Service Discovery in Ubiquitous Computing Environments

1
Service Discovery in Ubiquitous Computing
Environments

• Simon (Chih-Chieh) Han
• Simonhan_at_cs.ucla.edu

2
What is service

• Applications that perform computation or action
  on behalf of users
• Examples
• Server with IP 128.97.93.22
• HP color printer
• Files that has UCLA in it

3
Where is X???

• Ubiquitous Computing envisions
• Billions of computers, devices, and files
• How do we locate a particular service or device
  in the network?
• Service discovery tries to answer the where
  question

4
Issues

• Availability
• Given services are in the network, what is the
  success rate of discovery
• Scalability
• Must support billion services in the wide-area
  network
• Load balancing
• Fault-tolerance
• Allow frequent service joins/leaves, user
  joins/leaves
• Security
• Allow only authenticated access to services
• Allow discovery of trusted services
• Expressiveness
• Support wide range of query for services

5
First GenerationWhere is www.ucla.edu?

• Characteristics
• Client/Server model
• Global name space
• Targeted at specific services
• Discover computes in the network
• Examples
• DNS (Domain Name Service)
• NIS (Network Information Service)
• SAP (Systems, Application, Products in Data
  Processing)
• Collection of applications for business
• LDAP (Lightweight Directory Access Protocol)
• Global
• Locating Objects in global name space

6
Second GenerationWhere is the printer?

• Characteristics
• Client/Server and Decentralized model
• Security
• Service expressiveness
• Fault-tolerance
• Discover devices in the network
• Examples
• SLP
• Jini (Sun Microsystem)
• Salutation (Salutation Consortium)
• INS (MIT)
• SDS (Berkeley)

7
Service Location Protocol (SLP)

• User Agents (UA) perform service discovery
• Service Agents (SA) advertise the location and
  characteristics of services.
• Directory Agents (DA) collect service addresses
  from SAs and respond to service requests from UAs
• DA discovery is done either from DHCP or sending
  request to multicast group address
  (239.255.255.253).
• UA must have the driver for SA!

8
Jini Network

• Developed by Sun Microsystem
• Very similar to SLP
• DA -gt Lookup service
• SA -gt Service provider
• UA -gt Client
• Specific to JVM
• Device driver can be downloaded from DA

9
Salutation

• Developed by the Salutation Consortium
• Similar idea as SLP
• Services register with Salutation Managers (SLMs)
  and clients query SLMs to find devices
• Independent of Transport layer
• As oppose to TCP/IP in SLP
• Independent of programming language
• As oppose to Java in Jini

10
Intentional Naming System (INS)

• Part of MIT Project Oxygen
• Focus on service description, query and routing
• XML service description using attribute-value
  tree
• Support queries of any-cast and multi-cast
• Integrate query resolution and routing

11
How INS works
12
(Service Discovery Service) SDS

• Part of Berkeley Ninja project
• Similar discovery protocol as Jini Network
• XML service description
• Hierarchies of SDS servers for Load balancing
• Focus on security
• Certificate Authority for public key
• Capability Manger for capability-based access
  control
• Wide-area network support
• Multiple hierarchies
• Lossy aggregated service description
• Support simple query with attributes

13
How SDS works
Dashed lines periodic multicast Solid lines
one time RMI connection
14
Third GenerationWhere is music clip?

• Characteristics
• Fully decentralized model
• Scale to wide-area network
• Structure the network so routing is quick
• Robust to frequent joins/leaves
• partition data so joins/leaves require minimal
  work
• Discover files in the network
• Napster
• Gnutella
• CAN (Berkeley)
• Pastry (Rice University Microsoft Research)
• Tapestry (Berkeley)
• Chord (MIT)

15
Napster Gnutella

• Napster
• Centralized server cluster
• load balance among servers using DNS rotation
• no security passwords in plain text, no
  authentication, no anonymity
• Gnutella
• Flooding based search protocol
• If you dont have the file you want, query 7 of
  your partners.
• If they dont have it, they contact 7 of their
  partners, for a maximum hop count of 10.
• Scaling is problem!!!

16
Content-Addressable Network (CAN)

• d-dimensional virtual coordinate space to store
  (key, value)
• Entire coordinate space is dynamically
  partitioned among all the nodes, every node
  owns its individual, distinct zone.
• Using consistent hash to store/retrieve key
• Nodes keep virtual neighbors for routing/querying
  purpose.

17
Content-Addressable Network (CAN)

• When node joins
• Bootstrap
• Randomly pick point p
• Find point p and join routing

18
Queries in CAN

• Store/retrieve files
• (filename (K), data (V)) K is deterministically
  mapped onto a point P in the coordinate space
  using Uniform hash function.
• Simple greedy forwarding to neighbor with
  coordinates closest to destination coordinate
• Every CAN node maintains a coordinate routing
  table with pointers to 2d neighbors.
• d-dimensional coordinate space partitioned into n
  equal zones path lengths are O(dn1/d) hops.

19
Pastry

• Decentralized, scalable, and self-organizing
• Adapt to the arrival, departure and failure of
  nodes.
• Seeks to minimize the distance messages travel,
  according to a scalar proximity metric like the
  number of IP routing hops.
• In Pastry network,
• Each node has a unique 128 bits id, nodeId
• secure hash on node public key or IP address
• Presented with a message and a key, Pastry node
  efficiently routes the message to the node with a
  nodeId that is numerically closest to the key.

20
State of Pastry Node

• NodeIds are in base 4
• Leaf set
• L/2 numerically closest larger nodeID and L/2
  nodes with closest smaller nodeId
• Routing Table
• Chosen from scalar proximity metric, such as the
  number of IP routing hops or geographic distance.
• Neighborhood Set
• Stores closest nodes according to proximity
  metric

21
Routing in Pastry

• Forwards message to a node whose nodeId shares
  with the key a prefix that is at least one digit
  (or b bits) longer than the prefix that the key
  shares with the current nodeId
• If no such node is found in the routing table,
  the message is forwarded to a node whose nodeId
  shares a prefix with the key as long as the
  current node, but is numerically closer to the
  key than the current nodeId
• Routing table size O(log N)
• Number of hops O(log N)

22
Tapestry

• Nodes and objects have unique 160-bit IDs
• Nodes route messages using destination ID, to a
  node whose ID shares longer suffix
• Fault-tolerant using soft-state lease
• Periodic update service information

23
Routing in Tapestry

• Neighbor Map at node 0642

24
Routing in Tapestry

• the path taken by a message originating from node
  0325 destined for node 4598

25
Chord

• Provides peer-to-peer hash lookup service
• Lookup(key) ? IP address
• Chord does not store the data
• Efficient O(Log N) messages per lookup
• N is the total number of servers
• Scalable O(Log N) state per node
• Robust survives massive changes in membership
• Chord assigns keys to node with consistent
  hashing technique
• All nodes roughly receive the same number of keys
• When the Nth node joins ( or leaves ) the network
  only O(1/N) fraction of keys are moved
• Uses a flat namespaces 0 2m -1 (Key spaces
  used to identify both object and nodes)
• Node ID in the key space is obtained through
  hashing its IP address
• A node is responsible for all keys equal to or
  before its identifier until its predecessor
  nodes identifier in the key space.

26
Chord Identifier Circle
27
Simple Key Lookup

• Each node contacts its successor node on the
  identifier circle

28
Scalable Key Lookup

• Lookup with finger table (routing table)

29
Trend of Service Discovery

• Type of service
• Hosts -gt Devices -gt Files
• Architecture
• Client/Server -gt Decentralized
• Network size
• Local network -gt Global network
• Service Description
• Global name space -gt arbitrary attribute-value
  pairs
• Service mobility
• Static -gt highly mobile

30
Beyond Service Discovery

• What is the business model once we have service
  discovery
• How can we trust the service provider?
• How do we determine the quality of service?
• What are the services besides files and printers?
• Network simulator service?
• Bus schedule service?
• Stock quote?

31
References

• Christian Bettstetter, Christoph Renner, A
  Comparison Of Service Discovery Protocols And
  Implementation Of The Service Location Protocol.
• Steven E. Czerwinski, Ben Y. Zhao, Todd D. Hodes,
  Anthony D. Joseph, Randy H. Katz, An Architecture
  for a Secure Service Discovery Service.
• William Adjie-Winoto, Elliot Schwartz, Hari
  Ralakrishman, Jeremy Lilley , The design and
  implementation of an intentional naming system
• Peter Druschel and Anthony Rowstron. Pastry
  Scalable, distributed object location and routing
  for large-scale peer-to-peer systems. Submission
  to ACM SIGCOMM 2001.
• Sylvia Ratnasamy, Paul Francis, Mark Handley,
  Richard Karp, and Scott Schenker. A scalable
  content addressable network. Submission to ACM
  SIGCOMM, 2001.
• Ion Stoica, Robert Morris, David Karger, M. Frans
  Kaashoek, and Hari Balakrishnan. Chord A
  scalable peer-to-peer lookup service for internet
  applications. Submission to ACM SIGCOMM, 2001.