fcmlaucs hku hk Research Issues in Adapting Computing

fcmlau@cs. hku. hk Research Issues in Adapting Computing to Small Devices Francis C. M. Lau Department of Computer Science The University of Hong Kong APPT 2005 • Hong Kong APPT 05 (27/10/05, HK)

Popular Computing • 70’s: one computer, many users • 80’s: one computer, one user • 90’s: many (connected) computers, many users – distributed computing • Now: many computers, one user • What next? Ubiquitous computing? Grid computing? Service-oriented computing? … APPT 05 (27/10/05, HK) – Do we need that many computers? – Do we need a “computer”? 2

Windows Panic • Continue or quit or kill the window or details? • How would my mother know!? • The best computing is without the computer APPT 05 (27/10/05, HK) 3

Imagine … • APPT 2015 (Disneyland, Hong Kong) … APPT 2015 APPT 05 (27/10/05, HK) 4

My Future Handheld • All in one: phone, PDA, wallet, game console, remote terminal, ad-hoc networking … • That means a lot of software – big and fat!? • But what I really want is lean and thin and mean – Lean – simple software which does only what I want – Thin – that’s all the space in my pocket, or around my neck! – Mean – affordable APPT 05 (27/10/05, HK) LTM 5

LTM Computing • The (thin-) client-server paradigm, works but in limited ways • Too thin: a browser-only device, a remote terminal … • Future devices are more powerful than just that – But how powerful is powerful? APPT 05 (27/10/05, HK) 6

O my god! Software complexity Subramanian, 1999 APPT 05 (27/10/05, HK) 7

Rich Computing • My hard disk and VM can be as big as the Internet • Not just contents, but software come and go on demand • Pay-per-use subscription model • Browser plug-in’s and Java applets – not lean enough, and some require just too much resources • Software in terms of “functionalities” APPT 05 (27/10/05, HK) The Network is the Computer ™ 8

The Network Behind Every Device • Future digicams – Snap and go • The network is the memory • The network is the photo album • (The network is NOT the power) – Context aware APPT 05 (27/10/05, HK) 9

Edge Computing • Person. . device. . middleware (proxies). . Internet • The abstract cloud moves with the client – personalized “cuddleware”, nomadic computing proxies united Internet metropolis client APPT 05 (27/10/05, HK) 10

GRID E DEVI CE EDG The continuum: WHAT SHOULD BE DONE APPT 05 (27/10/05, HK) WHERE? 11

A Continuum • Your device. . other devices (peers) in the vicinity. . server(s) nearby. . the grid – “Neighborhood” is still important • network bandwidth can’t be infinite; latency will never be 0; network will never be trouble-free • What should be done where? – The end-device: completely passive (HAS) to completely active (DIY) APPT 05 (27/10/05, HK) 12

“Hire a Server” • Completely passive (remote terminal) – A waste of bandwidth – Responsiveness and interactivity (esp. for games) • Does it make sense? – MP 3, camera, PDA, … • Something must be done locally – Adaptation APPT 05 (27/10/05, HK) 13

Limits You Can’t Beat • Power? • Screen size: foldable, rollable, wearable, projectable? Or do without the screen (audible? ) – Convenience and privacy; and power consumption – Resolution helps a little bit • Why I don’t use a PDA (I just ordered an 24” LCD; but I still look at my watch) – Ergonomics is very important – Scrolling is a nuisance (now) – User experience! APPT 05 (27/10/05, HK) 14

Wearable Computers server display APPT 05 (27/10/05, HK) 15

Adaptation • Data – = not code • Content – A special kind of data – Text, paged, non-text – Pixel level and/or higher • UI – Higher level, structural, interactive – Buttons, icons, menus, panels and panes … – content • Code APPT 05 (27/10/05, HK) 16

Componentized Software • Executable code as puzzle pieces • Fast start, like contents • Software download pipelining – first-comefirst-execute, the most needed functionalities first APPT 05 (27/10/05, HK) 17

The Need for Code On Demand • How many functions have you ever used? A simplified view of Microsoft Word APPT 05 (27/10/05, HK) 18

Remember UNIX? • Small is beautiful • Make each program do one thing well • Choose portability over efficiency • Store data in flat ASCII files • Avoid captive user interfaces • Make every program a filter • … APPT 05 (27/10/05, HK) 19

UNIX Philosophy Applied • Small software components implementing single functionalities • Bigger functions runtimecomposable from small components • Download on demand, and disposable • Standard formats and interfaces (eg. XML) • Shared GUI APPT 05 (27/10/05, HK) 20

Service-Oriented Computing eb ic le a c es S v r e S W l dev n a C mal ? n w Do s p p a ce i rs - fo APPT 05 (27/10/05, HK) 21

Adapting X to Y • Many issues – – Psychology Aesthetics Readability, convenience, privacy … Economics • Context awareness – Because small devices are likely mobile – The user moves around APPT 05 (27/10/05, HK) 22

Awareness and Adaptation • An intelligent supporting architecture • Aware of the come and go of mobile users • Choices of functions, depending on situation: the location, the device, the user, … • Fast reacting nearby proxy • Proxies united, within certain locality APPT 05 (27/10/05, HK) 23

Summary 1: Basic Problems (I think) Worth Pursuing • Edge computing → “seamless” – New protocols for the edge • The continuum → the network is the computer – Collaborative models and mechanisms, esp. at the edge • The global grid → invisible, “PC” disappearing • The device – Adaptation • On-demand code composition – The SOC approach? • Content – HTML • UI description languages – New paradigms for user interaction in small devices (input and output) APPT 05 (27/10/05, HK) 24

Some HKU Projects • The Sparkle Project – “code adaptation” • PDF document content adaptation system – “content adaptation” • HTML adaptation – “UI/content adaptation” APPT 05 (27/10/05, HK) 25

Application = “facet” composition Sparkle • Mobile code on demand – Proxy to device, device to device • Features – – – Code On Demand Context Based Adaptation Proxy Caching on Code & State Migration Ontology Sparkle View & Universal Browser APPT 05 (27/10/05, HK) 26

Universal Browser 5000 4500 Time (ms) 4000 3500 3000 2500 2000 1500 1000 APPT 05 (27/10/05, HK) 500 0 27 Add Noise Add Find Flip Gaussian Open Shadow Edges Horizontal Vertical Blur Image Save as Sharpen Jpeg

PDF Document Adaptation APPT 05 (27/10/05, HK) 28

APPT 05 (27/10/05, HK) 29

APPT 05 (27/10/05, HK) 30

APPT 05 (27/10/05, HK) 31

Web Document Adaptation APPT 05 (27/10/05, HK) 32

(a) The original content (b) Only the bottom level is “compressed” (c) The two lowest levels are “compressed” APPT 05 (27/10/05, HK) 33

Handling Segments APPT 05 (27/10/05, HK) 34

Summary 2: When I Grow Older • I want a simpler life (others can enjoy all the complexities) – One device, LTM – No computer please; terminal is fine; plug and play – Single data storage by hire, access anywhere – Single sign-on, single environment (adaptive) APPT 05 (27/10/05, HK) 35

APPT 05 (27/10/05, HK) 36