13
Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Embed Size (px)

Citation preview

Page 1: Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Distributed SystemsBrief Overview

CNT 5517-5564Mobile & Pervasive Computing

Dr. Sumi HelalUniversity of Florida

Page 2: Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Distributed System Values

• Reliability• Availability• Dependability• Security• Performance: response time, throughput, etc• Scalability: vertical (concurrency) and

horizontal (distribution)

Fault-tolerance

Page 3: Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Distributed System Values, cont’d

• Consistency• Up-to-date-ness• Adaptability• Reconfigure-ability• Manageability• Extensibility• more

Page 4: Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Reliability

• Eternal:• Mean Time to Failure (MTTF)• Extended Reliability = Availability– (1) Employing Redundancy (replication)– (2) Employing specific Recovery Procedures– In either (1) or (2): • First detect failure• Second Mask it (redundancy), or Recover from it

(recovery)

Page 5: Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Availability• Past:

• Today: – As past, plus:– Available through a variety of networks – Available through a variety of devices– Available through a variety of locations (e.g., while mobile? )

4 nine’s = 0.99994 nine’s = 0.9999

Page 6: Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Replication in Distributed Systems

• Consider data, how can we use replication of data to increase its availability and use?

Page 7: Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Replication Management Protocols

Page 8: Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Distributed Systems Intricacies

• Synchronization: multiple clocks (difficult to agree on exact time)

• Concurrency: multiple simultaneous accesses potentially conflicting.

• Failures: high probability of failures (too many components). Complex failure modes (single, multiple simultaneous, network partition, …)

• Consensus: difficult to reach consensus (odds includes failures, lack of synchronization, …)

• more …

Page 9: Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Distributed Computing Models

• Client/Server Approach • Multi-tier Approach• Peer-to-Peer Approach • Agent based systems• Mobile Code (Applets, mobile agents, ..)• Service-oriented computing– Service registration/discovery– Service composition

• Cloud Computing

Page 10: Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

The Emerging Computing Model: Cloud, Edge & Beneath

• Cheaper to provide services to a thin client than to maintain a fat client in a changing world (service Science)

• Besides the Cloud, all other computing infrastructures become simply Edge Devices to the Cloud

• Do Sensors count in this new world of Clouds? Are Sensors Edge Devices?

Page 11: Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Cloud, Edge & BeneathSensors – to - Cloud

Page 12: Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Cloud, Edge & BeneathGateway – to Cloud

Page 13: Distributed Systems Brief Overview CNT 5517-5564 Mobile & Pervasive Computing Dr. Sumi Helal University of Florida

Cloud, Edge & BeneathSensors – to – Edge; Edge – to- Cloud

In N

etw

ork

Com

puti

ng

The Case for the Edge

Clo

ud C

ompu

ting