Browse Prior Art Database

Capacity Aware, Speculative F-SCH Channel Assignment

IP.com Disclosure Number: IPCOM000007984D
Original Publication Date: 2002-May-09
Included in the Prior Art Database: 2002-May-09
Document File: 4 page(s) / 41K

Publishing Venue

Motorola

Related People

John M. Harris: AUTHOR [+2]

Abstract

Our current SCH assignment algorithm is not aggressive enough nor speculative enough when the system is lightly loaded. WAP Downloads, in general, do not get any benefit from the SCH in CDMA given current CDMA Scheduling Algorithm. Currently, in the infrastructure, the PCF/SDU generates a F-SCH request only after a threshold amount of data is in the PCF/SDU queue waiting to be sent to a user. SCH assignments are typically made for a duration of some msecs. The proposal is to send Speculative assignment requests (SAR) from SDU to TSM (Time Slice Manager at BTS). TSM only schedules speculative slices when extra resources remain after all real assignment requests for that slice have been satisfied, and when that slice is about to expire. A Time Slice expires if it is not assigned by BTS_TS LeadTime in advance. Time slice assignment requires a scheduling horizon. If it is not assigned before then, it goes to waste. It is these otherwise wasted slices that will be used by the speculative requests. Speculative requests will be based on opposite direction activity, dormant to active transition, and recent same direction traffic. A mechanism for controlling the degree of speculation is also planned. If 90% of speculative request have been denied, then SDU will be less speculative. If all of the speculative requests have been accepted, then SDU will be more speculative (up to a maximum amount of speculative-ness). This reduces the backhaul capacity hit due to overly-speculative (e.g. point less) requests. Results in roughly a 1 second decrease in the download time.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 33% of the total text.

Capacity Aware, Speculative F-SCH Channel Assignment

John M. Harris , Ivan N. Vukovic

Performance Analysis Department

Network Advanced Technology

Motorola

I.                   Problem:

WAP Downloads, in general, do not get any benefit from the SCH in CDMA given current CDMA Scheduling Algorithm.  WAP/Microbrowser Handheld traffic is 99.5+% of the data traffic.  Generally, the WAP traffic does not get ANY benefit from the SCH & high speed data channels of CDMA 2000.  WAP bursts are too small relative to SCH scheduling window horizon (< 1 Kbyte).  Consumers are very sensitive to the push to display delay of these services.  One component of this delay is transmission time (1.2 seconds is typical)

In CDMA 2000, when a download is initiated, the user must be put on channel, unless the user is already on channel.  When the user is on channel, we refer to the user as being in the active state. Typically, if the user is in the active state, and it neither sends nor receives data for IA seconds, then the user is transitioned off of the RF channel, (e.g. transitioned from active to dormant.  If the user is active, then it does not need to wait for the channel to be rebuilt the next time it has data to send or receive. However, while the user is in the active state, it is tying up Walsh code, channel element, and some RF resources (at least 10% of a full rate frame).  

   Figure 1 : CDMA 2000 Dormant to Active Setup Delay Depiction (Same as for EDGE/UMTS if replace Active With the word TBF)

After the user gets on traffic channel, 2 other things must happen before the download begins: a) the user must upload the request, b) the server response delay and internet round trip delay must occur.  As described above, because the transaction delay perceived by the user on the 1st push is significantly larger than it is on subsequent pushes, the operator tends to set the inactivity timer to be longer.  This results in a significantly larger RF impact than would be the case if the IA timer were 3 seconds. 

                  SCH scheduling mechanisms generally use buffer occupancy or input rate schemes for determining when to assign a SCH to a user.  For example, in the infrastructure, (PCF or SDU) when a threshold amount of data is in the queue waiting to be sent to a user or the incoming rate reaches a certain threshold, a request for a SCH channel for that user is generated.  SCH assignments are typically made for a duration of ~320 msecs.

                  However, the problem is that WAP users only download ~400 bytes to 1.5 Kbytes per download.  These downloads are not large enough to trigger SCH channel assignment.  The SCH channel assignment schemes were designed for laptop traffic.  However, laptop traffic is a fraction of data traffic.

                  Additionally, these SCH schemes are designed for systems where the SCH resources are very much in demand.  Thus, they are somewhat ‘miserly’ about assigning the SCH.  Also, there is a 400msec+ SCH assignment horizon or window.  The SCH assignment is always made at least 1 or 2 time slices in advan...