Design  | SFC  |  NOS  |  Code |  Diary  | WWW |  Essay |  Cisco | [Home]

 

 

[Bill's Home Page]
[ Old page]

Essays


The best stock to have in the following decades are:
1950s:
Polariod 8,266% gain
Avon 3,799% gain
1960s:
Mascot 10,177% gain
Xerox 5,146% gain
1970s:
Keystone 2,393% gain
Chemfirst 2,170% gain
1980s:
Circuit City 8,265% gain
Mark IV 6,989% gain
1990s:
Dell 89,820% gain
CMGI 87,603% gain
[Star Wars - The Text Version]
Are you sick of 3D graphics or slick Flash animation, well here's the site for you.

Related links

 


 

 

 

 

 

Click on me, if you just want to see the magazine

Date: 6 August 2001
Ed: Billatnapier
Source: Chapter 8, Mastering Computing
Price: £10 (UK)/ $14 (US)
 
This is a special edition of the Ethernet Agent, and presents some of the most asked questions relating to Ethernet networks. More information can be found in Mastering Computing.
[Show only magazine] [Previous edition][Next edition]
 
Question: I would like some discipline in the design of my network. Thus, what are the main design steps?
It is important to properly design your network, as incorrect planning can cause problems in the future. The basic steps are:
Analyze requirements. This involves understanding and specifying the requirements of the network, especially its major uses. If possible future plans should be incorporated. One of the key features is the band-width requirements and the size of the network.
Develop LAN structure. This step involves developing a LAN structure for these requirements. Typically in organizational networks this will be based on a star topology using Ethernet hubs/switches.
Set up addressing and routing. The final step involves setting up IP addresses and subnets to add structure.

DEAR NET-ED
Question: How do I try to limit the number of colli-sions on an Ethernet segment?
Ethernet collisions occur when two nodes try and transmit onto a network segment at the same time. When the transmitting nodes detect this, they transmit a jamming signal to the rest of the network. All the other nodes on the network detect this, and wait for one of the two colliding nodes to get access onto the network segment. These collisions reduce the overall bandwidth of the network segment. An important concept is the collision domain, which defines the physical distance by which a collision is propagated. Repeaters and hubs propagate collisions, but switches, bridges and routers do not. Thus if you want to reduce the amount of collision insert either a switch, a router or a bridge in a network segment.
 
DEAR NET-ED
Question: What are the main requirements in designing a network?
Business requirements.
Technical requirements. The main issues are media contention, reducing excessive broadcasts (routing tables, ARP requests, and so on), backbone requirements, support for real-time traffic and addressing issues.
Performance requirements.
This is likely to involve a network load requirement analysis for the typical loading on the network, and also for the worst-case traffic loading. This will determine the requirement for client/server archi-tectures. An analysis should also be made for the impact of new workstations being added to the network. It should also involve an analysis of the requirements for application software, especially in its bandwidth requirements. Multi-media applications tend to have a large bandwidth requirement, along with centralized database applications and file servers.
New application requirements.
Availability requirements. This defines the usefulness of the network, such as response time, resource availability, and so on.

 

Question: I have a local Ethernet hub which I connect to. How far can I run a cable from the hub to my computer?
If you use Cat-5 horizontal cable, you can get a maximum distance of 100m (if you were to use fiber cable you could get up to 400m). A hub can thus cover an area of 200 meters square (assuming that the hub is located in the center of the area).
 
DEAR NET-ED
Question: Will I do damage if I connect using incor-rectly wired cable, also how do I know that I've connected everything correctly? For example, I have a fiber cable which has two connectors and both are the same, how do I get the TX to the RX, and vice-versa?
It's unlikely that you will do any damage if you connect your cables round the wrong way, as all the inputs and outputs are electrically buffered. This allows them to sustain short-circuits, and incorrect wiring. The key of knowing if your con-nection is working is to look at the 'keep-alive' signal, which is typically a green LED on the NIC, hub, switch or router. If it is active, or flashing, you have made a proper connection.
With fiber-optic connections, the transceiver unit will activate two green LEDs when you have made a correct connection. If they are not active, swap the connections round and reconnect.
 
DEAR NET-ED
Question: How do security people protect a network against external hackers?

Well it's a bit problem. A loss of any data could cost a lot of money. In fact one of the biggest problems is actually due to employees within a company, who can more easily do damage to a system than an external person can. The main devices used these days for protecting a network include:
Firewalls. These are basically routers (which routes data packets over interconnected net-works), which runs firewall software which blocks certain incoming and/or outgoing net-work addresses, and also for certain network applications (such as FTP or TELNET).
Network switches. These are devices which can be used to isolate networks from each other, by software configuration.
Passwords and User IDs. All logins must be protected with a strong password and network login.




 
DEAR NET-ED
Question: How do I go about gathering the data to define the requirements of a new network?
The most important information that is required is the structure of the organization and how information flows between the units, as the designed network is likely to reflect this structure. The information will include:
Understand the current network (if one exists) es-pecially its strengths and weaknesses.
Gather information on geographical locations.
Determine current applications, and future plans for each site and for the organization.
Develop organizational contacts. These will be the important people who will be involved in the devel-opment of the network. A mixture of technical and business skills always helps. Technical people tend to be driven by technology ('it should transfer files faster', 'it's easier to install', and so on), whereas business people tend to be driven by applications ('I just want access to a good spreadsheet', 'I want to be able to send e-mails to anyone in the company', and so on). It is also important to get someone involved who has experience of legal matters, and/or someone involved in Personnel matters.
Determine the requirements for external network connections. This is an important decision as the se-curity of the whole network may depend on the choices made on the external connections. Many large companies have a single point of connection to the external Internet as this allows organizations to properly manage internal and external connections to the Internet.
Determine key objectives of the organization, espe-cially related to mission-critical data and mission-critical operations. These should have top priority over other parts of the network. For example a hospital would declare its ambulance service as a mission-critical unit, whereas the cuts and bruises unit (if there was one) would not be.
Determine who is in control of information services. This may be distributed over the organization or over centralized in an MIS (Management Information Service) unit.
DEAR NET-ED
Question: I've analyzed the traffic on the network, and I've found that a large portion of the network traffic is related to broadcasts. How can I reduce their effect?
Broadcasts are sent out when a node wants help from other nodes. Typically this happens when a node requires the MAC address for a known net-work address. The broadcast domain defines the physical distance by which a broadcast will be propagated. Hub, bridges and switches all propa-gate broadcasts, but routers do not. Thus, if you want to reduce the number of broadcasts on a network segment, insert a router, and it will intel-ligently route data packets into and out of a network segment without too many broadcasts (as the router handles external data routing).

Question: Did you mention something about a cat?
Yes. There are five categories of UTP cables defined in EIA/TIA-568A. Cat-1 is only suitable for telephone communications, Cat-2 supports up to 4Mbps, Cat-3 supports up to 10Mbps, Cat-4 supports up to 16Mbps and Cat-5 sup-ports up to 100Mbps.
DEAR NET-ED
Question: You say that there's two types of Cat-5 cables which can connect to a network. How do I know which one I need?

Well there's a straight-through cable and there's a cross-connect.
Straight-through
Normally if you're connecting to the front of a hub or a switch to a node, you need a straight-though cable. The ports are typically marked with an x. For example, with a 24-port switch:
1x, 2x, 3x . 24x
the x means that it is crossed-over internally.
Cross-connect
This is used when connecting between hubs and switches. It is also used to connect to computers directly to each other.
Understand now?

DEAR NET-ED
Question: The Internet seems to be moving so fast. How do they manage to move so fast?

I suppose you mean: How do they manage to develop new methods and techniques to quickly, on a world-wide basis? Well if they left it up to In-ternational Standards organizations, it would take years to standardize anything, as many companies would have to come together, and all the countries of the world would also have to agree on the standard. The Internet has a better way. Basically an expert is asked to draft a specification on a particular area, and then the post an RFC (Request For Comment) on the Internet. These are then read by anyone who has an interest, who can make a comment on the document. Finally the standard can be adopted, as many companies are keen to make their systems comply with the RFC standard. For example if you wanted to find-out about the POP-3 protocol, then you would search on the Internet for RFC1939.
Question: So where will Gigabit Ethernet be used?

It is unlikely that Gigabit Ethernet will be used for WANs, as it still suffers from too many problems, which get worse over large physical areas. It is most likely to be used for network backbones, and in campus area networks.
 
DEAR NET-ED
Question: In your previous reply you said that I should get someone involved with a legal or with a Personnel Department background. Why? Surely they will hold back the project.
Not at all. A network design team which only involves technical people may not understand the complete requirements of the customer. In organizations, the customer will include the organization, itself, the workers, and the actual customers of the organization. Every organ-ization must also meet its legal and moral re-sponsibilities.
Networks and the Internet have caused all sorts of legal problems, which are extremely difficult to fully understand. A badly planned network which breaches civil liberties can be very costly in the future. A major decision is whether users are monitored in the applica-tions they are running, and the files that they are accessing (and even how fast they typed). These could be used by management to deter-mine how efficiently a worker is completing their tasks. Another major problem is users who access the Internet, either for mindless surfing, or for illegal downloads. Legal-minded individuals will be able to sort out these per-sonal matters, especially when it relates to who is allowed access to information on certain users.
DEAR NET-ED
Question: How do I try to limit the number of colli-sions on an Ethernet segment?
 
Ethernet collisions occur when two nodes try and transmit onto a network segment at the same time. When the transmitting nodes detect this, they transmit a jamming signal to the rest of the network. All the other nodes on the network detect this, and wait for one of the two colliding nodes to get access onto the network segment. These collisions reduce the overall bandwidth of the network segment. An important concept is the collision domain, which defines the physical distance by which a collision is propagated. Re-peaters and hubs propagate collisions, but switches, bridges and routers do not. Thus if you want to reduce the amount of collision insert either a switch, a router or a bridge in a network segment.
DEAR NET-ED
Question: Can I use the OSI model to design my network?
Yes. The OSI model can split the network up into identifiable areas. These are:
 
Physical layer. Network media (typically Cat-5 cable or fiber-optic cable), hubs and repeaters. Cables are normally run conforming to the EIA/TIA-568A standard. This layer should allow for future expansion.
Data link layer. Switches and bridges. These devices will define the size of the collision and broadcast domains.
Network layer. Routers, addressing. This layer filters data packets between network segments.

DEAR NET-ED
Question: Which is best, enterprise servers or work-group servers?
Well it all depends on your organization. Enter-prise servers are typically used when all the users within an organization require access to a single resource, such as with electronic mail. Work-group servers provide local access to data and application programs, and isolate traffic around these servers. Workgroup servers should be physically located where they are most required. Typically enterprise servers require to be more centralized in their location, and are more robust than workgroup servers, as the whole organization depends on them. Mirror servers (servers which have exact copies of the main enterprise server) can be used with an enterprise in order to reduce data traffic to the main server.

DEAR NET-ED
Question: As usual, I'm confused. Sorry. You have said that Ethernet connections have a cross-over, but when I look at my patch cable, there isn't a cross-over, and pin 1 wires to pin 1, pin 2 to pin 2, and so on? Where's the crossover?
You're totally correct, and so am I. The standard Ethernet connection must have a cross-over to connect the transmit to the receive, and vice-versa, but most hubs implement the cross-over inside the hub. Thus all you need is a straight-through cable. I've listed the standard cross-over connections in below, but most of the time you do not need a cross-over when you're connecting to the front of a hub or a switch. It is only at the back of the hub that you may need a cross-over cable. If in doubt look at the 'keep-alive' LED. If it is off after you con-nect, it's likely that you've got the wrong cable (or the power isn't on, or you've not connected the other end, or the power isn't on the computer, and so on).
Question: Sorry to bother you again, but does it matter which port I connect my workstation to the hub with. Do I have to start from port 1, then port 2, and so on.

No. Hubs and switches are autosensing and auto-matically use the port that you connect to. You should hopefully see an LED become active when you connect to the port. You can also connect to a cascaded hub/switch to any one of the ports.


Question: I've got a dual 10/100 switching hub. Can I communicate at 100Mbps, even though I only have a 10Mbps networking card?

No. The switching hub will automatically sense the speed of your networking card, and use that rate. The great advantage of buying a dual speed switch is that you can upgrade your network card over time.
Question: Wow! One billion bits in a second from Ethernet using standard Cat-5 cable. Surely it isn't possible.

It is, and the IEEE has standardized it. If the IEEE says that something works, then it works. In fact with a Gigabit Ethernet switch you can get even higher bandwidths.