Complete the following steps to determine the customer's network requirements:
Layer 2 switches forward broadcasts and multicasts, which becomes a scalability issue as flat switched networks become larger.
The network interface cards in a network station pass broadcasts and relevant multicasts to the CPU of the station. (Some NICS pass all multicasts to the CPU, even when they are not relevant, because they do not sallow you to select broadcasts and multicasts.)
The CPUs on network stations can become overwhelmed by processing broadcasts and multicasts. When investigating the effect of broadcast "radiation," Cisco's technical marketing group discovered that:
When provisioning hardware for an internetwork design, which we will
discuss in depth in Section 2 of Module 2,
be sure to take into consideration the broadcast behavior of the desktop
protocols. The following table shows recommendations for limiting the number
of stations on a single LAN based on the desktop protocol(s) in use. The
table also applies to the number of stations in a VLAN.
|
|
IP | 500 |
IPX | 300 |
AppleTalk | 200 |
NetBIOS | 200 |
Mixed | 200 |
The numbers shown are provided as a guide. Actual station limits depend on factors such as broadcast/multicast loads, IP addressing structure constraints, inter-VLAN routing requirements, management and fault isolation constraints, and traffic flow characteristics.
Avoid increasing the MTU to larger than the maximum supported for the media traversed by the frames in order to avoid fragmentation and reassembly of frames. When devices such as end nodes or routers need to fragment and reassemble frames, performance degrades.
For IP, use a protocol stack that supports MTU discovery. With MTU discovery, the software can dynamically discover and use the largest frame size that will traverse the network without requiring fragmentation. If your customers use IP implementations that support MTU discovery, make sure this feature is enabled. Some implementations default to a configuration with MTU discovery disabled.
The following table presents the importance of using maximum frame sizes.
|
|
|
|
1492 | 1518 (maximum) | 2.5% | 97.5% |
974 | 1000 | 3.8% | 96.2% |
474 | 500 | 7.4% | 92.6% |
38 (no PAD) | 64 bytes (minimum) | 50.0% | 50.0% |
1 (plus 37 bytes of PAD) | 64 bytes (minimum) | 98.7% | 1.3% |
In the IP suite, Transmission Control Protocol (TCP) supports windowing and flow control. Applications that run on top of TCP include the following:
Connectionless protocols usually do not implement error recovery. Most data link-layer and network-layer protocols are connectionless. Some transport-layer protocols, such as UDP, are connectionless.
Error recovery mechanisms for connection-oriented protocols vary. A good TCP implementation, for example, should implement an adaptive retransmission algorithm, which means that the rate of retransmissions slows when the network is congested. Using a protocol analyzer, you can determine whether your customer's protocols implement effective error recovery. In some cases, you can configure retransmission and timeout timers.
While reviewing these tables, remember that the true engineering solution
to most questions that characterize network traffic and performance is
"it depends." The data in the following charts is approximate and does
not take the place of a thorough analysis of the network in question.
|
|
E-mail message | 0.01 |
Spreadsheet | 0.1 |
Computer screen | 0.5 |
Document | 1 |
Still image | 10 |
Multimedia object | 100 |
Database | 1000 |
Note: The size of a computer screen depends on the type of screen, the number of pixels, and the number of colors. A "dumb" terminal application transfers much less data. Telnet, for example, sends each character that the user types in one packet. Responses are also very small, depending on what the user is doing.
A 3270 terminal application transfers about 4000 bytes, including characters and "attribute" bytes that define color and style.
The following table illustrates the amount of traffic overhead associated
with various protocols. Understanding the traffic overhead for protocols
can be a factor when choosing the best protocol for your network design.
|
|
|
Ethernet | Preamble = 8 bytes,
Header = 14 bytes, CRC = 4 bytes, Interframe gap (IFG) = 12 bytes |
38 |
802.3 with 802.2 | Preamble = 8 bytes,
header = 14 bytes, LLC = 3 or 4 bytes, SNAP (if present) = 5 bytes, CRC = 4 bytes, IFG = 12 bytes for 10 Mbps or 1.2 bytes for 100 Mbps |
46 |
802.5 with 802.2 | Starting delimiter = 1 byte, header = 14 bytes, LLC = 3 or 4 bytes, SNAP (if present) = 5 bytes, CRC = 4 bytes, ending delimiter = 1 byte, frame status = 1 byte | 29 |
FDDI with 802.2 | Preamble = 8 bytes,
starting delimiter = 1 byte, header = 13 bytes, LLC = 3 or 4 bytes, SNAP (if present) = 5 bytes, CRC = 4 bytes, ending delimiter and frame status = about 2 bytes |
36 |
HDLC | Flags = 2 bytes,
addresses = 2 bytes, control = 1 or 2 bytes, CRC = 4 bytes |
10 |
IP | With no options | 20 |
TCP | With no options | 20 |
IPX | Does not include NCP | 30 |
DDP | Phase 2 (long "extended" header) | 13 |
The first table shows the packets that a Novell NetWare client sends
when it boots. The approximate packet size is also shown. On top of the
packet size, add the data link-layer overhead, such as 802.3 with 802.2,
802.5 with 802.2, on FDDI with 802.2. Network-layer and transport-layer
overhead are already included in these examples. Depending on the version
of NetWare being run, the packets generated might be slightly different
than the ones shown here.
|
|
|
|
|
|
GetNearestServer | Client | Broadcast | 34 | 1 | 34 |
GetNearestServer response | Server or router | Client | 66 | Depends on number of servers | 66 if 1 server |
Find network number | Client | Broadcast | 40 | 1 | 40 |
Find network number response | Router | Client | 40 | 1 | 40 |
Create connection | Client | Server | 37 | 1 | 37 |
Create connection response | Server | Client | 38 | 1 | 38 |
Negotiate buffer size | Client | Server | 39 | 1 | 39 |
Negotiate buffer size response | Server | Client | 40 | 1 | 40 |
Log out old connections | Client | Server | 37 | 1 | 37 |
Log out response | Server | Client | 38 | 1 | 38 |
Get server's clock | Client | Server | 37 | 1 | 37 |
Get server's clock response | Server | Client | 38 | 1 | 38 |
Download login.exe requests | Client | Server | 50 | Hundreds, depending on buffer size | Depends |
Download login.exe responses | Server | Client | Depends on buffer size | Hundreds, depending on buffer size | Depends |
Login | Client | Server | 37 | 1 | 37 |
Login response | Server | Client | 38 | 1 | 38 |
The following table shows the packets that an AppleTalk station sends
when it boots. The approximate packet size is also shown. On top of the
packet size, add data link-layer overhead. Depending on the version of
Macintosh system software, the packets generated might be slightly different
than the ones shown here.
|
|
|
|
|
|
AARP for ID | Client | Multicast | 28 | 10 | 280 |
ZIPGetNetInfo | Client | Multicast | 15 | 1 | 15 |
GetNetInfo response | Router(s) | Client | About 44 | All routers respond | 44 if one router |
NBP broadcast request to check uniqueness of name | Client | Router | About 65 | 3 | 195 |
NBP forward request | Router | Other routers | Same | Same | Same |
NBP lookup | Router | Multicast | Same | Same | Same |
If Chooser started: | |||||
GetZoneList | Client | Router | 12 | 1 | 12 |
GetZoneList reply | Router | Client | Depends on number and names of zones | 1 | Depends |
NBP broadcast request for servers in zone | Client | Router | About 65 | Once a second if Chooser still open; decays after 45 seconds | About 3000 if Chooser closed after 45 seconds |
NBP forward request | Router | Other routers | About 65 | Same | Same |
NBP lookup | Router | Multicast | About 65 | Same | Same |
NBP reply | Server(s) | Client | About 65 | Depends on number of servers | Depends |
ASP open session and AFP login | Client | Server | Depends | 4 | About 130 |
ASP and AFP replies | Server | Client | Depends | 2 | About 90 |
Note: An AppleTalk station that has already been on a network remembers its previous network number and node ID and tries 10 times to verify that the network.node combination is unique. If the AppleTalk station has never been on a network or has moved, it sends 20 multicasts: 10 multicasts with a provisional network number and 10 multicasts with a network number supplied by a router that responded to the ZIPGetNetInfo request.
The following table shows the packets that a NetBIOS station sends when
it boots. The approximate packet size is also shown. On top of the packet
size, add data link-layer overhead. Depending on the version of NetBIOS,
the packets might be slightly different than the ones shown here.
|
|
|
|
|
|
Check name (make sure own name is unique) | Client | Broadcast | 44 | 6 | 264 |
Find name for each server | Client | Broadcast | 44 | Depends on number of servers | 44 if 1 server |
Find name response | Server(s) | Client | 44 | Depends | 44 if 1 server |
Session initialize for each server | Client | Server | 14 | Depends | 14 if 1 server |
Session confirm | Server | Client | 14 | Depends | 14 if 1 server |
The following table shows the packets that a TCP/IP station not running
DHCP sends when it boots. The approximate packet size is also shown. On
top of the packet size, add data link-layer overhead. Depending on the
implementation of TCP/IP, the packets might be slightly different than
the ones shown here.
|
|
|
|
|
|
ARP to make sure its own address is unique (optional) | Client | Broadcast | 28 | 1 | 28 |
ARP for any servers | Client | Broadcast | 28 | Depends on number of servers | Depends |
ARP for router | Client | Broadcast | 28 | 1 | 28 |
ARP response | Server(s) or router | Client | 28 | 1 | 28 |
The following table shows the packets that a TCP/IP station running DHCP sends when it boots. Although a DHCP client sends more packets when initializing, DHCP is still recommended. The benefits of dynamic configuration far outweigh the disadvantages of the extra traffic and extra broadcast packets. (The client and server use broadcast packets until they know each other's IP addresses.)
The approximate packet size is also shown. On top of the packet size,
add data link-layer overhead. Depending on the implementation of DHCP,
the packets might be slightly different than the ones shown here.
|
|
|
|
|
|
DHCP discover | Client | Broadcast | 576 | Once every few seconds until client hears from a DHCP server | Depends |
DHCP offer | Server | Broadcast | 328 | 1 | 328 |
DHCP request | Client | Broadcast | 576 | 1 | 576 |
DHCP ACK | Server | Broadcast | 328 | 1 | 328 |
ARP to make sure its own address is unique | Client | Broadcast | 28 | 3 | 84 |
ARP for client | Server | Broadcast | 28 | 1 | 1 |
ARP response | Client | Server | 28 | 1 | 28 |
DHCP request | Client | Server | 576 | 1 | 576 |
DHCP ACK | Server | Client | 328 | 1 | 328 |
Case Studies
In this section, you will complete our analysis of the customer's requirements and business constraints.
Read each case study and complete the questions that follow. Keep in mind that there are potentially several correct answers to each question.
When you complete each question, you can refer to the solutions provided by our internetworking experts. The case studies and solutions will help prepare you for the Sylvan exam following the course.
In this section, you will review the following case studies:
Now that you are familiar with CareTaker Publications' requirements,
answer the following questions. Refer to the readings in this section to
help you characterize the network requirements.
1. On a separate sheet of paper, diagram the
packet flow of information for the new e-mail system.
In the following space
write down any open issues or follow-up questions regarding the e-mail
system.
Now that you are familiar with PH Network Services' requirements, answer
the following questions. Refer to the readings in this section to help
you characterize the network requirements.
1. On a separate sheet of paper, diagram the
packet flow for the referral approval process.
2. Identify the customers performance requirements.
Now that you are familiar with Pretty Paper Ltd.'s requirements, answer
the following questions. Refer to the readings in this section to help
you characterize the network requirements.
1. On a separate sheet of paper, diagram the
packet flow of client/server traffic for the new business
software and traffic
for the pattern design documents for film production based on the requirements
as you understand
them.
2. Identify the customers performance requirements.
Now that you are familiar with Jones, Jones, & Jones' requirements,
answer the following questions. Refer to the readings in this section to
help you characterize the network requirements.
1. On a separate sheet of paper, diagram the
packet flow of information for the e-mail application.
2. On a separate sheet of paper, diagram the
packet flow of information for the CD-ROM Library
Pack application.
3. Identify the customers performance requirements.
If you are done with this section, click here to advance to Module 2.