Campus Networks: GE or ATM?

Ahmet Tevfik İnan

Yıldız Technical University 80750 - İstanbul / TURKEY

Abstract. With the evolution of computer technology, computer networks became more popular. Especially the wide use of Internet and related technologies changed the LAN traffic model from the conventional 80/20 rule (80% of the traffic stays in the workgroup, 20% moves over the backbone) to 20/80 rule [1]. By the development of bandwidth hungry and delay sensitive applications, design and operability of the networks became a serious problem. This paper overviews physical media types, topologies, network technologies, and analyzes benefits and drawbacks of challenging technology like Gigabit Ethernet (GE) and Asynchronous Transfer Mode (ATM) in terms of technology and cost of ownership in the design stage of a new Campus Network.

1 Computer Networks

Interconnected collections of autonomous computers are named computer networks [2]. The benefits and drawbacks of computer networks are shown in Table 1.

Table 1. Benefits and Drawbacks of Computer Networks

Benefits

/ Drawbacks
  • Resource sharing
  • Incremental growth (ease to expand)
  • Placing power where it is needed
  • Autonomy
  • Redundancy
/
  • Backup issues
  • Security issues
  • Interoperability issues
  • System failure issues

To design a "campus network" based on open standards, the layered structure of OSI reference model must be followed. So we will first focus on different physical media characteristics and then go on with topology and LAN standards (media access technologies) concluding with a comparison of challenging networking technologies.

1.1 Physical Media

There is a wide range of media available to the network designer and the one chosen must suite, Environmental Requirements (speed and physical distance), Cost Requirements and Operational Requirements (modularity, extendibility, ease of installation and maintenance).

Coaxial cable is a type of communication cable in which a solid center conductor is surrounded by an insulated spacer, which in turn surrounded by a braided wire. An insulating layer then covers the entire assembly. The comparison of two types of coaxial cabling is shown in Table 2.

Table 2. Coaxial Cabling Specifications

Impedance (ohm) / Thickness (mm) / Transmission speed (Mbps) / Segment
length (m) / Max transceivers per segment
Thicknet / 50 / 10 / 10 / 500 / 100
Thinnet / 50 / 5 / 10 / 185 / 30

Twisted pair cables are so named because pair of wires are twisted around one another. Each pair consist of two insulated copper wire twisted together to reduce cross talks and noise effect. In data network applications, higher quality cables known as DGM (Data Grade Medium) must be utilized. 3 types of twisted pair cables are widely used in data networks. UTP (Unshielded Twisted Pair) has an impedance of 100 ohm. The categories used are shown in Table 3. ScTP (Screened Twisted Pair) is 100 ohm UTP with a single foil screen surrounding all 4-pair in order to minimize Electro Magnetic Interference (EMI) and outside noise. STP (Shielded Twisted Pair) is 150 ohm twisted pair cables individually wrapped in a foil shield and enclosed in an overall outer braided wire shield to eliminate EMI and outside noise.

Table 3. UTP Cabling Specifications [14]

Category / Cat-1 / Cat-2 / Cat-3 / Cat-4 / Cat-5 / Cat-5e / Cat-6 / Cat-7
Type / Voice / Data / Data / Data / Data / Proposal / Proposal / Proposal
Frequency(MHz) / - / 4 / 16 / 20 / 100 / 100 / 250 / 600

Table 4. Twisted Pair Cabling Benefits and Drawbacks

Benefits / Drawbacks
  • Simple, easy to install
  • Flexible
  • Low weight
  • Easy spliced and connected
/
  • Possible to tap
  • EMI (if not shielded)
  • Attenuation

Fiber optic cabling is the technology where electrical signals are converted into optical signals, transmitted through a thin glass fiber and reconverted back into electrical signals. Light source can be LED (Light Emitting Diode) or LD (Laser Diode). The specifications of are shown in Table 5.

Table 5. Fiber Optic Specifications [3]

5-10/125 SMF / 50/125 MMF / 62.5/125 MMF / 85/125 MMF / 100/140 MMF
Core / 5-10 / 50 / 62.5 / 85 / 100
Cladding / 125 / 125 / 125 / 125 / 140
Attenuation / 0.8 dB/Km / 3-4 dB/Km / 3,75-6 dB/Km / 5 dB/Km / 5-6 dB/Km
Frequency / > 1000 MHz / > 400 MHz / 160 MHz / 200 MHz / 10-100 MHz

MMF (Multi Mode Fiber) allows many paths of light to propagate down the fiber optic path. It has good coupling from inexpensive LEDs light sources and use of inexpensive couplers and connectors. SMF (Single Mode Fiber) has a core diameter that is so small that only a single mode of light is propagated. Small core of single-mode fiber makes coupling light into the fiber more difficult. So more expensive lasers must be used as a light source. Single-mode fiber is capable of supporting much longer segment lengths than multi-mode fiber.

Table 6. Benefits and Drawbacks of Fiber Optic Cable

Benefits / Drawbacks
  • Larger bandwidth
  • Longer distance
  • No EMI, no cross-talk, no attenuation
  • Impossible to tap
/
  • Difficult to install
  • Special connection techniques

Using radio waves as a communication media brings to mind the idea of mobility. But atmosphere poses some limitations due the changing weather conditions and national and international radio communication regulations. Frequencies between 300 GHz and 100 THz are called infrared [26].

Table 7. Radio Waves Frequencies [26]

Frequency
3 KHz / 30 KHz / VLF (Very Low Frequency.) / Long Range Communication.
30 KHz / 300 KHz / LF (Long Frequency)
300 KHz / 3 MHz / MF (Middle Frequency) / Multi Dimensional
Radio Communication
3 MHz / 30 MHz / HF (High Frequency)
30 MHz / 300 MHz / VHF (Very High Frequency)
300 MHz / 3 GHZ / UHF (Ultra High Frequency)
3 GHz / 30 GHz / SHF (Super High Frequency) / Microwave Communication
30 GHz / 300 GHz / EHF (Extremely High Frequency) / Space Communication

Table 8. Benefits and Drawbacks of Radio Waves

Benefits / Drawbacks
  • No physical link
  • Node can be mobile
/
  • Depends on atmospheric conditions

Table 9. Benefits and Drawbacks of Infrared

Benefits / Drawbacks
  • No physical link
/
  • In door use
  • Shadowing

1.2 Topology

The pattern of interconnection of nodes in a network is called topology. The key issues in the topology are cost, flexibility and reliability. The benefits and drawbacks of Star, Bus, Ring, Tree and Mesh topologies are shown in Table 10.

Table 10. Benefits and Drawbacks of Various Topologies

Benefits / Drawbacks
STAR /
  • Ease of service
  • One device per connection
  • Centralized diagnosis,problem determination
  • Simple access
  • Long cable length
/
  • Central node dependency
  • Wiring closet space required

BUS /
  • Short cable length
  • Simple wiring layout
  • Easy to extend
/
  • Adding a station stops network
  • Fault diagnosis is difficult
  • Fault isolation is difficult
  • Broken cable causes failure
  • Limited number of users (shared media)
  • Must listen the media (CSMA/CD)

RING /
  • Short cable length
  • No wiring closet required
  • Suitable for optical fiber
/
  • If node fails, network fails
  • Difficult to diagnose faults
  • Difficult to reconfigure
  • Topology affects the access

TREE /
  • Easy to extend
  • Fault isolation is simple
/
  • Depends on the root

MESH /
  • Redundant
/
  • Routing problem
  • Complex cabling/interfacing

2 LAN Standards (Media Access)

2.1 Ethernet

Announced as a product in 1980. But the development work dates backs to the early 70's. (Created by XEROX in 1972. In 1980, Digital, Intel and Xerox - DIX [3] developed Ethernet Version 2.0). Many of the original ideas come from the ALOHA wide area network used in the University of HAWAII. [4] Ethernet is a base-band network with a bus topology and data transmission rate of 10 Mbps. The access method is CSMA/CD (IEEE 802.3) In a CSMA/CD environment a station can access the network any time. Before sending data, stations listen to the network until no traffic detected. A collision occurs when two stations transmit simultaneously. In this situation both transmission are damaged and stations must retransmit at some later time. A back-off algorithm determines when the colliding stations should retransmit. The 10Base Ethernet Specifications is shown in Table 11.

Table 11. 10Base Ethernet Specifications [5]

10Base5 / 10Base2 / 10BaseT / 10BaseFL
Transmission speed / 10 Mbps / 10 Mbps / 10 Mbps / 10 Mbps
Media / Coax (50) / Coax (50) / UTP / Fiber(62,5)
Topology / Bus / Bus / Star / Star
Devices attached to / Transceiver / NIC by BNC / Hub/Switch / Hub/Switch
Segment length (m) / 500 / 185 / 100 / 2000

2.2 Fast Ethernet

The IEEE Higher Speed Ethernet Study Group was formed to assess the feasibility of running Ethernet at speeds of 100Mbps. The Study Group established several objectives for this new higher speed Ethernet but disagreed on the access method. An issue was whether this new faster Ethernet would support CSMA/CD to access the network medium or some other access method. The Study Group divided into two camps over this access method disagreement: The Fast Ethernet Alliance and the 100VG-AnyLAN Forum. Each group produced a specification for running Ethernet at higher speeds: 100BaseT and 100VG-AnyLAN respectively [5].

100BaseT. 100BaseT use the existing IEEE 802.3 CSMA/CD specifications. It supports all applications and networking software currently running on existing Ethernet infrastructure. In addition, 100BaseT supports dual speeds of 10 and 100Mbps, and using Fast EtherChannel technology it is possible to achieve 4x100Mbps bandwidth (800Mbps full duplex). The 100Base Ethernet Specifications is shown in Table 12

Table 12. 100Base Ethernet Specifications [5]

100BaseT4 / 100BaseTX / 100BaseFX
Transmission speed / 10/100 Mbps / 10/100 Mbps / 10/100 Mbps
Topology / Star / Star / Star
Devices attached to / Hub/Switch / Hub/Switch / Hub/Switch/PP
Number of repeaters / 2 / 2 / 1
Segment length (m) / 100 / 100 / 400
Cable type / 4 Pair UTP/ScTP
Cat 3-4-5 / 2Pair UTP/ScTP
Cat 5 / Fiber
(MMF 62.5/125)

Benefits and drawbacks of using Fast Ethernet over is shown in Table 13.

Table 13. Benefits and Drawbacks of Fast Ethernet Technology

Feature / Benefits / Drawbacks
True Ethernet Technology / Compatible with 10 Mbps / Reduces number of allowable repeater in a segment
Auto Negotiation / Transparent interoperability in mixed Ethernet environments / Full feature not implemented by all vendors
10/100 Scalability / Transparent migration from 10Mbps to 100Mbps / 100 Mbps operation require cable or hub upgrade
Uses same physical layer interface as FDDI / Proven functionality and compatibility / None
Supported by all major network component providers / Large selection of compatible equipment / None

100VG-AnyLAN. HP developed 100VG-AnyLAN as an alternative to CSMD/CD for newer time sensitive application such as multimedia. The access method is based on station demand. A node waiting to transmit signals its request to the hub/switch. If the network is idle, the hub immediately acknowledges the request and the node begins transmitting a packet to the hub. If more than one request received at the same time, the hub uses a round robin technique to acknowledge each request in turn. High priority requests, such video conferencing applications, are serviced ahead of normal priority requests [5].

Table 14. 100VG-AnyLAN Specifications

Node-Hub distance (m) / Hubs tree deep / End to end distance (m)
4-pair (Cat-3 UTP) / 100 / 3 / 600
2-pair (Cat-4/-5 UTP) / 150 / 3 / 900

2.3 Token Ring (TR)

IBM originally developed the Token Ring in the 1970s. Although dissimilar in some respects, IBM's Token Ring Network and IEEE 802.5 are generally compatible (Table 15). Stations are directly connected to MAU (Media Access Unit), which can be wired together to form one large ring. Token passing networks move a small frame, called token, around the network. Possession of the token grants the right to transmit. If a node receiving the token has no information to send, it passes the token to the next end station. Each station can hold the token for a maximum period of time. If the station possessing the token does have information to transmit, it seizes the token, alters one bit of the token, which turns the token into a start of frame sequence, appends the information it wants to transmit, and sends this information to the next station on the ring. While the information is circling the ring other stations wanting to transmit must wait. Therefore a collision cannot occur in Token Ring networks. The information frame circulates the ring until it reaches the intended destination station, which copies the information. The token continues to circulate and is removed when it reaches the sending station, then in turn checks if the frame is copied by the destination. In addition, a priority mechanism also exists in Token Ring networks [6].

Table 15. Token Ring Specifications [6]

Transmission speed (Mbps) / Media Type / 4-16 / Data Grade Medium
Ring diameter (m) / Maximum closet per ring / 800 / 12
Maximum Node-Closet distance (m) / 100 (in multiple closet ring)
Maximum Closet-Closet distance (m) / 200

2.4 FDDI (Fiber Distributed Data Interface)

FDDI is a standard defined by ANSI for 100 Mbps LAN communication using fiber optical medium designed as a backbone network and desktop delivery system for high-end workgroups. FDDI uses dual ring or dual ring to tree topology and a token passing media access method. Up to 1000 stations can be connected to an FDDI network, with up to 3km between stations. Compared to Fast Ethernet, FDDI is complex and costly to implement. Its advantage is its large domain diameter (which is 50 km) and inherent redundancy [7] (Table 16).

Table 16. Benefits and Drawbacks of FDDI

Feature / Benefit / Drawback
Stable Industry Standard / Multiple vendors available / Requires new hub, protocols and drivers. Expensive for desktop
Security / Difficult to tap.
Improved immunity / Requires fiber optic cable
Redundancy / Cable/device failures do not disturb critical applications / Requires expensive dual attach devices

2.5 ATM (Asynchronous Transfer Mode)

Organizing different streams of traffic in separate calls allows the user to specify the resources required and allows the network to allocate resources based on these needs [8]. The session's establishment, which is the most time consuming and complex process, is done once and fixed size cells are routed directly from source to the destination without any trouble. During the session establishment, traffic type specification brings the required prioritization to sustain the necessary Quality of Service (QoS). ATM operates from 25Mbps to 622 Mbps and serves as an excellent backbone merging Fast Ethernet and FDDI networks. It is also suitable for WAN connection, LAN backbone, and desktop networking technology. And it may become a protocol of choice for meeting the extremely high bandwidth requirements of real-time, bi-directional multimedia applications merging voice, video and data. ATM does not directly support the way LAN currently work, but instead must be configured to emulate conventional LAN operation [9]. ATM is a switched, connection-oriented LAN and WAN technology that allows a virtually unlimited number of users to have dedicated, high speed connections with each other and with high-performance network servers. There are three primary differences between ATM and conventional shared-media networks such as those based on Ethernet, Token Ring and FDDI. These three differences are ATM's use of dedicated media, its fixed length and its connection-oriented nature [10]. ATM tries to emulate Ethernet networks via LANE (LAN Emulation) and IPOA (IP over ATM).

2.6 Gigabit Ethernet (GE)

Gigabit Ethernet (GE) is an extension of IEEE 802.3 Ethernet standard. GE increases speeds tenfold over Fast Ethernet to 1000Mbps or 1Gbps. To accelerate from 100Mbps to 1Gbps, several changes need to be made to the physical interface. GE is identical to Ethernet from data link layer upward. To sustain necessary speed, IEEE 802.3 and ANSI X3T11 Fiber Channel technologies are merged (Fig. 1.). The MAC layer of GE is similar to those of standard Ethernet and Fast Ethernet. The MAC layer of GE supports both full and half-duplex transmission. The characteristics of Ethernet, such as collision detection, maximum network diameter, repeater rules, and so forth, are the same of GE. As it is in Fast EtherChannel technology, 4x1 GE can form a Gigabit EtherChannel to obtain 4 Gbps (8 Gbps full duplex) bandwidth. "GE is an evolution not revolution"[7].

3 Design Criteria for a Campus Network

As mentioned in the previous sections the major point to care about campus networks are, environmental requirements, cost and operational. In addition, Network Complexity must be taken into account to. Use of different protocols, vendors, products and type of applications brings the Network Complexity. But, it is always not possible to implement a single technology due to different needs. So it is usual to see heterogeneous networks, which takes advantage of all challenging technologies.

Fig. 1. Layered Structure of Gigabit Ethernet [5]

Table 17. 1000Base Ethernet Specifications

1000BaseLX / 1000BaseSX / 1000BaseCX / 1000BaseT
Laser type / LW / SW / N/A / N/A
Transmission media / SMF / MMF / MMF / Copper
(150 Ohm) / Copper
(100 Ohm)
Media diameter (micron) / 5/9 / 50 / 62,5 / 50 / 62,5 / N/A / N/A
Distance (m) / 1300 nm / 3000 / 550 / 550 / N/A / N/A / N/A / N/A
850 nm / N/A / N/A / N/A / 550 / 250 / N/A / N/A
2pair STP / N/A / N/A / N/A / N/A / N/A / 25 / N/A
Cat-5 / N/A / N/A / N/A / N/A / N/A / N/A / 100

3.1 Environmental Requirements

Intranet experiences have shown that networks are evolving. The Client/Server type distributed computing applications flipped the classical 80/20 rules [1]. In addition, support for more complex data streams such as graphics, animations and sound requires more bandwidth (Table 18).

Today the challenging technologies for campus networks bring 100Mbps FDDI, 100/200Mbps Fast Ethernet, 155/622Mbps ATM, 400/800Mbps Fast EtherChannel, 1000Mbps GE and 4/8Gbps Gigabit EtherChannel as common speeds on copper or fiber links within the range of 100m up to few kilometers.

3.2 Cost Requirements

The cost of network ownership is the sum of capital equipment, support staff and facility costs with a percentage of 48%, 36% and 18% respectively. The cost of ownership goes beyond the price/performance of initial capital equipment purchased and takes into account the long-term effects of technology change on support staff and facilities. As a consequence, the complexity drivers for two competing technology represents 31.2 person/month for the ATM implementation conversely 16.2 person/month for GE. This impact, called complexity inflation, is multiplied with fixed staff cost [11].

3.3 Operational Requirements

While designing the network infrastructure besides the speed, one had to take into account the non-blocking architecture and solve the over-subscription problem at distribution points in order to obtain the required performance. Can a device handle the traffic load that all its interfaces can accept or generate? Non-blocking means that a device's internal capacity matches or exceeds the full capacity bandwidth requirements of its ports and will not drop packet due to architecture [13]. To provide relative balance, non-blocking switches must support trunk link that are several times faster then the incoming links and their back plain capacity must be greater or equal to the switching load generated by all its ports.