VHT Metrics Considerations April 2008 doc.: IEEE 802.11-08/0465r2 April 2008 VHT Metrics Considerations Date: 2008-04-18 Authors: Darwin Engwer, Nortel Networks Darwin Engwer, Nortel Networks

April 2008 doc.: IEEE 802.11-08/0465r2 April 2008 Abstract This presentation reviews some important considerations in the development of an analytical framework for VHT. In particular, relevant metrics are discussed. Please view in slideshow mode to see animations. Darwin Engwer, Nortel Networks Darwin Engwer, Nortel Networks

Outline Objectives Definition Analytical Framework – Part I April 2008 Outline Objectives Definition Analytical Framework – Part I BSS Maximum Throughput Metric (M1) Analytical Framework – Part II Multiple BSS Maximum Throughput Metric (M2) Analytical Framework – Part III Maximum BSS Throughput Metric in space (M1) … Darwin Engwer, Nortel Networks

April 2008 Metrics Objectives Devise key metrics for characterization and comparison of different wireless systems. Darwin Engwer, Nortel Networks

April 2008 doc.: IEEE 802.11-08/0465r2 April 2008 What are metrics, and how do they differ from system measurement and performance reports? Metrics characterize a given system in terms of the abstract entities operating in a perfect environment. Metrics are computations that expose the capabilities of a given system. Performance reports (TGT) are the outcomes of specific tests run on a real world implementation operating in a specific, constrained environment for a specific set of use case scenarios. Measurement reports (802.11k) indicate the current state of an operating network implementation in the real world. [Metrics describe what the system is capable of under ideal conditions.] Performance reports indicate how well an implementation performs in a given scenario. These are useful on their own and also in relation to the best case values, the metrics. Measurement reports indicate how well an implementation performs in a real world setting, in real time. These are useful on their own and also in relation to the best case values (the metrics) and the performance reports for similar scenarios. Darwin Engwer, Nortel Networks Darwin Engwer, Nortel Networks

Analytical Framework – Part I April 2008 Analytical Framework – Part I Consider a universe; a void with nothing in it. In the middle of this void, place an AP with a set of associated STAs communicating with the AP. The AP is a point of access to a non-802.11 LAN of infinite capacity (through a DS and portal, not shown). A constellation of n STAs is associated with the AP. Each STA is attempting to exchange a maximum amount of data with an entity x on the non-802.11 LAN. The entity x has infinite throughput capacity. Darwin Engwer, Nortel Networks

A single AP in a void April 2008 April 2008 doc.: IEEE 802.11-08/0465r2 April 2008 A single AP in a void STA AP With a constellation of associated and active MUs. Darwin Engwer, Nortel Networks Darwin Engwer, Nortel Networks

Infrastructure BSS Maximum Throughput (M1) April 2008 Infrastructure BSS Maximum Throughput (M1) Now consider that collectively the AP and the set of STAs are capable of some MAXIMUM throughput. STAs BSS Throughput Max 1 2 3 4 5 Note: Data is for illustration purposes only. Darwin Engwer, Nortel Networks

Infrastructure BSS Maximum Throughput April 2008 Infrastructure BSS Maximum Throughput where, i = index of current STA n = number of associated and active STAs T = throughput at MAC SAP of STA i of n Darwin Engwer, Nortel Networks

Analytical Framework – Part II April 2008 Analytical Framework – Part II Consider a universe; a void with nothing in it. In the middle of this void, place a set of APs with a set of associated STAs communicating with each AP. Each AP is a point of access to a non-802.11 LAN of infinite capacity (through a DS and portal, not shown). A constellation of n STAs is associated with each AP. Each STA is attempting to exchange a maximum amount of data with an entity x on the non-802.11 LAN. The entity x has infinite throughput capacity. The APs are collocated. The APs operate on different channels within a set of k channels. Darwin Engwer, Nortel Networks

A set of collocated APs in a void April 2008 doc.: IEEE 802.11-08/0465r2 April 2008 A set of collocated APs in a void AP STA AP STA AP STA With a constellation of associated and active MUs. Darwin Engwer, Nortel Networks Darwin Engwer, Nortel Networks

Multiple Infrastructure BSS Maximum Throughput (M2) April 2008 Multiple Infrastructure BSS Maximum Throughput (M2) Now consider that collectively the set of APs and the associated sets of STAs are capable of some MAXIMUM throughput. APs ESS Throughput (Mbps) Max Note: Data is for illustration purposes only. Darwin Engwer, Nortel Networks

Multiple Infrastructure BSS Maximum Throughput (M2) April 2008 doc.: IEEE 802.11-08/0465r2 April 2008 Multiple Infrastructure BSS Maximum Throughput (M2) where, i = index of current STA n = number of associated and active STAs j = index of current AP m = number of APs k = number of channels T = throughput at MAC SAP of STA i of n wrt AP j of m (j, m, k determine number of APs on current channel) Note that a given *system* might employ multiple bands, e.g. dual concurrent 11b + 11a. Darwin Engwer, Nortel Networks Darwin Engwer, Nortel Networks

M2 examples April 2008 802.11a 802.11b Mbps Max Max APs ESS Throughput (Mbps) Max 802.11a Max 802.11b APs Note: Data is for illustration purposes only. Darwin Engwer, Nortel Networks

April 2008 Space The metrics M1 and M2 seek to quantify aspects of the maximum throughput capability of the system. In doing so, the M1 and M2 metrics ignore any consideration of space (or range) of the APs wrt the STAs. Darwin Engwer, Nortel Networks

Analytical Framework – Part III April 2008 Analytical Framework – Part III Consider a universe; a void with nothing in it. In the middle of this void, place an AP with a set of associated STAs communicating with the AP. The AP is a point of access to a non-802.11 LAN of infinite capacity (through a DS and portal, not shown). A constellation of n STAs is associated with the AP. Each STA is attempting to exchange a maximum amount of data with an entity x on the non-802.11 LAN. The entity x has infinite throughput capacity. The data rate varies depending on the distance between the AP and each STA. Darwin Engwer, Nortel Networks

An AP in a void, with zones at different rates April 2008 doc.: IEEE 802.11-08/0465r2 April 2008 An AP in a void, with zones at different rates STA AP With a constellation of associated and active MUs. Darwin Engwer, Nortel Networks Darwin Engwer, Nortel Networks

April 2008 M1 in space The M1 metric now expands from a single value to array of values, one per supported MCS, M1□. Each value in the array represents the maximum throughput achievable using the corresponding MCS, and hence reflects the maximum throughput within an abstract range boundary. Darwin Engwer, Nortel Networks

April 2008 Energy The metric M1□ seeks to quantify one aspects of the maximum throughput capability of the system. However, the M1□ metric ignores any consideration of the energy required to lift the bits into a space. Without this correlation the ability of a system to fill a given space with data in unknown. Darwin Engwer, Nortel Networks

Energy Factor [add energy factor notes here, see file 38] April 2008 doc.: IEEE 802.11-08/0465r2 April 2008 Energy Factor [add energy factor notes here, see file 38] When we consider energy we usually do so in terms of the transmit power (EIRP), but from a system level perspective the amount of work required to actually “lift” the bits into the space is a much more valuable metric. Work is measured in units of Watt seconds, aka Joules. … See file 38 Darwin Engwer, Nortel Networks Darwin Engwer, Nortel Networks

Latency [Add latency metric notes here] April 2008 Darwin Engwer, Nortel Networks

Metrics Sub-summary M1 – max BSS throughput M2 – max ESS throughput April 2008 Metrics Sub-summary M1 – max BSS throughput M2 – max ESS throughput M1□ - max BSS throughput vs MCS M1□e – max BSS throughput vs MCS vs energy L1 – latency metric Given those we can begin to consider parallel communications … Darwin Engwer, Nortel Networks

Direction Factor [add direction factor notes here] Uplink traffic April 2008 Direction Factor [add direction factor notes here] Uplink traffic Downlink traffic Combination of both uplink and downlink (show effect on the formulas) STA-to-STA traffic Darwin Engwer, Nortel Networks

Parallel Communications Metrics April 2008 Parallel Communications Metrics [add parallel communications metrics notes here] How to quantify various aspects of the system when parallel communications techniques are in use. Filling a room with data Note that with VHT we need to move away from a device centric view and instead look at what’s happening in the air. Our objective is to maximize utilize of the air (time, space, energy and frequency). Darwin Engwer, Nortel Networks

Impact of various parallel communications techniques April 2008 Impact of various parallel communications techniques Multi-user MIMO Downlink metric increases by number of independent streams factor Network coding Metric increases by total number of bits delivered, i.e. this includes the coded data Max value is 1.5x ?? Broadcast of group addressed data Metrics are unaffected; same data to same recipients yields no advantage Single hop relay Contribution to metrics is constrained to the min() of the inputs to and outputs from the relay station. MG ? Darwin Engwer, Nortel Networks

Conclusion Metrics are helpful to quantify various approaches. April 2008 Conclusion Metrics are helpful to quantify various approaches. More work is needed … Darwin Engwer, Nortel Networks

References April 2008 11-07-0412-01-0wng-looking-ahead-to-future.ppt 11-07-0419-01-0000-very-high-throughput-study-group.ppt 11-07-0574-01-0000-draft-par-and-5criteria-gigabit-tg.doc 11-07-0724-01-0vht-vht-study-group-thoughts.ppt 11-07-2863-00-0vht-how-should-we-manage-the-process-for-the-proposed-vht-activity.ppt 11-07-2866-00-0vht-vht-possibilities.ppt “Mobile Cooperation Usage Models”, IEEE 802.11 submission, 2008-01-13, Marc de Courville (Motorola) et al. 11-08-0081-02-0vht-mobile-cooperation.ppt “Below 6 GHz 11VHT PAR&5C’s Proposal”, IEEE 802.11 submission, 2008-03-17, Marc de Courville, Darwin Engwer et al. 11-08-0219-04-0vht-below-6ghz-11vht-par-5c-s-proposal.ppt “Infrastructure Mesh Broadband Wireless System: Example of Cooperative Wireless”, Intel Cooperative Wireless Workshop, 2007-04, N. K. Shankaranarayanan (AT&T) Darwin Engwer, Nortel Networks

Revisions r0 – 2008-04-16 For first presentation to VHT SG. April 2008 Revisions r0 – 2008-04-16 For first presentation to VHT SG. r1 – 2008-04-18 s3, s10, s12, s13: Changed “ESS” to “Multiple BSS”. s8, s12, s14: Added “illustrative data only” notes. s28: Added list of revisions. s30: Added equation for c. r2 – 2008-05-01 s8, s9, s12, s13: Clarified the BSS type as “Infrastructure BSS”. Darwin Engwer, Nortel Networks

April 2008 Backup Slides Darwin Engwer, Nortel Networks

Multiple BSS Maximum Throughput (M2) April 2008 doc.: IEEE 802.11-08/0465r2 April 2008 Multiple BSS Maximum Throughput (M2) where, i = index of current STA n = number of associated and active STAs j = index of current AP m = number of APs k = number of channels c = number of APs on channel of AP j of m APs T = throughput at MAC SAP of STA i of n with c APs on the channel Note that a given *system* might employ multiple bands, e.g. dual concurrent 11b + 11a. Darwin Engwer, Nortel Networks Darwin Engwer, Nortel Networks

Technology axes for VHT to consider April 2008 Technology axes for VHT to consider Darwin Engwer, Nortel Networks

Technology axes for VHT to consider April 2008 doc.: IEEE 802.11-08/0465r2 April 2008 Technology axes for VHT to consider Darwin Engwer, Nortel Networks Darwin Engwer, Nortel Networks