Service Definitions

Introduction

• Explanation of how many users of a shared packet channel can all experience performance levels that would normally be associated with a single user in a Switch circuit

• WaveRider 11 Mbps wireless channel as an example for describing the performance of a packet-based channel

Packet

• Data is transmitted in bursts called frames• Frames consist of overhead information,

necessary for synchronization and control, and the actual transmitted information

• Whole frame is considered one packet

Packet Channel

Start of TransmissionEnd of Transmission

Sent Info Channel Overhead

Channel Burst (Frame)

Sent Info Channel Overhead

Next Frame

Start of TransmissionEnd of Transmission

Inter Frame Space

Service Definitions

• So, what do the performance numbers look like when multiple users are receiving from the same channel?

• The simple answer is that it depends primarily upon the service!

Nature of Internet Traffic

Nature of typical Internet services:• normally very ‘bursty’ and asymmetrical• usage has been modeled by statistical

distribution (Poisson etc) which demonstrates how Internet traffic takes advantage of a packet channel

WEB Browsing

• “Heavy” Browsing: A person initiates a service request (clicks on web page). Typical of human generated traffic patterns, these requests are modeled according to a Poisson (Exponential) distribution (i.e. random events with a particular characteristic)

• “Heavy” browsing results in users generating requests on the average of once a minute, each request averaging 3K bytes and the server giving responses, each averaging 30K bytes.

Typical Service Levels

Average Service Eventsper Hour

Average InformationTransferred per event

Average Asymmetry(hub:remote)

Internet Browsing Approx. 10:1Low load 5 2.5K Bytes

High Load 60 30K BytesEmail send/receive Approx. 1:1

Low load 1 500 BytesHigh Load 10 2000 Bytes

FTP get/put Approx. 1:1Low load 1 5 K Bytes

High Load 10 50 K Bytes

Bandwidth Limiters & Slow Backhauls

• Transmission facilities that are lower performance than the EUM/CCU link will result in direct throughput reductions.

• A user cannot get more than 1.5 Mbps end to end throughput from an EUM whose traffic is routed through a T1 link.

• If a T1 is used to backhaul CCU traffic, then every EUM that is serviced by that CCU will not get better throughput than T1 rates.

• In fact, under load, Queues can build up that will result in delays that further reduce throughput.

Bandwidth Limiters & Slow Backhauls

• Unlike the common transmission facility, a bandwidth limiter can act on individual EUMs and individual users.

• Controlling particular EUMs or individual users service capability will increase the probability of the remaining ones maintaining a higher performance level.

• One can still not guarantee delay performance using the Bandwidth limiter.

Summary

• Understand the End-Users expectations when it comes to Internet services. Network design required to meet these expectations.

• Many of the factors affecting end user performance are not associated with the CAP/EUM combination. Operators and end users should be aware of this.

• Link factors are: Delay and Delay Variation; data throughput and availability are the important factors in determining “how fast is the link”.