This chapter is from the book
This chapter is from the book
This chapter is from the book
Switches
Fibre Channel switches are one of the most powerful components of a SAN. They are responsible for the efficient and high-speed switching of frames over a storage network. These switches are the basis of the switched Fabric topology, where the switches are interconnected to form the Fabric. The Fabric, in turn, can support numerous point-to-point connections, individual nodes, and arbitrated loops.
Unlike Fibre Channel hubs that are generally used to implement the arbitrated loop topology and extend the effective distance of a loop, Fibre Channel switches offer enhanced and more complex functionality. As a part of the Fabric, Fibre Channel switches are responsible for the following:
Providing a dedicated bandwidth of 100 Mbps and higher per port. This means that all switch ports can engage in separate transactions simultaneously without affecting other transactions.
Providing high-speed switching of frames from source to destination ports.
Providing frame flow control during communication. Switches use services such as buffer-to-buffer credit for this purpose. Buffer-to-buffer credit, also known as BB_Credit, determines the maximum number of buffers available per port if buffer flow control is in use.
Providing Fabric services, such as Fabric login, Simple Name Server (SNS), State Change Notification (SCN), and Registered State Change Notification (RSCN).
NOTE
See Chapter 6 for more information about Fabric services.
A typical Fibre Channel switch offers 8 to 16 ports. Therefore, a single switch can successfully support a small-scale SAN. A small-scale SAN can consist of a few hubs, application servers, database servers, and high-performance storage devices. A small-scale SAN can also be extended effectively by cascading two or more switches. However, the larger the number of cascaded switches, the greater the possibility that the cascaded links can end up as bottlenecks in switch-to-switch transactions. In this case, redundant links might prove to be an effective solution as they provide an alternate path between Fabric switches.
Types of Switches
Fibre Channel switches are generally classified into three categories—loop switches, Fabric switches, and directors. Loop switches are the most cost-effective category of switches that are used to connect an FC-AL to the Fabric. They support full-duplex data transfers and are ideal for low-bandwidth devices. Mostly, they are implemented for connecting legacy loop devices to the Fabric and tape consolidation. Typically, most loop switches support eight ports.
Fabric switches offer a high-speed category of switches that form the Fibre Channel Fabric. They support full-duplex data transfers that are not affected by simultaneous communi-cations. They can efficiently support small SANs and act as the consolidation point of larger SANs. Fabric switches are generally implemented in SANs where storage devices are distributed geographically, such as storage wide-area networks (SWANs). A typical Fabric switch offers 16 to 32 ports.
Directors are a high-bandwidth, high-availability (99.999%), and high-performance category of switches that offer fully redundant, hot-pluggable components that can include processors and various switching elements. As a result, these switches are considered to be the most reliable. Their downtime is estimated to be about five minutes or less in a year. This makes them the most expensive category of switches. Directors are implemented in mission-critical and performance-intensive applications where downtime must be as low as possible. They are also used to implement the backbone for enterprise SANs. Directors offer 32 or more ports.
The price of switches can range from about $7500 to $45,000, depending on the function-ality provided by the switch. Many vendors offer Fabric switches. IBM, Vixel, Compaq, HP, Gadzoox, Brocade Communications, Inc., Hitachi, and McData are some of the most well-known vendors in the field of Fibre Channel switches.
Cisco Systems, Inc., offers the Catalyst 6000 and Catalyst 4000 family of switches for storage networking solutions. These switches are described in the following sections.
Catalyst 6000 Family Switches
The Catalyst 6000 family of switches is actually a set of Catalyst 6500 and Catalyst 6000 series of switches. These switches were designed to offer the following features that are extremely important in gigabit Fibre Channel networks:
Extensive intelligent network services
Exceptional scalability
Optimal price and performance
High availability
Support to a wide range of interfaces that are used in Fibre Channel networks
Figure 4-10 shows a Catalyst 6000 family switch.
)
Figure 4-10 Catalyst 6000 Family Switch
The functionality of members of this family include the following:
High performance—The Catalyst 6513 switch from the Catalyst 6500 series offers switching speeds up to 210 million packets per second. The switch also offers 1152 10/100 and 388 gigabit ports per 7-inch rack.
Switch Fabric Module 2—The Catalyst 6500 family offers scalable switching bandwidth up to 256 Gbps, while ensuring the highest level of availability.
Content Switching Module (CSM)—The Catalyst 6000 and 6500 families offer CSM that helps in tracking network sessions and server load conditions in real-time. CSM also directs each session to the appropriate server.
Optical Services Modules (OSM)—The Catalyst 6500 series switches offer optical WAN interfaces, deep packet buffers, and PXF IP Services Processors for high-speed IP service applications.
Supervisor Engine 2—The Catalyst 6000 and 6500 families offer a CEF-based architecture, which helps in increasing the overall system performance to 100+ Mbps.
Cisco Express Forwarding (CEF)—The Catalyst 6500 family offers CEF, which in collaboration with Supervisor Engine 2 and Gigabit Ethernet Modules delivers next generation solutions for dynamic service provider and enterprise networks.
Gigabit Ethernet Switching Module—The Catalyst 6500 family offers Gigabit Ethernet Switching Modules that are ideal for gigabit backbones, server farms, and high-density wiring closets.
Intrusion Detection System (IDS)—The Catalyst 6000 family switches integrate switching and security functionality to provide comprehensive attack detection against unauthorized and malicious access.
Quality of service (QoS)—The Catalyst 6000 and 6500 families offer high QoS, high-availability, and security.
NOTE
For more information on Catalyst 6000 series switches, refer to the site www.cisco.com/warp/public/cc/pd/si/casi/ca6000.
Catalyst 4000 Family Switches
The Catalyst 4000 family of switches offers modular switches (Catalyst 4003 and Catalyst 4006) that are an integral part of Cisco's Architecture for Voice, Video, and Integrated Data (AVVID) solutions. AVVID offers scalable and manageable solutions for both SAN- and NAS-based networks.
The main features of the Catalyst 4000 family switches include the following:
IP access to storage
Support to multimedia traffic, such as IP telephony, unified messaging, and Virtual Private Networks (VPNs)
Scalability and future growth
NOTE
For more information on Catalyst 4000 series switches, refer to the site www.cisco.com/warp/public/cc/pd/si/casi/ca4000.
Figure 4-11 shows a Catalyst 4000 family switch.
)
Figure 4-11 Catalyst 4000 Family Switch
Fibre Channel Switches Versus Hubs
While designing and implementing a storage network, one source of confusion is when to use hubs and when to implement switches. There are relevant considerations when deciding between hubs and switches including shared bandwidth, network population, services, security, and future growth. The following sections discuss these considerations.
Shared Bandwidth
Except for switched hubs, the bandwidth offered by managed and unmanaged hubs is shared. This means that if a 10-port hub offers 100 Mbps bandwidth, each individual hub port gets a bandwidth of only 10 Mbps. On the other hand, a 10-port 100 Mbps switch provides 100 Mbps bandwidth to each individual switch port.
Some applications, such as sustained full-motion video streams and streaming tape backups, are extremely sensitive to delays. As a result, hubs are not the right choice in storage environments where audio and video data is exchanged or tape backup data is streamed predominantly.
Network Population
The total number of nodes supported by the storage network also affects the choice of a hub or switch. If the number of nodes is smaller (5 to 40 nodes), hubs seem to be an ideal solution. However, for a large number of nodes (more than 100), switches are the best solution because the larger the network population, the greater the need of switching frames from source to destination. For networks that consist of 50–100 nodes, a mix of hubs and switches offers an optimal solution.
Services
In a large storage network populated with a large number of devices, discovering the required nodes or ports can be a tedious and complex task. Moreover, if a node goes down, other nodes might not know about the change in its status and try to contact it. This can result in long request queues for the failed node. Services, such as Fabric login, SCN, and RSCN, are very helpful in the quick and efficient discovery of the required disk, port, or node and their corresponding status. These services, however, are not obligatory, and nodes need to register themselves for notifications regarding changes in topology and individual device status.
The SCN service allows an N_Port to send a change notification to another N_Port that has registered its interest in any change of the given node or port. Therefore, only the port that was notified about the change would know about the change in status. The other members of the Fabric do not know about the change. RSCN offers a better solution to this problem. All the registered nodes are informed about the change simultaneously, thus allowing more than one node or port to know about the change. This helps in the fast discovery of the destination node or port.
Security
If security is a requirement in your storage network, it is advisable to use switches. This is because most of the hubs do not allow zoning, which is a common feature of Fibre Channel switches. Zoning prevents access to nodes or ports that do not belong to the same zone, thereby reducing the risk of unauthorized, malicious attacks to confidential corporate data.
Dynamic zoning also proves to be very helpful in networks where nodes (such as portable computers) are often physically relocated. Instead of moving hubs around and physically rewiring the entire setup, dynamic zoning allows the relocation of nodes without the need of physical rewiring. At the same time, the security setup is not affected.
Future Growth Potential
In a situation where the number of nodes in a storage network is small but is estimated to increase in the near future, switches prove to be an intelligent choice. Switches are highly scalable and have a higher potential for future growth. In contrast, scalability of hubs is limited because of shared bandwidth and security restrictions.