The Hierarchical Design Model
Cisco has used the three-level Hierarchical Design Model for years. The hierarchical design model divides a network into three layers:
Access: Provides end-user access to the network. In the LAN, local devices such as phones and computers access the local network. In the WAN, remote users or sites access the corporate network.
- High availability via hardware such as redundant power supplies and redundant supervisor engines. Software redundancy via access to redundant default gateways using a first hop redundancy protocol (FHRP).
- Converged network support by providing access to IP phones, computers, and wireless access points. Provides QoS and multicast support.
- Security through switching tools such as Dynamic ARP Inspection, DHCP snooping, BPDU Guard, port-security, and IP source guard. Controls network access.
Distribution: Aggregation point for access switches. Provides availability, QoS, fast path recovery, and load balancing.
- High availability through redundant distribution layer switches providing dual paths to the access switches and to core switches. Use of FHRP protocols to ensure connectivity if one distribution switch is removed.
- Routing policies applied, such as route selection, filtering, and summarization. Can be default gateway for access devices. QoS and security policies applied.
- Segmentation and isolation of workgroups and workgroup problems from the core, typically using a combination of Layer 2 and Layer 3 switching.
Core: The backbone that provides a high-speed, Layer 3 path between distribution layers and other network segments. Provides reliability and scalability.
- Reliability through redundant devices, device components, and paths.
- Scalability through scalable routing protocols. Having a core layer in general aids network scalability by providing gigabit (and faster) connectivity, data and voice integration, and convergence of the LAN, WAN, and MAN.
- No policies such as ACLs or filters that would slow traffic down.
A set of distribution devices and their accompanying access layer switches are called a switch block.
The Core Layer
Is a core layer always needed? Without a core layer, the distribution switches must be fully meshed. This becomes more of a problem as a campus network grows larger. A general rule is to add a core when connecting three or more buildings or four or more pairs of building distribution switches. Some benefits of a campus core are:
- Adds a hierarchy to distribution switch connectivity
- Simplifies cabling because a full-mesh between distribution switches is not required
- Reduces routing complexity by summarizing distribution networks
Small Campus Design
In a small campus, the core and distribution can be combined into one layer. Small is defined as fewer than 200 end devices. In very small networks, one multilayer switch might provide the functions of all three layers. Figure 1-1 shows a sample small network with a collapsed core.
Medium Campus Design
A medium-sized campus, defined as one with between 200 and 1000 end devices, is more likely to have several distribution switches and thus require a core layer. Each building or floor is a campus block with access switches uplinked to redundant multilayer distribution switches. These are then uplinked to redundant core switches, as shown in Figure 1-2.
Data Center Design
The core layer connects end users to the data center devices. The data center segment of a campus can vary in size from few servers connected to the same switch as users in a small campus, to a separate network with its own three-layer design in a large enterprise. The three layers of a data center model are slightly different:
- Core layer: Connects to the campus core. Provides fast switching for traffic into and out of the data center.
- Aggregation layer: Provides services such as server load balancing, content switching, SSL off-load, and security through firewalls and IPS.
- Access layer: Provides access to the network for servers and storage units. Can be either Layer 2 or Layer 3 switches.
Network Traffic Flow
The need for a core layer and the devices chosen for the core also depend on the type of network traffic and traffic flow patterns. Modern converged networks include different traffic types, each with unique requirements for security, QoS, transmission capacity, and delay. These include:
- IP telephony signaling and media
- Core Application traffic, such as Enterprise Resource Programming (ERP), Customer Relationship Management (CRM)
- Multicast multimedia
- Network management
- Application data traffic, such as web pages, email, file transfer, and database transactions n Scavenger class traffic that requires less-than-best-effort treatment
The different types of applications also have different traffic flow patterns. These might include:
- Peer-to-Peer applications such as IP phone calls, video conferencing, file sharing, and instant messaging provides real-time interaction. It might not traverse the core at all, if the users are local to each other. Their network requirements vary, with voice having strict jitter needs and video conferencing using high bandwidth.
- Client-Server applications require access to servers such as email, file storage, and database servers. These servers are typically centralized in a data center, and users require fast, reliable access to them. Server farm access must also be securely controlled to deny unauthorized users.
- Client-Enterprise Edge applications are located on servers at the WAN edge, reachable from outside the company. These can include email and web servers, or e-commerce servers, for example. Access to these servers must be secure and highly available.
No comments:
Post a Comment