Download presentation
Presentation is loading. Please wait.
Published byMarsha Monica Hensley Modified over 8 years ago
1
The Biologically Inspired Distributed File System: An Emergent Thinker Instantiation Presented by Dr. Ying Lu
2
CAS (Complex Adaptive Systems) Systems that have – a large number of members with simple functions – limited communications among them CAS property – be able to adapt quickly to changing environmental conditions
3
Emergent Computation An emergent computation model – agents follow simple rules to affect their states and/or environment to produce a system wide result, an emergent computation – all computations (e.g. aggregation, resource allocation, classification, assignment, path selection, decision, etc.) can be obtained by emergent computations of simple activities
4
Regular vs. Emergent Computation A regular computation – a CPU computation like arithmetic and logic operations An emergent computation – make use of regular computation or other emergent computations to achieve its outcomes
5
Challenge How to identify the relationship between the emergent computation and the local agents’ properties or actions? Necessary and sufficient conditions to obtain certain emergent computations? How a specific property produces an emergent outcome?
6
Long-term Goal Control and manage the agent’s properties to obtain desired global outcomes
7
Determining Factors for Global Outcome Local properties & actions Propagation models: – different propagation approaches lead to different global, emergent self-* properties – CAS propagation model concerns the spread and its affects on the amplification of the agents’ actions to eventually give an emergent result
8
Emergent Thinker Paradigm The CAS emergent computation model is used as the building block for the paradigm Emergent function services (self-* property ) are provided to application by the CAS
9
Biologically Inspired P2P Distributed File System (BPD) An instantiation of the Emergent Thinker paradigm BPD – an alternative to deterministic techniques proposed in P2P and DFS – modeling natural behavior in its foundation services to solve distributed systems’ design challenge – environment: computing devices with ad hoc behavior (i.e. joining and leaving network); no central server or controller
10
Overview of BPD A P2P system with hundreds or thousands of computing devices (peers) Each peer execute basic, independent actions with minimum or no communication among them Emergent computations achieved by the actions provide computing services required by the DFS A user or application accesses File System services for its file management needs through calls to the DFS emergent computation engine that resides in each peer
11
Necessary Services for DFS Allocation Retrieval Replication Discovery For each BPD DFS service provided, there is an independent spatially decentralized domain of agent actions that execute on the same physical P2P system
12
CAS Algorithms For DFS allocation services For DFS discovery services Based on squirrel hoarding mechanism, both are achieved as emergent function services
13
Allocation Service in BPD Squirrels hoard acorn in dispersed caches, where they are allocating resources (land space) to storage demands (acorns) in such a way that resources are balanced Allocate data acorns evenly among nodes in BPD
14
Discovery Services in LDS Large Distributed System (LDS) property: – dynamic, self-organization, and ad-hoc connectivity and operation of its decentralized members – only constant: variability of the member’s connectivity to LDS (on/off/fail/disconnected) Adaptable, scalable search
15
Discovery Services in LDS (cont) Previous search algorithms: – structured search: too rigid – blind search (poor resource utilization) – informed search (such as PlanetP, requiring tables or indices maintenance, not suitable for LDS with extremely variable member population)
16
Emergent Search in Large Distributed Systems Emergent search – based on CAS, it is local to foster peer independence – emergent outcomes result from the member activities – compared to blind search, it minimizes messages by grouping several searches (acorn identifiers) within one message – CAS algorithm provides a system-wide, scalable search with reliability
17
Squirrel Emergent Search Each location has – its own independent squirrels – shared resources, e.g. shared files (data acorns)
18
Squirrel Emergent Search (cont) A new search arrives at a location, the location’s squirrel puts the acorn id in a bag together with other acorn ids already existing and hoards them in nearby locations
19
Squirrel Emergent Search (cont) If a bag with acorn ids is placed in a location, the acorn ids are searched within this location
22
Questions?
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.