●
Architecture implies a model
●
Design implies the actual network topology designed by
a person or persons
●
Five distinct areas of an enterprise network:
○
Core backbone
○
Campus
○
Data center
○
Branch/WAN
○
Internet edge
●
Characteristics for a good network design: Modular,
resilient, and flexible
●
Cisco campus architecture:
○
Access Layer – Used to grant user, server or edge
devices access. Usually incorporates switches with large amount of ports.
Features security, access control, filters, management, etc
○
Distribution layer – Aggregates wiring closets,
switches segment groups and WAN connections and isolates network problems in
campus. Acts as a service and control boundary between access and core layers
○
Core layer (backbone) – High-speed, usually a minimum
of 10 Gigabit Ethernet, high level of availability. Built for scalability and
fast convergence
●
Layer 2 switches can do some QoS and network access
control based on IP address.
●
Layer 4 switching implies switching based on protocol
sessions – uses not only source and destination IP addresses in switching
decisions but also IP session information contained in the TCP and UDP portions
of the packet.
●
Layer 7 switching is based on application information –
implies content-intelligence.
●
Layer 1 – electric signals, Layer 2 – Frames, Layer 3 –
Packet, Layer 4 – TCP or UDP segments – TCP or UDP port numbers, Layer 7 –
embedded application information in data payload
●
In Layer 2-only switches, ASICs handle frame
forwarding. Supports some Layer 3 funcationality such as IGMP snooping and QoS
marking
●
In layer 3 switching, provides routing domains. Packet
forwarding in layer 3 is handled by ASICs and other specialized circuitry.
Validates integrity of L3 packet header via L3 checksum, verifies and
decrements TTL, rewrites source and destination MAC addresses, updates Layer 3
CRC during L3 rewrite, processes and responds to any option information in the
packet such as ICMP record, updates forwarding statistics for network
management applications and applies security controls and classification of
service if required
●
Multilayer switches support QoS markings and limit the
scale of spanning tree by segmenting Layer 2
●
Cisco Catalyst 6500 - Scalable modular switch up to 13
slots, supports up to 16 10-Gig ethernet interfaces per slot in
over-subscription model, Up to 80 Gbps of bandwidth per slot, supports layer 2
and 3 features, optionally up to layer 7 features, designed with redundancy in
mind, supports Layer 3 NSF (Non-stop forwarding), etc
●
Cisco Catalyst 4500 - scalable module switch with up to
10 slots, supports multiple 10 Gigabit ethernet interfaces per slot, supports
layer 2 and 3 switching, and optionally supports redundancy for power and
supervisor slots
●
Cisco Catalyst 4948G, 3750, and 3560 family -
Fixed-port scenarios usually for access layer with up to 48 1Gbps access layer
ports and 4 10-Gbps ethernet interfaces for uplinks to distribution layer,
layer 2 and 3 switching, and not architected with redundancy
●
Cisco Catalyst 2000 family - Layer 2 usually, fixed
port with up to 48 1Gbps access layer ports and multiple 10Gbps ethernet
uplinks and not redundant
●
Nexus 7000 family - Premier data center switches,
modular switch with up to 18 slots, supports up to 230Gbps per slot, supports
Nexus OS (NX-OS), 10-slot chassis is built front-to-back, supports redundancy
●
Nexus 5000 and 2000 family - low-latency switches
designed for deployment in the access layer of the data center. Layer 2 only.
Nexus 5000 designed for 10Gbps ethernet applications and also support Fibre
Channel over Ethernet (FCOE)
●
Collapsed core = design combines the core and
distribution layers into a single area
●
Hardware switching = the act of processing packets at
any layers 2 through 7 via specialized hardware components referred to as
application-specific integrated circuits (ASIC)
●
ASICs can reach throughput at wire speed without
performance degradation for advanced features such as QoS marking, ACL
processing or IP rewriting
●
Many ASICs use specialized memory referred to as
ternary content addressable memory (TCAM)
●
Generally, the size of the ASIC memory is relative to
the cost and application of the switch.
Common Traffic Types:
|
Traffic Type
|
Description
|
Traffic Flow
|
BW
|
|
Network Management
|
I.e. BPDU, CDP, SNMP, SSH, etc
|
Traffic is found flowing in all
layers
|
Low
|
|
VoIP
|
2 Types - signaling between end
devices and data packets from voice itself. Data to and from IP phones is
usually on separate VLANs for QoS purposes
|
Generally moves from access
layer to servers in the core layer or data center
|
Low
|
|
IP Multicast
|
Sent from a particular source
address to a group MAC address. Switches need to be configured to keep this
traffic from flooding to devices that haven’t requested it and routers need
to ensure that multicast traffic is forwarded to the network areas where it
is requested
|
Market data applications are
usually contained within the data center. Other traffic such as IP/TV and
user data flows from access layer to core layers and to the data center
|
Very high
|
|
Normal data
|
Typical application data
|
Traffic usually flows from the
access layer to core layer and to the data center
|
Low to mid
|
|
Scavenger class
|
Includes all traffic with
protocols or patterns that exceed their normal data flows. Used to protect
the network from exceptional traffic flows that might be the result of
malicious programs. Also used for less than best-effort type traffic
|
Traffic patterns vary
|
Mid to high
|
●
Client-server farm applications apply the 20/80 rule -
20% traffic stays on LAN and 80% leaves segment for centralized servers,
internet, etc
●
Service-Oriented Network Architecture (SONA) is the
Cisco architectural approach to designing advanced network capabilities
●
SONA provides guidance, best practices and blueprints
for connecting network services and applications to enable business solutions.
It illustrates that the network is the common element that connects and enables
all components of the IT infrastructure
●
SONA outlines three layers of intelligence in the
enterprise network:
○
The Networked Infrastructure Layer - Where all the IT
resources are interconnected across a converged network foundation. The IT
resources include servers, storage and clients. The network infrastructure
layer represents how these resources exist in different places in the network.
Objective is for customers in this layer to have anywhere and anytime
connectivity
○
The Interactive Services Layer - Enables efficient
allocation of resources to applications and business processes delivered
through the network infrastructure
○
The Applications Layer - Includes business applications
and collaboration applications. Objective for customers in this layer is to
meet business requirements and achieve efficiencies by leveraging the
interactive services layer.
●
Borderless Networks - model enables businesses to
transcend borders, access resources anywhere, embrace business productivity,
and lower business and IT costs. Framework focuses more on growing enterprises
into global companies thus “borderless”
●
When applied to a building, the Cisco Campus
Architecture divides networks into the building access, building distribution,
and building core layers as follows:
○
Building access layer - Layer used to grant user access
to network devices. In the network campus, the building access layer generally
incorporates switched LAN devices with ports that provide connectivity to
workstations and servers. In a WAN environment, this layer at remote sites can
provide access to the corporate network across WAN technology. Access layer
provides HA, convergence and security
○
Building distribution layer - Aggregates the wiring
closets and uses switfches to segment workgroups and isolate network problems.
Availability, fast path recovery, load balancing and QoS are important at this
layer. Uses a combination of Layer 2 and multilayer swiches to segment
workgroups. Commonly used to terminate VLANs from access layer switches.
○
Building core layer/campus backbone - High speed
backbone designed to switch packets as fast as possible. Must be HA, high
speed, redundant, scalable, and adaptable to network changes. Redundant Layer 3
connections throughout for faster convergence. Should not perform any packet
manipulation in software at this level
●
Additional redundancy can be provided on layer 3
switches with first-hop redundancy protocols (FHRP) such as hot standby routing
protocol (HSRP)
●
Distribution layer - Aggregates and segments the access
layer, summarizes routing to access layer, always dual-connected to upstream
core layer, and optionally applies packet filtering, security features, and QoS
features
●
Why a core- Without it, distribution layer needs to be
fully meshed and this design isn’t scalable.
●
Recommended practice - deploy dedicated campus core
layer to connect three or more physical segments such as building in the
enterprise campus or four or more pairs of building distribution switches in a
large campus.
●
Campus core more scalable when using 10-Gigabit or 1
Gigabit density to scale and LAN convergence optionally with additional WAN and
MAN convergence
●
Small campus or large branch network = 200 end devices
or less
●
Medium-sized campus = 200-1000 end devices. Usually
access layer switches with uplinks to distribution multilayer switches.
●
Large campus networks = 2000+ end users.
●
Three layers of the data center design:
○
Core layer - Provides high-speed packet switching
backplane for all flows going in and out
○
Aggregation layer - Provides important functions such
as service module integration, Layer 2 domain definitions, spanning tree
processing, and default gateway redundancy
○
Access layer - connects servers physically to the
network
●
PPDIOO - Prepare, Plan, Design, Implement, Operate,
Optimize
No comments:
Post a Comment