University of Minnesota GeoFeed: A Location-Aware News Feed System Jie BaoMohamed F. Mokbel Chi-Yin Chow Department of Computer Science and Engineering.

University of Minnesota

GeoFeed: A Location-Aware News Feed System

Jie Bao Mohamed F. Mokbel Chi-Yin Chow

Department of Computer Science and EngineeringUniversity of Minnesota – Twin Cities

Department of Computer ScienceCity University of Hong Kong

2

Background

Become one of the most popular Web services!!!

Social NetworkingServices

(e.g., Facebook & Twitter)

3

■ News Feed function Display a set of messages/news from friends / subscribed

news agents

■ Examples: Social networking system, i.e., Facebook, Twitter

News Aggregators, i.e., My Yahoo!, iGoogle

What is News Feeds?

4

Motivation

■ Traditional News Feed Organized by either message issuing time, e.g.,

Twitter, or some user requirements, e.g., Facebook Spatial relevance is overlooked, user gets the same

news feed from different log on locations

■ Motivating Scenarios Travelling user is more interested in the

news/messages that are close to her current location to explore the new place

Stationary users may NOT be interested in the news/messages that are issued very far from their locationsIf the news feed functionality is aware of the inherent locations of users and messages, more relevant news

feed will be delivered

5

“Locations” in Existing Social Networking Systems

■ Unfortunately not “real” location awareness currently Share only user’s current location, e.g., Google Latitude Use location information as a tag , e.g., Facebook Place View all the messages in a spatial range, e.g., Twitter Nearby

Facebook PlaceGoogle Latitude Twitter Nearby

“Real” Location-Aware News Feed1. Social Relevance

Messages from friends/ subscribed news agents

2. Spatial Relevance Message relevant to the user’s location

6

Location-Aware News Feeds

■ Location-Based Messages Issuer: user/ news agent Spatial extent:

point/range Location-Aware News

Feeds

Recent k spatial relevant messages from each of my friendsM2

M3

M5

M4

M1

Message Content Spatial Timestamp

M6 Local Sale S6 15:30

M4 An accident S4 14:21

M1 Work finished S1 11:40

Message Content Spatial Timestamp

M5 Raining S5 14:30

M3 A nice bar S3 14:10

M2 Eating at bar S2 14:04

A location-based query is issued to retrieve the most recent k=2 relevant messages from Alice

A location-based query is issued to retrieve the most recent k=2 relevant messages from Bob

M6

Carol

Example:Carol wants her news feed from friends (Alice

and Bob)Alice’s Messages

Bob’s Messages

7

An Overview of GeoFeed

■ For a user U with N friends, GeoFeed abstracts location-aware news feed to a set of N location-based queries, such that: The N location-based queries are fired upon U logging on to the system Each location- based query is directed to one friend to retrieve the set of

k relevant messages

■ GeoFeed employs three approaches for each location-based query Spatial Pull approach Spatial Push approach Shared Push approach

■ GeoFeed employs a decision model that decides upon the best approach to evaluate each query such that: The system computational overhead is minimized Each user U will get the required news feed in TU time units

8

GeoFeed Preliminary :Problem Formulation

■ Given: User location User friend list User response time requirement User activity patterns, i.e., offline time and update

frequency

■ Find: Best approach among spatial pull, spatial push,

and shared push approaches, to evaluate q once u logs on to the system next time

■ Objective: Provide location-aware news feed for the user Guarantee a the response time that u will encounter to

get all the requested location-aware news feeds Minimize the computational overhead for all queries in

the system

9

The Spatial Pull Approach in GeoFeed

■ Spatial Pull approach Do nothing when the user offline Once the user logs on, compute al the queries for the

user

Advantage: No extra overhead during offline period Disadvantages: High user response time and not

efficient for the user with short offline time

Alice

SpatialFilter

Bob

Grid Index

1. location-based query

2. Alice’slocation

3. Get cell

4. Messages in the cell5. Relevant messages

Messages

10

The Spatial Push Approach in GeoFeed

■ Spatial Push approach Maintain a materialized view for the pre-computed

messages Once the user logs on, the answer is ready

Advantage: Users are very happy with very low response time

Disadvantages: System is overwhelmed with maintaining large number of views that may not be necessary

Materialized view

Bob

Grid Index

3. Range query

1. location-based query

New message

OtherMaterialized

views

Other Friends

4.Update2. Relevant messages

Alice

11

The Shard Push Approach in GeoFeed

■ Shared Push approach Share one view among queries for the nearby friends Once the user logs on, the answer is ready

Advantages: Users are still very happy with very low response time, and system overhead could be significantly lower

Disadvantages: Users need to be close enough, continuously check if views can be shared

Bob

Grid Index

3. Range query

1. location-based query

New message

Sharedmaterialized

viewNearby Friends

4.Update2. Relevant

messages

Alice

Filter

12

GeoFeed Cost Model

■ Spatial pull approach (based on per user-friend evaluation) Response time

Evaluating the location query

■ Spatial push approach (based on per user-friend evaluation) Response time/Query processing cost

Return messages from materialized view System overhead

Cost to update the materialized view with the user’s the offline time and the friend’s update frequency

■ Shared push approach (based on per cell evaluation) Response time

Return messages from the shared view with filtering System overhead

Cost to update the shared view with the user’s update frequency and friends’ minimum offline time

13

Challenges in Decision Model

■ Main Challenges: Guarantee a response time requirement for the user Do not overwhelm the system Consider the wide diversity of the user activity patterns in

social networking systems, e.g., offline times, update frequencies

■ To favor user response time More spatial push approaches will be adapted System is overkilled to maintain a large number of

materialized views and continuous queries

■ To favor system overhead More spatial pull approaches may be adapted Users suffer significant delays to get their news feeds

14

Which is the Best Approach for a Query

■ Consider the wide diversity in user activities in social networking systems e.g., offline times and update frequencies

A

B

C

D

E

F

A

B

C

D

E

F

A

B

C

D

E

F

System-wide decision Per-User decision Per-Query decision(GeoFeed)

A

B

C

D

E

F

OR

Users Friends

Users Friends

Users Friends

15

GeoFeed Decision Algorithm

■ Step 1. Response Time Guarantee For each user, this step uses our cost model to decide the

MAX number queries (N) to be evaluated by the spatial pull approach

■ Step 2. Spatial Pull & Push Selection For each user, this step selects N queries to be evaluated by

the spatial pull approach based on our cost model

■ Step 3. Shared Push Refinement For each user, this step attempts to share the execution of

his/her friends’ queries that are selected to be evaluated by the spatial push approach.

16

Experiments (1/4)

■ Data Sets Get final 646,697 tweets issued in State of Minnesota Use location information in tweets

Coordinate locations Semantic location, e.g., a city name (use Google

Geocoder)

■ Experimental Settings Based on a Postgresql database Based on the statistics from Facebook A set of evaluation experiments to get the parameters

to build the cost model and decision algorithm

17

Experiments (2/4)

■ Inside GeoFeed Decision model

■ Insights: With the increase of Tu, more spatial pull approaches are selected. When Tu=0 no spatial pull approaches are applied When Tu=∞, GeoFeed aims to only minimize the system overhead

through employing much of the spatial pull approach. Comparing two figures shows that with a smaller offline time, more

spatial push approaches are applied.

(a) Offline time = 1 hour (b) Offline time = 8 hours

18

Experiments (3/4)

■ Compare with traditional approaches

■ Insights: Pure spatial pull has bad response time Pure spatial push had bad system overhead

19

Experiments (4/4)

■ System overall overhead

Insight: GeoFeed with shared push refinement has the similar

response time but saves significant in system overhead

20

System Prototype

■ Sindbad: A Location-Aware Social Networking System (SIGMOD 2012 demo)

21

Conclusion

■ Location-Aware News Feeds Social relevance, i.e., a user’s friends/subscribed

news agents Spatial relevance, i.e., messages overlap user’s

location

■ GeoFeed is an efficient system equipped with a smart decision algorithm, which chooses the best approach among spatial pull, spatial push and shared push to evaluate location-aware news feed: Guarantee the user’s required response time Minimize the system overhead

22

Thanks