for the network industry. In Ethernet networks, the loss of transmitted data is permissible, which is absolutely unacceptable for Fiber Channel networks. DCB is an Ethernet enhancement technology that provides lossless data transport. FCoE - FC transport protocol over "enhanced" Ethernet.
Although DCB and FCoE are used together to build a holistic solution, these are two different technologies, and we will consider them separately
What is required from Ethernet to ensure its applicability for FC transport? This question has been the subject of close attention by the networking industry and the development of standards over the years, and several terms have emerged to denote a set of Ethernet enhancements for data center needs:
• CEE (Converged Enhanced Ethernet) - a term proposed by IBM and actively used in the framework of the standardization process of Fiber Channel over Ethernet
• DCB (Data Center Bridging) is a term used in the process of standardizing various Ethernet enhancements in IEEE organization committees.
• DCE (Data Center Ethernet) is a term used by Cisco to designate a new generation Ethernet architecture for data center tasks.
')
The term “Data Center Bridging” will be used to denote “improved Ethernet” in the following article.
The main components of the Data Center Bridging solution include:
• Providing lossless transport (Ethernet). To do this, you must fulfill two conditions:
o The internal architecture of the switch should provide lossless data switching . In particular, on Cisco Nexus switches, this is accomplished using the concept of VOQ (Virtual Output Queuing - organization of virtual output queues), which provides for the transfer of data through the switching matrix to the output port only if it is able to receive this data
• There must be flow control mechanisms that will not allow more information of this type to be sent to the network than the network can transmit at a particular point in time (similar to what is provided in Fiber Channel networks through the Buffer-to-buffer Credits mechanism). DCB (Priority-based Flow Control, IEEE 802.1Qbb) was created in DCB, which functions separately for each traffic class and allows you to suspend the transfer of storage network data without affecting the control data and client-server sessions. Manage bandwidth available to different types of traffic. This is necessary in order to determine the levels of service provided for the different types of traffic on the consolidated network, and thereby make it possible for them to be transferred together without sacrificing quality. This problem is solved in the DCB architecture using the ETS (Enhanced Transmission Selection, IEEE 802.1Qaz) mechanism.
• Automatic detection of settings of interacting devices. This is necessary to automatically provide consistent flow control and dedicated band settings for different types of traffic, the logical state of “virtual lanes”, etc., as well as to determine whether the cooperating device supports DCB functions, i.e. define the DCB “domain” boundary within an Ethernet network. This task is solved by the DCBX protocol (Data Center Bridging Exchange), developed in the framework of IEEE 802.1Qaz.
Fiber Channel protocol information over the "enhanced" Ethernet in the FCoE architecture occurs with full preservation of the frame structure of the Fiber Channel - Fiber Channel frames are equipped with additional headers and are placed inside the Ethernet frames.
It should be noted that, from the Fiber Channel frame transfer (encapsulation) requirement, there are two significant additional FCoE requirements for DCB transport:
• Support for large Ethernet frames . Since the maximum frame size of a Fiber Channel exceeds 2 kilobytes, it cannot be encapsulated into a standard Ethernet frame of up to 1518 bytes, therefore, support of larger frames — so-called jumbo frames — is required to transfer Fiber Channel over Ethernet.
• Work at a speed of 10 Gbit / s and above . Modern servers, as a rule, are connected to a storage network at a speed of 4-8 Gbit / s, therefore, naturally, we cannot use speeds of 1 Gbit / s or lower for consolidated transport.
Today, FCoE together with DCB are widely used, especially at the level of connecting servers to the data center network, making it possible to halve the number of network adapters and switches and reduce cable infrastructure. Technology I / O consolidation is an effective way to reduce the cost of creating and operating data centers, increases the flexibility and speed of deployment of new services.
Additional materials:
Cisco data center switches
Cisco data center solutions
Ethernet is undoubtedly the most popular networking technology today. During the existence of Ethernet, the maximum data transfer rate increased from 10 Mbit / s to 100 Gbit / s. Ethernet has pushed many other protocols out of the market due to its simplicity, low cost and broad support for the network industry. In Ethernet networks, the loss of transmitted data is permissible, which is absolutely unacceptable for Fiber Channel networks. DCB is an Ethernet enhancement technology that provides lossless data transport. FCoE - FC transport protocol over "enhanced" Ethernet.
Although DCB and FCoE are used together to build a holistic solution, these are two different technologies, and we will consider them separately
Data Center Bridging (DCB) technology
What is required from Ethernet to ensure its applicability for FC transport? This question has been the subject of close attention by the networking industry and the development of standards over the years, and several terms have emerged to denote a set of Ethernet enhancements for data center needs:
• CEE (Converged Enhanced Ethernet) - a term proposed by IBM and actively used in the framework of the standardization process of Fiber Channel over Ethernet
• DCB (Data Center Bridging) is a term used in the process of standardizing various Ethernet enhancements in IEEE organization committees.
• DCE (Data Center Ethernet) is a term used by Cisco to designate a new generation Ethernet architecture for data center tasks.
')
The term “Data Center Bridging” will be used to denote “improved Ethernet” in the following article.
The main components of the Data Center Bridging solution include:
• Providing lossless transport (Ethernet). To do this, you must fulfill two conditions:
o The internal architecture of the switch should provide lossless data switching . In particular, on Cisco Nexus switches, this is accomplished using the concept of VOQ (Virtual Output Queuing - organization of virtual output queues), which provides for the transfer of data through the switching matrix to the output port only if it is able to receive this data
• There must be flow control mechanisms that will not allow more information of this type to be sent to the network than the network can transmit at a particular point in time (similar to what is provided in Fiber Channel networks through the Buffer-to-buffer Credits mechanism). DCB (Priority-based Flow Control, IEEE 802.1Qbb) was created in DCB, which functions separately for each traffic class and allows you to suspend the transfer of storage network data without affecting the control data and client-server sessions. Manage bandwidth available to different types of traffic. This is necessary in order to determine the levels of service provided for the different types of traffic on the consolidated network, and thereby make it possible for them to be transferred together without sacrificing quality. This problem is solved in the DCB architecture using the ETS (Enhanced Transmission Selection, IEEE 802.1Qaz) mechanism.
• Automatic detection of settings of interacting devices. This is necessary to automatically provide consistent flow control and dedicated band settings for different types of traffic, the logical state of “virtual lanes”, etc., as well as to determine whether the cooperating device supports DCB functions, i.e. define the DCB “domain” boundary within an Ethernet network. This task is solved by the DCBX protocol (Data Center Bridging Exchange), developed in the framework of IEEE 802.1Qaz.
Fiber Channel over Ethernet (FCoE) technology
Fiber Channel protocol information over the "enhanced" Ethernet in the FCoE architecture occurs with full preservation of the frame structure of the Fiber Channel - Fiber Channel frames are equipped with additional headers and are placed inside the Ethernet frames.
It should be noted that, from the Fiber Channel frame transfer (encapsulation) requirement, there are two significant additional FCoE requirements for DCB transport:
• Support for large Ethernet frames . Since the maximum frame size of a Fiber Channel exceeds 2 kilobytes, it cannot be encapsulated into a standard Ethernet frame of up to 1518 bytes, therefore, support of larger frames — so-called jumbo frames — is required to transfer Fiber Channel over Ethernet.
• Work at a speed of 10 Gbit / s and above . Modern servers, as a rule, are connected to a storage network at a speed of 4-8 Gbit / s, therefore, naturally, we cannot use speeds of 1 Gbit / s or lower for consolidated transport.
Today, FCoE together with DCB are widely used, especially at the level of connecting servers to the data center network, making it possible to halve the number of network adapters and switches and reduce cable infrastructure. Technology I / O consolidation is an effective way to reduce the cost of creating and operating data centers, increases the flexibility and speed of deployment of new services.
Additional materials:
Cisco data center switches
Cisco data center solutions
Source: https://habr.com/ru/post/229891/
All Articles