Download Create the foundation for data storage and retrieval, equivalent to

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Digital Object Architecture:
Building Information Management
Infrastructure for Networks
20 September 2010
Larry Lannom
Corporation for National Research Initiatives
http://www.cnri.reston.va.us/
http://www.handle.net/
Three Initial Networks
• About 30 – 35 years ago, DARPA funded the creation
of three seminal packet networks – ARPANET, Packet
Radio, Packet Satellite
• The Internet came about from a desire to link the
three of them
• Ethernet occurred in parallel, led by Xerox Parc
researchers, and other network types followed
• The resulting architecture was independent of the
number and type of networks or who ran them.
Corporation for National Research Initiatives
Key Decisions
• The Internet would be a global information system.
• An open-architecture would be used to combine
different networks based on open and well-known
interfaces, protocols & objects.
• A new communications-oriented host protocol
(TCP/IP) would be created to replace the original
ARPANET host protocol (NCP).
• The concept of global addressing and IP addresses
would be introduced to identify individual machines
anywhere on the global Internet.
Corporation for National Research Initiatives
Comments on the Key Decisions
•
•
•
•
•
The architecture is robust in the presence of many
different network types and many outages.
Gateways provided IP routing and Network
"Impedance Matching".
TCP accommodated end-end protocol:
— different packet sizes, duplicates, error
detection, losses due to tunnels, mountains,
jamming, etc.
Separate network administrations were permitted,
which allowed the Net to grow.
DNS not technically critical, but helped users.
Corporation for National Research Initiatives
Understanding the Big Picture
• Many things were done well from the outset; with
20/20 hindsight, some could have been done better.
• The context was critical:
–
–
–
–
Mostly mainframes, few time-sharing systems
No PCs, workstations, LANs
One dominant carrier in the US
Government facility initially
• What is important at the time may be only apparent
with hindsight; but also what seems important at
the time may not turn out to be so important later
on.
Corporation for National Research Initiatives
Infrastructure Development
• What is so hard about it?
– Making it scalable over platforms, size and time
– Achieving Critical Mass
• Getting Buy in:
– Pleasing many essential participants
– Displacing prior capabilities
– Structuring matters to deal with concerns about empire
building
• It’s a lot easier to create brand new capabilities than
to affect existing means of operation.
Corporation for National Research Initiatives
Infrastructure Creation is a Subtractive Process
• Infrastructure reduces a common, shared capability
to its basic and essential attributes.
• These attributes are not always recognized or
understood up front.
• Upon further scrutiny, capabilities are usually deleted
from a well-conceived architecture over time.
• Consensus develops when no more can be removed
without disabling the infrastructure.
Corporation for National Research Initiatives
What is the Information Management Problem?
• Managing information in the Net over very long
periods of time – e.g., centuries or more.
• Dealing with very large amounts of information in
the Net over time.
• When information, its location(s) and even the
underlying systems may change dramatically over
time.
• Respecting and protecting rights, interests and value.
Corporation for National Research Initiatives
A Meta-level Architecture
• Allows for arbitrary types of information systems.
• Allows for dynamic formatting and data typing.
• Can accommodate interoperability between multiple
different information systems.
• Allows metadata schema to be identified and typed.
Corporation for National Research Initiatives
Digital Object Architecture: Motivation
• To reformulate the Internet architecture around the notion
of uniquely identifiable data structures.
• Enabling existing and new types of information to be reliably
managed and accessed in the Internet environment over
long periods of time.
• Providing mechanisms to stimulate innovation, the creation
of dynamic new forms of expression, and to manifest older
forms.
• While supporting intellectual property protection, finegrained access control, and enable well-formed business
practices to emerge.
Corporation for National Research Initiatives
Digital Object Architecture Technical Components
• Digital Objects (DOs)
– Structured data, independent of the platform on which it was created
– Consisting of "elements" of the form <type,value>
– One of which is its unique, persistent identifier
• Resolution of Unique Identifiers
– Maps an identifier into "state information" about the DO
– Handle System is a general purpose resolution system
• Repositories from which DOs may be accessed
– And into which they may be deposited
• Metadata Registries
– Repositories that contain general information about DOs
– Support multiple metadata schemes
– Can map queries into unique DO specifications (via handles)
Corporation for National Research Initiatives
What is a Digital Object?
• Defined data structure, machine independent.
• Consisting of a set of elements:
– Each of the form <type,value>
– One of which is the unique identifier
• Identifiers are known as "Handles":
– Format is "prefix/suffix"
– Prefix is unique to a naming authority
– Suffix can be any string of bits assigned by that authority
• Data structure can be parsed; types can be resolved within the
architecture.
• Associated properties record, and transaction record, contain
metadata and usage information.
Corporation for National Research Initiatives
Interoperability & Federated Repositories
• Create a cohesive interoperable collection of
repository-based systems.
– Initially, perhaps, around a core set of projects,
content, applications and/or organizations
• Demonstrate interoperability between different
repository collections.
• Develop procedures to insure continued accessibility
to key archival information.
Corporation for National Research Initiatives
Repository Notion
Logical External Interface
DOP
Digital
Object Protocol
Any Hardware & Software
Configuration
Corporation for National Research Initiatives
Repositories & Digital Objects
Objects may be
Replicated in
Multiple Repositories
Repository
• Each Digital Object has its own unique & persistent ID.
• Content Providers assign IDs.
• Could be upwards of trillions of DOs per Repository.
Corporation for National Research Initiatives
The Handle System
• Distributed identifier service on the Internet
• First general purpose resolution system
• Can be used to locate repositories that contain digital objects
given their handles – and more!
• Other indirect references
– Public Keys, Authentication information for DOs
• Accommodates interoperability between many different
information systems
Corporation for National Research Initiatives
Attributes of the Handle System
• The basic Architecture of the Handle System is flat,
scaleable, and extensible.
• Logically central, but physically decentralized.
• Supports Local Handle Services, if desired.
• Handle resolutions return entire "handle records" or
portions thereof.
• Handle Records are also:
– digital objects
– signed by the servers
– doubly certificated by the system.
Corporation for National Research Initiatives
Resolution Mechanism
Handle
Handle Record
Multiple Sites
Multiple Servers
Handle System
<www.handle.net>
•
•
•
•
System is non-nodal
Scaleable & Distributed
Supports global (and local) resolution
Has backup for reliability, mirroring for efficiency
Corporation for National Research Initiatives
Conclusions
• Managing Digital Objects for long-term access is a key
challenge.
• Initial technology components are available; industry is
expected to generate more over time.
• Third-party value-added providers in the private sector
will ultimately shape the long-term evolution.
• Interoperability and reliable information access is a
critical objective.
• A diversity of applications (with user-friendly interfaces)
need to be developed & deployed.
Corporation for National Research Initiatives
Phone Guy Perspective
Corporation for National Research Initiatives
Purpose of Digital Object
Create the foundation for data storage and retrieval, equivalent to
what packet data did for communication.
• Today's architectures and paradigms, including leading edge
technology, operate on the circuit switched telephone
equivalent of data storage.
– A "dumb" system for payload data storage ("the circuits").
– A separate system for management, control, and metadata
("the signaling network").
• As a consequence, these systems are limited in robustness,
security, interoperability, extensibility, cost effectiveness,
vendor independence, and functionality.
Urs Muller, Net-Scale
Today's Paradigms
Data storage
Data Data
Data
Data
Data
Data
Authentication
Data management
•
•
•
•
•
Access control
Key management
Provenance infrastructure
Version control
Metadata
Request
Data
User
Examples:
• Documentum (EMC)
• SharePoint, MOSS 2007 (Microsoft)
• FileNet (IBM)
• 10g, Stellent (Oracle)
• LiveLink (OpenText)
• Alfresco (open source)
Urs Muller, Net-Scale
What Happens When Data Is Moved
•
•
•
•
•
Data management
Data storage
Data
Data
Data
Loss of access control
Loss of key management
Loss of provenance infrastructure
Loss of version control
Loss of metadata
Data
Data
Data
Urs Muller, Net-Scale
Data
Limitations of Today's Paradigms
•
•
•
Use of separate and different systems for storage of the (payload) data and the
data management.
– Creates a centralized system.
– Poor interoperability.
– Heavy vendor and product dependence.
The data management system is a fragile huge single point of failure which
requires heavy protection to make a solution usable.
– This is similar to the signaling network and out of band data in a circuit
switched traditional telephone network.
Poorly suited to reach these key requirements for the DoD:
– High degree of global data distribution and replication (a super robust
network, data is available where needed).
– Vendor independence.
– Interoperability among vendors and multiple technology generations (like
the Internet).
– Access control "travels" with the data and does not need to be replicated
each time the data is copied onto a different system (e.g., a laptop).
Urs Muller, Net-Scale
Digital Object Architecture
Data
Data
Data
Data
•
•
•
•
•
Access control
Key management
Provenance infrastructure
Version control
Metadata
Data
Digital Object Repository
Urs Muller, Net-Scale
A Digital Object Is Moved
Data
Data
Data
Data
Data management remains intact:
• Access control
• Key management
• Provenance infrastructure
• Version control
• Metadata
Data
Digital Object Repository
Urs Muller, Net-Scale
Data
A Solid Foundation
The Digital Object Architecture provides a solid foundation for the creation of:
• A highly distributed, robust, and scalable data storage and retrieval
infrastructure.
– Digital Objects are self-contained and don't depend on a separate centralized
data management subsystem. This dramatically improves scalability.
• A highly secure data storage and retrieval infrastructure.
– By eliminating a centralized security paradigm which is a single point of failure
and greatly vulnerable to attacks.
– Security is distributed. A successful attack reveals very little reward (each
digital object has to be attacked separately).
• A highly "future proof", extensible, interoperable, and vendor independent
data storage and retrieval infrastructure.
– By greatly reducing the complexity for exchanging data without breaking
access control, provenance, version control, etc.
The Digital Object Architecture provides a far superior foundation for realizing
these essential properties compared to today's paradigms.
Urs Muller, Net-Scale
Comparison to Data Communication
Circuit Switched (old phone)
(~ traditional architectures)
• Data has no "intelligence" and is managed by
a large central system (signaling network).


Packet Based (Internet)
(~ Digital Object Architecture)
• Data management information is embedded
with the data itself (packet header).
– The packet itself knows what it is,
where it is coming from and where it is
going to.
– The network can be simpler, far more
flexible and robust.
Today, few people dispute that packet routing is superior to circuit switching for data
communication.
– A few decades ago the differences were not so clear. After all, data can easily be
exchanged over a circuit-switched network.
Compared with today's paradigms, the Digital Object Architecture will lead to far more
flexibility, diversity, technology independence, and overall usage for data storage and
retrieval.
Urs Muller, Net-Scale
Example From the Real World
• Circuit switched past: When a 5ESS switch was down, all calls
to the affected area were out, leaving a whole region without
communication.
• Current Internet: On December 19, 2008 three undersea cables
were cut between the Middle East and Europe. Data traffic was
severely impacted but communication remained intact.
We expect the Digital Object Architecture to create a paradigm
shift for data storage and retrieval similar to the impact the
Internet had on data communication.
Urs Muller, Net-Scale
Digital Object Architecture
Where Are We?
• Handle System
• Up and running since the early 90s
• Core architecture stable from the late 90s
• www.handle.net
• Digital Object Repository
• In daily use in multiple projects
• Available open-source since the start of 2010
• www.dorepository.org
• Introductory article in Jan/Feb D-Lib Magazine
• Digital Object Registry
• In daily use in multiple projects
• Available open-source since May, 2010
• www.doregistry.org
Corporation for National Research Initiatives
Information Management on Networks
Resolution
Client
<?xml version="1.0"?>
<note>
<?xml version="1.0"?>
<to>John</to>
<note>
<from>Jane</from>
<to>John</to>
<heading>Reminder
<from>Jane</from>
<body>Don't
forget me!
<heading>Reminder
</note>
<body>Don't forget me!
</note>
Repositories / Collections
<?xml version="1.0"?>
<description>
<?xml version="1.0"?>
…….
<description>
<?xml version="1.0"?>
</description>
…….
<description>
</description>
…….
</description
>
Resource Discovery
Search Engines, Metadata Databases, Catalogues, Guides, etc.
Identifier Resolution System
Information Management on Networks
Administrative
Client
<?xml version="1.0"?>
<note>
<?xml version="1.0"?>
<to>John</to>
<note>
<from>Jane</from>
<to>John</to>
<heading>Reminder
<from>Jane</from>
<body>Don't
forget me!
<heading>Reminder
</note>
<body>Don't forget me!
</note>
Repositories / Collections
<?xml version="1.0"?>
<description>
<?xml version="1.0"?>
…….
<description>
<?xml version="1.0"?>
</description>
…….
<description>
</description>
…….
</description
>
Resource Discovery
Identifier Resolution System
Search Engines, Metadata Databases, Catalogues, Guides, etc.
Corporation for National Research Initiatives
Information Management on Networks
Administrative
Client
<?xml version="1.0"?>
<note>
<?xml version="1.0"?>
<to>John</to>
<note>
<from>Jane</from>
<to>John</to>
<heading>Reminder
<from>Jane</from>
<body>Don't
forget me!
<heading>Reminder
</note>
<body>Don't forget me!
</note>
Repositories / Collections
<?xml version="1.0"?>
<description>
<?xml version="1.0"?>
…….
<description>
<?xml version="1.0"?>
</description>
…….
<description>
</description>
…….
</description
>
Resource Discovery
Search Engines, Metadata Databases, Catalogues, Guides, etc.
Identifier Resolution System
Corporation for National Research Initiatives
Federation
• Federation in information systems makes
sense when
– a set of varying features exists across the
federates, which is the reason for multiplicity
• Includes organizational boundaries, locations,
content types, etc.
– a set of common features exists across federates,
which is usuallly the reason to perform
federation
• Shared topics, common audience, etc.
Corporation for National Research Initiatives
Challenges - Conceptual
• Identifying the type of aggregation:
– Aggregate objects ahead of time, before query?
– Merge search responses from federates by issuing a distributed
query?
– Or, anything in between?
• Identifying the level of semantic interoperability
– Enforce complete semantic interoperability across all the data
stored in the federates?
– Use only the least common denominator (from a data semantics
point of view) among the federates?
• Federate topology
– Are all federates directly connected to each other? (fully-connected
mode)
– Is each federate connected to only its neighbor? (peer-peer mode)
• These criteria can be visualized as a Federation Spectrum
Corporation for National Research Initiatives
Federation Spectrum
Level of data Interoperability
Complete Semantic Interoperability
No Semantic Interoperability
(Ad Hoc Mix)
Challenges - Technical
• Depending on the criteria chosen for federation,
various technical requirements arise. These may
include:
– Designing a storage model to aggregate objects into a
common store that identifies the relationship between
multiple metadata instances describing a single object
– Designing cross-walking algorithms to translate and
map heterogeneous data into a common model
– Designing a query model to gather and rank search
results from multiple federates
– Ensuring scalability, reliability, and security without
compromising performance
Corporation for National Research Initiatives
Existing technologies
• Digital Object Registry (basis for ADL-R)
– Provides a data model to encapsulate related
metadata instances together
– Enables aggregation of objects from fullyconnected mode to peer-peer mode
– Uses the Handle System to uniquely identify
objects and metadata instances across all
federates
Corporation for National Research Initiatives