PERVASIVE COMPUTING SPCT- F2012 MOBILE COMPUTING

Transcription

1 PERVASIVE COMPUTING SPCT- F2012 MOBILE COMPUTING Jakob E. Bardram

2 Jakob E. Bardram 2

3 Mobile CompuGng intro Today s Agenda Chapter 16 Mobile and Ubiquitous Compu:ng, pp of Distributed Systems: Concepts and Design. G. Coulouris, J. Dollimore, T. Kindberg. 5 th edi:on, vola:lity & vola:le systems associa:on and interopera:on (sensing and context awareness) security adapta:on Jakob E. Bardram 3

4 Job Announcement We re looking for a student programmer in the PIT lab Working with Android programming PHP Deployment at the Rigshopitalet hence Danish speaking The MONARCA project mobile phone applica:on for bipolar disorder Jakob E. Bardram 4

5 Jakob E. Bardram 5

6 Mobile compugng in relagon to distr. systems and pervasive compugng (Satyanarayanan, 2001) Jakob E. Bardram 6

7 Different names... handheld compugng the device fits in the user's hand examples: mobile phones, PDAs, cameras mobile compugng users carry computers with them wireless connec:vity (Wifi, Bluetooth, GPRS, etc) example: laptops, smart phones, PDAs ubiquitous/pervasive compugng computers everywhere "one person, many computers "weave themselves into the fabric of life example: intelligent whiteboards Jakob E. Bardram 7

8 ... and more related areas wearable compugng user wears device on the body, integrated in clothes, watches, etc typically operated without user manipula:on example: user iden:fying badge, body sensors Jakob E. Bardram 8

9 VolaGle systems Jakob E. Bardram 9

10 volaglity & volagle systems all previously mengoned compugng forms have one thing in common: changes are common rather than excepgonal. note: volatility Examples of is volagle not a defining properges: parameter of mobile and ubiquitous computing systems. There are other systems out there that are also highly volatile, e.g. peer-to-peer file sharing applications connec:vity set of users hardware sobware excepgons devices and communica:on links fail communica:on characteris:cs change associa:ons are created and destroyed Jakob E. Bardram 10

11 Smart Spaces a smart space is any physical place with embedded services example: mee:ng room with intelligent whiteboard ubiquitous compugng happens within smart (physical) spaces mobile compugng happens both in- between and within "within" example 1: Magic Touch office environment (p271 in course book) "within" example 2: easyadl cogni:ve prosthesis Jakob E. Bardram 11

12 Example RecGcularSpaces JE Bardram, S Gueddana, S Houben, S Nielsen. ReticularSpaces: Activity- Based Computing Support for Physically Distributed and Collaborative Smart Spaces. In Proceeding of CHI 2012, ACM Press, Jakob E. Bardram 12

13 Q&A What did you just see in terms of...?... mobility... connec:vity... vola:lity Jakob E. Bardram 13

14 physical mobility Mobility... devices visit and leave the space (carried by human agents or by robots) logical mobility a sobware component might move in or out of a space, or of a device oben caused by physical mobility devices appears and disappears (and you really don t know why) "degree of volaglity = rate of change" Jakob E. Bardram 14

15 the typical mobile device (today) mobile devices are small with the following characterisgcs: limited energy bafery- powered devices balance performance, connec:vity, storage, bafery life resource constraints limited resources (processor speed, storage capacity, network bandwidth) adapt and design algorithms for limited resources use infrastructure to augment resources sensors and actuators needed to enable integra:on with the physical world sensors measures physical parameters actuators affect physical parameters Jakob E. Bardram 15

16 volagle connecgvity wireless connecgvity: WiFi, Bluetooth, GPRS, 3G, 4G, infrared,... disconnecgon connec:vity is not con:nuous, e.g. devices pass in and out of range variable bandwidth and latency bandwidth changes when users move about variable rougng communica:on without stable infrastructure ("ad- hoc") Jakob E. Bardram 16

17 spontaneous interoperagon in volagle systems, components change associagons oden associagon: logical relagonship between communicagng components. Examples: mobile device associated with local printer laptop client associated with an server (associa:on independent of connec:vity: client and server associated even if disconnected) in a smart space, associa:ons change because mobile components take the opportunity to interact with local components. associa:on spontaneity is oben driven by proximity in physical space. interoperagon: concrete interacgons during associagon. Examples: the mobile device sends data to be printed to the printer the laptop client receives the latest s from the server Jakob E. Bardram 17

18 lowered trust and privacy security in distributed systems relies on trusted hardware and sodware components trust: in mobile & ubiquitous systems, trust is problemagc because of the spontaneous interoperagon: components have never seen each other before components belong to different individuals/organiza:ons trusted third- party resource might not exist privacy: human agents might not like to be silently tracked by sensors in smart spaces other people might be able to directly or indirectly find out what you are doing, when, and where Jakob E. Bardram 18

19 associagon appearing devices need to associate with services and devices in smart spaces step 1: get an address on the network (the network bootstrapping problem) example solu:ons: DHCP (with infrastructure), ZeroConf (without infrastructure), UPnP (generalised plug- and- play) step 2: associate with available and relevant services in the smart space Jakob E. Bardram 19

20 Case Study Service Creates an object that implements the service Register the service object with lookup service(s) Keep it alive and running Lookup and Discovery Mul:cast request protocol Unicast discovery protocol Mul:cast announcement protocol Jakob E. Bardram 20

21 Step I Service RegistraGon Jakob E. Bardram 21

22 Step II Service Lookup Jakob E. Bardram 22

23 Step III Service Usage Jakob E. Bardram 23

24 JINI features... Centralized uses a lookup service (LUS) requires IP networking (DNS, DHCP, (zeroconf)) Object- oriented requires standardized (Java) interfaces needs to be shared / dynamically loaded i.e. needs compile :me considera:ons Java- based requires JVM, dynamical code loading, RMI Uses RPC using RMI implies non vola:le networking Jakob E. Bardram 24

25 associagon appearing devices need to associate with services and devices in smart spaces step 1: get an address on the network (the network bootstrapping problem) example solu:ons: DHCP (with infrastructure), ZeroConf (without infrastructure), UPnP (generalised plug- and- play) step 2: associate with available and relevant services in the smart space Jakob E. Bardram 25

26 associagon appearing devices need to associate with services and devices in smart spaces step 1: get an address on the network (the network bootstrapping problem) example solu:ons: DHCP (with infrastructure), ZeroConf (without infrastructure), UPnP (generalised plug- and- play) step 2: associate with available and relevant services in the smart space the scale problem: if available services components are many (e.g. > 1000) how can the device quickly choose with which component(s) to associate itself? Example solu:on: discovery services (allows the newcomer to ask someone who knows what is indeed present in the space) Jakob E. Bardram 26

27 associagon appearing devices need to associate with services and devices in smart spaces step 1: get an address on the network (the network bootstrapping problem) example solu:ons: DHCP (with infrastructure), ZeroConf (without infrastructure), UPnP (generalised plug- and- play) step 2: associate with available and relevant services in the smart space the scale problem: if available services components are many (e.g. > 1000) how can the device quickly choose with which component(s) to associate itself? Example solu:on: discovery services (allows the newcomer to ask someone who knows what is indeed present in the space) the scope problem: limit to devices within the smart space (and all of them), not, not outside it. Example solu:on: to adhere to the "boundary principle" which aligns logical topology to spa:al structures that make sense to humans example of a problema:c situa:on: find a hotel printer within room (not next door room) Jakob E. Bardram 27

28 associagon appearing devices need to associate with services and devices in smart spaces step 1: get an address on the network (the network bootstrapping problem) example solu:ons: DHCP (with infrastructure), ZeroConf (without infrastructure), UPnP (generalised plug- and- play) step 2: associate with available and relevant services in the smart space the scale problem: if available services components are many (e.g. > 1000) how can the device quickly choose with which component(s) to associate itself? Example solu:on: discovery services (allows the newcomer to ask someone who knows what is indeed if this cannot present be in done the space) automatically, the device/system the scope needs problem: to involve limit to devices the human within the smart space (and all of them), not, not agent, e.g outside rely on it. Example "physical solu:on: association" to adher to the "boundary principle" which aligns logical topology to spa:al structures that make sense to humans example of a problema:c situa:on: find a hotel printer within room (not next door room) Jakob E. Bardram 28

29 physical associagon examples: manual text input: let the human agent input in what room she/he is in. the room number can be linked directly to a specific printer on the network "physically constrained channel": e.g., digital id of room embedded in the music in the room (the digital id is thus only available within the room itself) (no human input needed) symbol capture: let the human use a mobile device to photograph a visual tag ("glyph") fifed somewhere in the space or on the printer itself to obtain the unique network iden:ty of the room or printer which is then looked up using a directory service direct associa:on: encode the network id of the device directly in the symbol (no directory service needed) read visually or through short- range radio (e.g. NFC, RFID) temporal/physical correla:on (e.g. press bufons/shake devices at the same :me) Jakob E. Bardram 29

30 interoperagon "a goal for ubiquitous and mobile compu:ng is that a component should have a reasonable chance of interopera:ng with a func:onally compa:ble component, even if the lafer is in a different type of smart space than from the one for which it was originally developed main challenge: interface incompagbility the object- oriented approach: use "translator" components which act as proxies for devices that need to communicate but do not understand each other benefit: each component interface can be very specialized and thus powerful drawback: for N device types you need to have up to N^2 translators the data- oriented approach: define one general interface that allows for a smaller set but well defined func:onali:es useful for a wide range of devices benefit: most devices can be made to speak the (limited) language drawback: more specialized devices cannot offer their special services Jakob E. Bardram 30

31 object- oriented example camera and a digital frame should be able to interoperate camera and digital frame disagree on interface solu:on: adapt the digital frame interface to one cameras understand many adaptors required (one for each camera/frame type or brand) for each missing adaptor we have a "lost opportunity" for interoperagon! this is the JINI approach a less complex standardized data- oriented interface is more feasible Jakob E. Bardram 31

32 data- oriented example camera and a digital frame should be able to interoperate camera discovers the frame service and its HTTP interface camera invokes HTTP POST/PUT on frame with a message body containing content type and image data camera and frame could first nego:ate viable content types this is the REST model combined with mdns / bonjour for discovery Jakob E. Bardram 32

33 security & privacy in mobile compugng Jakob E. Bardram 33

34 security & privacy in mobile compugng security is more difficult in mobile compugng cryptography public key cryptography requires too much CPU power one global symmetric key for all nodes makes systems vulnerable N 2 /2 symmetric keys may require too much memory energy spending ahacks can cause denial- of- service examples of spontaneous interoperagon problems: wireless services should only be available inside a smart space coworkers should be able to exchange documents on public WiFis wireless heart monitors should work easily with any pa:ent and not mix them up Jakob E. Bardram 34

35 spontaneous device associagon secure spontaneous associagon approaches: users enter PIN on devices (e.g., Bluetooth) keys sent via a physically constrained channel limited range (e.g., NFC) limited angle (e.g., infrared) limited :ming (e.g., wireless mouse pairing) various ways to use the physical channel send session key authen:cate a public key validate insecurely transmifed key Jakob E. Bardram 35

36 locagon based authengcagon authengcagng users of services in smart spaces can be based on the loca:on of clients rather than iden:ty e.g., using sound that does not escape the smart space that cannot travel far within a small :me interval loca:on based methods can compromise privacy: several different physical aspects (Bluetooth MAC, WLAN MAC, CCTV, credit card transac:ons,...) may now be associated by the system Jakob E. Bardram 36

37 devices in volagle systems are heterogeneous variability in processing power I/O capabili:es network bandwidth memory energy capacity adaptagon applicagons and services have to adapt to current resource availability Jakob E. Bardram 37

38 (technical) context- aware adaptagon services may adapt the content they provide to the context of the client examples adapt streamed video to screen size adapt encoding to device- present decoders and resources compress content to reduce bandwidth requirements content adaptors ("proxies") in the smart space poten:ally any device may serve content content adaptagon on the fly increasing set of permuta:ons of clients and services media- specific and lossy compression example: black & white images instead of color images Jakob E. Bardram 38

39 adapgng to changing resources network bandwidth and energy are crigcal resources, but can change at rungme OS support for adaptagon applica:on reserves resources (e.g., reserves a certain bandwidth for streamed video) user no:fied of changes in resources (e.g., when bandwidth is low, allows use to select fewer frames or black & white) applica:on no:fied of resource changes (e.g., when bandwidth is low, automa:cally switch to a stream with less requirements) smart space support for adaptagon equip the space with compute servers, "cyber foraging extend applica:ons to adapt to the presence of such servers example: a compute server may convert speech to text Jakob E. Bardram 39

40 mobile (phone) plakorms Android (Google) J2Me ios (Apple) Symbian (Symbian FoundaGon (Ericsson, Nokia, Motorola &Psion, now Nokia)) Meego (Nokia) HP webos (HP, formerly Palm) Windows Mobile (Microsod) BlackBerry 7.1 OS (RIM) Jakob E. Bardram 40