Search ITS   | User Guides |   | Services |   | Policies |   | Forms |   | Help |   | Home |  

TNS Services

TNS LAN Overview

• Short History
• TNS Supported LANS
• Current LAN Concept
• LAN Port Options
• Ethernet Switched Features
• Life Cycles and Equipment Status
• Internet 2 Ready
• Current LAN Design Guidelines

LAN Overview

A Short History of Ethernet LAN's:

In the relatively short time since computers were first connected together, Local Area Network (LAN) technologies and performance have improved substantially. The earliest Ethernet networks were constructed from a single length of coaxial cable that was tapped once for each network device. This style of interconnection appeared in the early 1980s, when computer networking was accomplished with Thicknet (known now as 10BASE-5) and later using IEEE 802.3 standards. The mechanical process of interconnecting computers improved slightly with the adoption of Thinnet (known now as 10BASE-2) which eliminated the need to tap cable. Although interconnecting devices was much easier with Thinnet's coaxial connectors, this technology continued to string network devices together on a single length of coaxial cable. Layer 2 packets transmitted between devices utilizing either of these standards are received by all other devices on the cable. The group of devices (also called a segment) which can receive transmissions from all other connected devices is called a collision domain. A packet transmission protocol, CSMA/CD for these standards, was necessary to control the orderly transmission of information over a single collision domain. In this overview, the group of devices (segment) which can receive a Layer 2 broadcast is referred to as a broadcast domain. Additionally, the group of devices (segment) that can receive unicast Layer 2 packets not directly addressed to the device is referred to as a repeated segment.

Over time the single wire implementations were replaced by Ethernet repeated segment HUBs and RJ-45 style cabling (10BASE-T). The change to modular 10BASE-T components was a huge improvement in the methods used to interconnect connect LAN devices, but the performance limitations of single broadcast domains and CSMA/CD remained. Local Area Networks grew to port densities where hundreds of computers would be sharing the same 10Mb broadcast domain. LAN administrators quickly discovered that large broadcast domains were inconsistent with network performance and data privacy.

Layer 2 Ethernet bridges and Ethernet switches do not forward unicast packets out a port unless the destination device is located behind the port. Because of this feature, bridging and switching were two of the first methods used to limit the size of collision domains and repeated LAN segments. The deployment of switches and bridges brought increases in LAN performance and data privacy. As the price of Layer 2 switching technology decreased and port densities on these switches increased, LAN administrators started to deploy switches to the very edge of the network. Although repeated hubs can still be found in smaller networks and SOHO applications, Ethernet switches have almost entirely replaced repeated hub devices in modern LANs.

Back to Top

TNS supported LANS:

Although TNS (nee OTC) did install and maintain Thicknet LANs (the first was Land and Water Building in spring 1987), the first designed, installed and maintained LANs were based on 10BASE-T technology and used repeated segment hubs and bridges. TNS continued to deploy repeated segment hub technology from various vendors (mostly from DSI, NCR, and 3Com) during the late 80s and most of the 90s. Some of these hub devices could be stacked and frequently hub stacks were populated to 192 ports.

Typical Shared Segment Hub Stack

For a period time (1996 - 2000) TNS employed switches (Marconi ES-3810 and ESX-2400 and 3Com 3300) in its LAN designs that supported ATM uplinks and provided ATM Forum ELAN capable Ethernet ports at the edge . This ATM technology and protocol allowed LAN administrators to achieve high bandwidth intra-LAN links to the IB Core (OC-3 and OC-12) and also to segment their repeated segment networks. Similar to the repeated segment hubs, stacking in these switches was also desired. The selected ATM uplink / 10Mb Ethernet port switches had high densities (The ESX-2400 could be populated with 120 10/100Mb 10Base-T ports) which reduced the overall price/port for switched Ethernet LANs. As the advances in chip densities and backplane speeds of frame based Ethernet switch technology allowed it to compete with ATM based switches on price and performance, many vendors stopped providing ATM capable edge equipment and TNS changed its LAN models accordingly.

The first steps toward an all frame based Ethernet LAN occurred in the summer of 2000 when ATM uplinks and ATM Forum ELANs were discontinued for new LAN designs using the 12 and 24 port Ethernet 10/100BASE-T switches (3Com 3300 family). These switches, although still stackable, were used to break-up large repeated segments and collision domains of hub only LANs. They were also installed at the network edge to provide 10/100Mb desktop, workstation and server ports. Later that same year, TNS added the HP4000, an Ethernet only switch, to its offerings. This high density chassis based switch, which was not stackable, was used to replace many of the older repeated hub devices on TNS maintained LANs both at the LAN edge and at switch or repeated segment hub uplink aggregation points.

Back to Top

Current LAN concept:

The current TNS LAN design paradigm requires Ethernet switches that provide appropriate functions at two different network locations: the very edge of the network and at the aggregation point where the LAN connects to the IB router. The switch that is deployed at the LAN edge needs to provide a high density of ports for network end devices and also provide the capability of a high bandwidth uplink. This uplink can be directly connected to the IB router in a single switch LAN, or it can function as an inter-switch link to an aggregation switch. In a multi-switch LAN, the switch that directly connects to the IB router needs to be able to aggregate multiple inter-switch links without restricting the bandwidth between the links. In the past, TNS sought devices for the edge that could be connected together to provide large quantities of ports, and managed, as if it was a single switch or "stack." As the demand for larger bandwidths available from an edge device port to the IB backbone router increased, " stackable solutions" became undesirable from a performance and cost perspective. TNS's designs are using smaller devices at the edge and larger uplinks to the aggregation point instead of large port densities in an edge device. Our current offerings for the edge, while stackable, will not be stacked beyond two devices, and, only if the stack is located in a different telecommunications closet than that of the Aggregation switch.

While the need for the aggregation function has always existed in LANs with multiple Ethernet switches and repeated hubs, TNS has started to specify Ethernet switch devices selected specifically for the purpose of edge switch aggregation. These aggregation switches typically have a small number of high speed ports, copper and fiber, for inter-switch aggregation links or for connecting and centrally locating servers and high performance workstations. These switches also provide a high bandwidth switching fabric (also called a backplane) to allow non-blocking transmission of data between these inter-switch links and the IB uplink.

The current LAN design paradigm, with the edge and aggregation point concept is shown in the following diagram:

Back to Top

LAN Port Options:

At present TNS offers three categories of LAN edge ports: accountable, mobility and wireless. Accountable ports are normally deployed where the LAN administrator can control who's computer will be connected to a port and the IP address it will use. For accountable ports the LAN administrator is responsible for documenting the relationship between an IP address and a valid PSU faculty/staff/student who will be using the address at any given time. For areas where LAN ports will be used at different times by different members of the PSU community, mobility or wireless ports can be provided. Mobility ports are suitable for locations where PSU faculty/staff/students will bring a portable device, usually a laptop PC, and connect to the network via a network cable. The relationship between IP addresses and who is using them is administered and recorded by devices that provide the IP address (DHCP) and authenticate the person (PSU Access). Wireless ports, radio connections to access points, provide a more convenient method of connecting to a network as the device does not need to be connected to a wall port by a cable. The relationship between IP address and the person using the address is maintained in a similar manner to mobility ports. Both the mobility port and the wireless port are associated with two different TNS services as described below.

As laptop computers became more popular, members of the PSU community that used laptops quickly recognized the possibilities presented by the ability to connect their laptops to network connections outside their offices. ITS responded to customer requests for this type of port by developing the Mobile Computing Service. The Mobile Computing Service allows anyone with an Ethernet network card, a web browser and a valid PSU Access account to gain access to the Internet at wire speeds (10Mb half-duplex or 100Mb full-duplex depending on location). As mobile technology evolved to include wireless, ITS has enhanced the mobility concept with the ITS wireless service (ITS Wireless SecureNet).

The ITS Wireless SecureNet Service provides wireless connectivity to any wireless device capable of LAN communication using the 802.11b standard (WiFi) and an operating system compatible with the PSU anywhere service. Wireless connectivity allows the benefits of mobile computing without the need to be tethered to a wall outlet. Wireless technology, however, is a shared broadcast technology and does not provide the same bandwidths to an individual device as would be available over a wired switched Ethernet connection.

Back to Top

Ethernet Switch Features:

As described above, TNS uses two categories of switches in its LAN designs: those suitable for the very edge of the network and those suitable for the aggregation point. TNS also offers these switches in two levels of capabilities: Basic and Advanced. LANs constructed with the Basic switch provide raw connectivity, a set of security features consistent with AD-20, and packet forwarding based on best effort. The Advanced Features class of switches is targeted toward customers looking to the near future where packets may need to be classified and forwarded based on the importance of the packet. Under conditions of network congestion, these markings can be used to give packets a higher forwarding priority than others, which reduces delay and jitter. Alternatively, packets with lower forwarding priorities may be queued, increasing delay and jitter, or dropped depending on the severity of the congestion. TNS has edge switches capable of these Quality of Service (QoS) features now and is developing IB Core services that will utilize Differentiated Services (Diff Serv) Quality of Service throughout the PSU intranet.

Basic Features include:

• Wire Speed (non-blocking)
• 10Mb, 100Mb, and 1000Mb Ethernet User ports
• Basic Security Features
  Port Mirroring
  Anti-Snooping
  Ability to Enable/Disable ports
  Loopback Protection
• Multicast Support (IGMP Snooping)
• Low Cost
• IPv6 support

Advanced Features Include:

• Wire Speed (non-blocking)
• 10Mb, 100Mb, and 1000Mb Ethernet User ports
• Basic Features Security
• Multicast Support (IGMP Snooping)
• Packet Filtering
• Packet Classification with Diff Serv
• Diff Serv code point honoring
• Differentiated Physical Queueing on Uplink

Back to Top

Life Cycle and Equipment Phases:

All organization with TNS maintained components are strongly encouraged to develop their own LAN equipment life-cycle plans. Although life-cycling is often considered primarily a function of support - frequently driven by continued vendor support - it is equally important from the perspective of performance and features. Organization need to create a viable financial plan to accomplish the complete life cycle replacement of components as a part of the original plan for component purchase. Early adopters of technology may prefer a shorter life-cycle length while others may prefer longer timeframes. Note that the monthly TNS maintenance fee does not support life-cycling - it is merely the ongoing cost of component maintenance.

TNS will, from time-to-time, introduce new LAN components, intended to provide updated services and replacement of older devices providing the same services. This equipment procurement cycle often starts with TNS submitting an RFP for new LAN devices, with cycles as frequent as about two years. Once new components are selected, the existing components move to a new phase of their TNS life-cycle. The final phase is often dictated by the vendor of each components who have their own cycles that typically include an end-of-sale date and an end-of-support date (sometimes called end-of-life). TNS's goal is to introduce new products such that the vendor's end-of-support date does not precede the TNS end-of-support date. While TNS considers product and vendor viability when it selects a new device, the realities of the market may force a vendor to drop support for a product prematurely. In these cases, TNS may need to shorten the life of this device and announce a new TNS end-of-support date.

Components used to in LANs move through three phases during their individual life-cycles - Current, Supported, and Expiring. A more detailed explanation of life cycling and the equipment phases can be found at Life-cycle of Customer Owned Components.

The Current and Supported LAN equipment by category:

TNS LAN Equipment in Edge Use Date: 06/12/2003

TNS LAN Equipment in Aggregation Point Use Date: 06/12/2003

Back to Top

Internet 2 Ready:

Every year all the LANs connected to PSU's Integrated Backbone are evaluated for their Internet 2 readiness. There are a minimum set of LAN related features and performance that are necessary to realize the connectivity available through Internet 2. The criteria used to determine Internet 2 readiness are reviewed yearly and modified to match the advances in the capabilities of Internet 2. Currently a LAN must meet the following criteria to be Internet 2 Ready:

100Mb or larger IB uplink
All switched Layer 2 links
IGMPv2 capable switches (multicast capable)

Back to Top

Current LAN Design Guidelines:

Routed Layer 3 network vs Flat Layer 2 network

TNS maintains the view that routed layer 3 networks have advantages over flat layer 2 networks. Any PC connected to a LAN, even if no applications requiring network access are running, needs to spend CPU time inspecting layer 2 broadcast packets and all layer 2 unicast packets in a repeated segment. As such, the ideal size of a TNS designed LAN's layer 2 broadcast domain is 200 devices or less. TNS will not design new LANs with over 400 devices and strongly recommend that existing TNS maintained LANs larger than 600 devices be segmented by the use of additional backbone interfaces.

The 200 device goal for a layer 2 broadcast segment is consistent with the efficient use of IPv4 address space as a 256 address subnet is normally the largest IP space allocated. A LAN with more than 254 (253 if redundant backbone service applies) will need multiple IPv4 address ranges. This can cause performance degradation when intra-LAN packets need to cross subnets as they will need to traverse the same backbone uplink twice. In the Layer 3 model these packets will not be intra-LAN and although they still need to traverse the router, they occur in each backbone uplink only once.

Effective Bandwidth to the Edge

While there are many measures of LAN performance, a metric to size uplinks and LAN switches are limited. One that TNS uses can be related to the concept of oversubscription. Oversubscription describes the ratio of maximum connected load to available bandwidths at any given LAN location, an uplink or device backplane. As packets are aggregated by LAN switches that an edge device's packets travel through, the concept assumes that the offered load to a device's backplane or uplink will be less than the maximum connected load. This is because not all devices connected to a Switch will be running at full maximum capacity all the time. TNS uses 100 : 1 or less for edge device uplinks to aggregation switches, less than 5 : 1 for switch backplanes (although non-blocking switches are preferred) and less than 500 : 1 for the entire LAN to the IB backbone. We consider the oversubscription rate of a LAN device or uplink an reasonable (simple) predictor of delay, jitter and packet drops.

Another metric TNS uses is the "effective bandwidth to the IB core on a saturated network." This measure relates to bandwidth available to any station on a LAN, congestion on a link, or congestion in a device. Effective bandwidth, as TNS uses the term, refers to the bandwidth available to a single device or LAN link if all connected devices are all transmitting and receiving at maximum capacity and each is receiving an equal share of the maximum bandwidth available.

The following show how these metrics apply to the evaluation of four LANs. These LANs represent the progression a customer would have (or will) seen if TNS has maintained their LAN for the last several life cycles of LAN equipment.

Some examples:

Shared Segment Hubs

Shared Segment Hubs (199x - 2004)

The shared segment network above, with 353 devices attaches highlights some of the disadvantages to an entirely shared segment LAN. First, all devices share the same collision domain, which implies that each could get an effective bandwidth of 28Kb. In an actual network this bandwidth number will be about 30% less as a shared segment gets less efficient when it is operating near saturation. Also, all packets are seen by every device. Unless anti-eavesdropping hubs are used, this situation could present a data security and privacy issue. Ethernet switches are usually used to reduce the size of repeated broadcast domains as in the next example.

Repeated Segment Hubs and Switch (1994 - 2004)

Repeated Segment Hubs and Switch

In the repeated segment hubs and switch example, the 24 port hub is replaced with an Ethernet switch. Now, instead of one large repeated segment, there is one large broadcast domain (373 devices) and three repeated segments. Note that the effective bandwidth a port at the edge of the network receives is dependent on the size of the segment the port is on. Also, as the Ethernet switch has its own ports, these ports have a much larger (almost 200 times that of the ports in the 192 port stack) effective bandwidth to the Integrated Backbone. This is a good location for servers, both local and internet accessible, and for workstations requiring larger bandwidth than most other devices. To get more effective bandwidth to the edge of the network, switches are needed at edge and larger interconnecting links are required. The next example shows a fully switched network.

Switched Ethernet Networks (1994 - )

This is an example of our previous design paradigm. Ports are available at the aggregation switch for servers or workstations needing larger bandwidths toward the backbone, or centralized connections to devices at the LAN edge. The devices connected to the 76 port switch are near the minimum effective bandwidth to the IB core. It should be noted that they have a much larger effective bandwidth to both devices at the aggregation point and on the same switch. Devices connected to the 41 port switch have an effective bandwidth to the core that is almost 20 times better than those on the 100Mb connected switch. To get more effective bandwidth to the edge required another paradigm change.

Switched Ethernet Network with Aggregation Switch. (2003 - )

This network represents our present paradigm. Both the specialized aggregation switch and the new edge switches are non-blocking. In designs that are constructed exclusively with the new switches only the device uplinks can be oversubscribed. All traffic between ports on the same switch can occur at line rates. The large effective bandwidths at the edge reflect the larger bandwidths to the aggregation switch. Please note: if servers or high performance workstations are directly connected to the aggregation switch, as they normally would be, the effective bandwidths to the edge devices are reduced.

Back to Top

The Pennsylvania State University © 2004. All rights reserved. Alternative Media Statement and Nondiscrimination Policy
This site maintained by Telecommunications and Network Services, a unit of Information Technology Services.
Provide site feedback to TNSWebmaster@mail.tns.its.psu.edu. 07/27/07