An Overview of Internet Oriented DARPA Projects

1
An Overview of Internet Oriented DARPA Projects

• Web Engineering
• Microcontrollers
• VLSI Design

AuthorsVladimir Jovicic, Prof. Dr. Veljko
Milutinovic
Kick off
2
What is DARPA?

• An American DoD(Department of Defense) sponsored
  org supposed to make high-risk research in
  order to go beyond normal ways of development.
• When the theoretical work is done, to pursue the
  ideas towards making prototype systems.
• Consists of offices faced towards de-
• velopment (advanced tech, info sys-
• tem, defence science) and support
• offices (contracts management, etc).

3
When Why What So Far?

• Established in 1958,as the response to the
  Soviet launch of Sputnik.
• The day of founding - February 7th, and the name
  ARPA (Advanced Research Projects Agency).
• From March,1972 the letter D was added to the
  name, and that was for Defence.
• Achievements so far recognized world-wide
  Internet, hardware description languages, 32-bit
  200MHz RISC
• (during 1980-ties) , TCP/IP, and so on.

4
-DARPA, you say? Never mind them, its my
concern. Better watch out for Belgrade Faculty of
Electrical Engineering. They could be TROUBLE!
5
(No Transcript)
6
The March Rabbit proudly presents
Web Engineering
7
WWW - Evolving to Information Infrastructure

• The object is the evolution of protocols,
  services and capabilities of the WWW, for-
  defense, - government, - commercial use.
• The content n programming paradigmminimize
  the difference between authors and programmers.
• The work is based on these items of approach
• Designing new sub-languages that can imbed new
  capabilities and interaction models on the Web
• Automating content consumption and production
• Improving and designing semantics for new
  contents

8
What is the CURL?

• The prototype name of the new environment.
• It is to provide a wide set of developing
  tools, from a HTML-like to powerful
  object-oriented languages.
• Automatically acquises 3D geometric models of
  real-world , using mobile cameras and
  reconstruction software.
• Enables the development of new algorithms
  that replicate web content, which for the
  problem of leveling load.

9
think I didnt know that ?
Mmmm Curley.
Achieved so far - Curl on the popular platforms
(x86, Windows or Unix) - automatically
reconstructing the 3D model - developing the
protocols that eliminate the Web hot spots
10
Distributed clients for the WWW

• Read update access, regardless to the bandwidth
  situation (bad connection, or none at all)
• The phases of the project (1 and 2)
• The Caubweb a new approach in the data flowon
  the WWW (phase 1).
• The weblets
• Sharing the idea, and developing Caubwebfor the
  popular computer platforms.

11

• Taming the Internet beast...
• Its always up to the tactics.

12
Introduction

• Provides read and update access to Web-based
  information, under conditions of variable
  bandwidth, and intermittent network
  connectivity.

• The real goal is to accomplish connection
  portable computing with Web environment

- support information selection
-retrieval of information
-publication
-sharing
13
The Phases of Developing Project
to share with your beloved ones?
Isnt it the nicest thing...

• Exploring the work in situations of alternating
  periods of high bw connectivity and no
  connectivity ( phase 1)

• Combining multiple-users intocoherent team
  information space,that can avail as much
  infoas possible ( phase 2 )

14
The Caubweb

• Elaborating phase 1, the working team developed
  a platform-portable, browser-neutral HTTP
  stream transducer (a kind of a Proxy, running
  on an user machine)

• This Proxy (The Caubweb) can denote subsets of
  information web (the weblets ), using a sort
  of weblet developing language.

• I could usea better proxy...

15
Thats better!

• Generally spoken, it pre-fetches weblets during
  the high bw connectivity, so they can be
  accessed even when connection is closed.

• Later on, when the connection is re-assembled,
  information can be updated from the cached
  pages, previously modified on Proxy.

16
continued from the previous slide

• In the phase two, as mentioned, teams are not
  only to cache info of their own interests,
  but to also populate their Proxies, in order
  to permit location of other relevant resources.

• Doing so allows - to have ongoing access to
  relevant information - to modify it, even while
  being disconnected.

17

• The milestones to reach for the first
  version of project, was placing multiple user
  caches into a coherent "team information
  space (OASIS).

• The development started in 1997, and since
  then, the main concepts have been achieved ,
  considering a prototype of Caubweb, or
  system implementations for Windows, Unix and Mac.

18
The Hard Real-Time Execution of Java( PERC )

• The quest here is to use high-level Java
  attributes, for the domain of embedded
  real-time software.

• Javas most important benefits

-pure object orientation
-automatic garbage collection
-reusable software
-portability of source and binary (byte code)
representations
19

• When working in Java, programmers produce
  twice as much code as they would develop
  using C or C

• Even greater benefits of Java is reducing of
  software maintenance cost, by providing
  strong type consistency check.

• This concept, unfortunately, lacks
  implementation of a real-time determinism,
  combined with the previously mentioned
  conveniences.

20
The Approach Objectives

• There are two major objectives

- providing appropriate virtual machine
development tools
- developing and promoting disciplines for
portable, real time software components.

• For the first objective, some real-time
  semantics that werent sufficiently constrained
  in official Java Language specification, are
  defined now

- thread priorities, priority inheritance
- real time constants ( timeouts, etc.) are made
visible
- real-time garbage collection
21

• Among all things, it has been added a pre-run
  native compilation and linking tools for
  out-of -ROM execution.

• For the second objective, classes for
  real-time support have been designed and added,
  to make sure that all the real-time components
  identify its own resource needs.

• Identification is to be done during the
  processes of configuration and negotiation,
  before the activity even begins to execute.

22
and milestones achieved so far

• The Java source-code compiler (Percolator),with
  support for real-time syntaxes.
• Contest switches reduced form 3ms on 300
  microseconds,and time-driven task jitter
  improved for 10 times.
• Released commercial version of PERC 2.0, that
  besides full compatibilitywith Java 1.1 .lang,
  .util, .io and .net libraries makes possible to
  write portable device drivers.

23
The Network Centrics

• The focus - network behavior during the data
  updates that are transmitted over
  low-bandwidth connections.

• The research should design a system able to
  track and pre-fetch updated Internet data
  -for remote storage -for browsing at
  locations with limited or sporadic network
  connectivity .

24
and network centrics so far

• The research itself took following directions

- first, a controlled network test-bed
environment was created, for data-caching
analysis and protocol experiments
25

• Then, the experiments are to be made in
  order to determine ability to transfer fresh
  Internet content while automatically
  throttling back that flow, to accommodate for
  operational data transfer.

• The project team realized that, so far, all
  their needs can be fulfilled with commercial
  caching servers.

This is what I callnetworking
26
The Adaptive Web Caching

• The immediate objective of this project is to
  develop a Web caching system, that is -
  self-configuring - highly adaptive - allowing
  both WWW and other data dissemination
  applications to scale to the global information
  infrastructure, and even beyond.

27

• Later on, the research would be focused towards
  understanding the behavior of large-scale
  systems made of heterogeneous and autonomous
  components.

28
approaching part one

• First, to build a self-organizing data transport
  substrate, and to induce self-organizing
  information flow on top of it

- It is needed to build the mechanisms for
organizing autonomous Web caches servers
into multiple, overlapping, multicast groups.
- Each group is a subnet, and each overlapping
point is a router.
- Collectivity of groups makes an application
level topology, made of caches, connected as
clients severs.
29
...

• So, each Web page builds its own dissemination
  a tree, while the chache groups reorganize
  themselves, due to changes in operational
  environment.

30
...approaching part two

• Secondly, to design algorithms and protocols
  to forward Web references through this maze of
  cache groups.

• When a Web page request has been made, it is
  usually being

- served by the local subnet
- forwarded to cache groups in the direction
towards the origin or a replicated server
- forwarded towards another cache, judged as
most likely to have the referenced object.
31

• The algorithms for forwarding web requests are
  based on a "URL routing table" containing URL
  prefixes, IPs ...

• The temperature measurement tool graphics
  indicate the clients interest in specific Web
  pages.

32
achievements

• The basic design of the Adaptive Web Caching
  system (AWC) it consists of

- mechanisms for organizing into multiple
groups (as mentioned before).

- heuristic research algorithms for directing
Web references through cache-group maze.

• The main idea is temperature measurement the
  way to capture clients interests for specific
  pages.

• When a page with high popularity is detected,
  server sends its URL to the cache(s),
  protecting origin server from actual demands

33
The Safe Execution Environments

• Develops technology to provide safe execution of
  applets, active controls, agents and
  downloaded programs.

• It provides the mediators that monitor execution
  of programs and control their access to
  information.

• Safety information policies are to be
  user-defined, and they are to specify what
  resources can be accessed or modified.

• For the accessible resources, copies are to be
  provided, so that real data is hidden from the
  executing program. With this provided, user can
  undo registry entries, or file changes.

34
...done so far

• Developed a safe execution environment for Web
  browsing. (it ensures of Netscape Explorer
  running with Java ActiveX)

• Developed a safe execution environment for
  running office products. (protects from
  possible malicious macros using the same tool as
  above)

• Developed a Personal Web Annotator that
  dynamically adds information to pages being
  downloaded to reflect local information.

• Developed an Ad Buster for eliminating unwanted
  advertisements from downloaded web pages.

35
Network-Attached Secure Disks
Got your attention?

• The goal is to produce storage systems that
  are - off-the-shelf - massively parallel -
  widely distributed - with low-latency file
  access.

• Inorder to do so, it is necessary to develop -
  storage architectures - interfaces - protocols
  that reduce access latency.

36
The Approach

• Today's parallel distributed storage systems
  need to inherit compelling scalability of the
  switched-networking technologies.
• To realize this potential, it is important to
  make some changes in systems architecture -
  avoid repeated store-and-forward copying of file
  segments through data "gateways. - counter
  the unacceptability of unavailability - manage
  the global resources of distributed and
  parallel devices - limit data insecurity caused
  by diverse organization loyalties

37
The NASD project
Attached
Secure
Disks
Network

• NASD comprehensively reduces access latency by
  - promoting devices to first-class network
  clients - streamlined transfer protocols -
  transfer with high-bandwidth - restructuring
  file and storage-systems-software layering - to
  enable offloading to devices and clients - to
  support diverse high-level file-system
  functionalities - increasing the quality of the
  workload information available at devices
  - to improve self-management decisions - to
  improve device-specific optimization

38
done scince 1997

• Developed a preliminary NASD-optimized parallel
  file system and demonstrated linear bandwidth
  increases with increasing client demand and
  storage resources.

• Developed a shared striped storage simulator
  and scalable optimistic synchronization
  protocols for server clusters

• Developed and implemented a native object
  storage system for NASD devices.( replacing
  UNIX file system )

• Expanded the prototype implementation platforms
  

- DEC, at first ( Alpha, Digital UNIX )
- Pentium 200MHz
/ Linux
39
and done even more...
and Im stillLATE!
oh,mythey reallytook their time...

• Instrumented and measured Alpha-based NASD
  prototype hampered by no hardware acceleration
  for copying, the prototype code spends 70-95
  of its work in the communications protocol stack
  (DCE/UDP/IP) and still achieves 60-90 of the
  speed of the fastest operations.

• According to the newest VLSI trends,
  statistical data mining, multimedia processing,
  and disk-embedded acceleration for the data-
  base search, the NASD disks will soon have more
  cycles than NASD needs.

40
The DAML Darpa Agent Markup Languge

• Most of the information on the WWW is currently
  being represented using the HTML.
• HTML allows us to visualize the information on
  the Web, but it doesn't provide capability to
  describe the information in ways usable for
  software programs,to find or interpret it.

ALGOL?
41
markup languages

• The XML allows information to be more accurately
  described using tags.
• Unfortunately, the XML has a limited capability
  to describe the relationships (schemas or
  ontologies) with respect to objects.
• The DAML provides a rich set of constructs,with
  which to create ontologies, and to markup
  information, so that it is machine readable and
  understandable.

42
Microcontrollers
43
End-to-End Reservation Servicesin Real-Time Match

• Our objective is to create distributed real-time
  operating system services supported by a
  guaranteed end-to-end resource reservation
  paradigm.
• To this will be added scalable micro-kernel,
  and a performance monitoring infrastructure
  (control visualization tools).
• So far, the real-time applications span
  between - battlefield simulations - signal
  processing - sensor data fusion - autonomous
  navigation, etc.

44
(No Transcript)
45
continoued

• Future real-time platforms will be more dynamic
  with needed support for - multimedia
  user-interfaces - distributed audio and video
  tele-conferencing on demand - remote
  retrieval of multimedia documents -
  tele-medicine
• Such a wide span of needs, shall require wide
  span of requirements in terms of -
  latency - user interfaces - computational power
  and evolvability

46
The Approach

• A uniform OS-enforced system which spans the
  complete spectrum of requests, can improve
  productivity of mentioned, real-time embedded
  systems - by supporting predictable and
  analyzable timing behavior for temporal
  correctness. - Using kernel mechanisms for the
  usage of system resources with guaranteed
  temporal protection boundaries between
  applications. - Enabling guaranteed but adaptive
  behavior of applications in
  response to changing mission dynamics. -
  Facilitating development with a powerful and
  flexible set of integrated control tools.

47
...continoued

• Being capable of supporting varying timing and
  processing requirements - from real low
  latencies to high latencies - from low to high
  data rate - simple to complex multimedia data.
• Things done so far - the processor reservation
  scheme in RT-Mach. - resource monitoring tool
  (rt-mon) which allows both the visualization
  and interactive control of the resource
  allocations. - other things, such as protocols,
  a/v conferencing environment, and so on

48
VLSI Design
49
Smart MemoriesA Universal Computing Element

• Develop the computing infrastructure (both
  hardware and software) to support the next
  generation of embedded applications

• The applications and processors become more
  complex.

• Today's solution of building custom silicon for
  each application will need to change.

• Smart memories enable one to create a more
  general component that will provide close to
  the power, performance and manufacturing cost
  of a custom solution.

• This effort will cover the model all the way
  from the software components down-to VLSI
  design.

50

• The development of memory-chips, during the
  history.
• However, simple information storage is hardly a
  thing to be characterized as smart.

51
The Approach of Smart Mems

• Two important changes in computers systems -
  the basic wire limits of the underlying VLSI
  technology. - the changing nature of the
  application space toward more streaming data.

• In future advanced chips, wires will need active
  repeaters to help reduce the long wiring
  delays.

• By making them into switches we create
  re-configurable wires with the same performance
  as dedicated wires

• These wires can be use to connect mem-banks with
  each other, or with the processor, and with
  usage of re-configurable logic, they can make
  smart memory block.

52
...whats it good for?

• For parallel applications, smart memories will
  be used to enhance the functionality of
  individual nodes. - For example, to improve
  instruction bandwidth for database
  applications with poor instruction cache
  behavior.

• For stream-based applications, the
  interconnect will be reconfigured to deliver
  the performance of hardwired designs.

• For legacy applications, the smart memory will
  be used to implement mechanisms for dynamic
  parallelism, automatically using speculative
  execution techniques.

53
.and what is needed for it

• Unified, software development environment for
  this system, supported by sophisticated static
  and dynamic programming tools.

• The start will be in collecting a number of the
  new applications, developing programming
  models that express the parallelism of these
  applications.

• Then, working on compiler and OS tools, that
  will efficiently map these models on to
  SmartMemory hardware.

• Finally, performance analysis tools will provide
  feedback to help the designer tune his/her
  application, while later tools will do some of
  this tuning automatically.

54
Wireless Adaptive Mobile Information Systems

• Highly adaptive, compact, and efficient radio
  transceiver capable of supporting data rates as
  high as 60 Mbps and operating within the 2.4
  GHz ISM band.

• The expected results of this work are

- state-of-the-art CMOS chip set
- optimum conformal antenna array
- set of access protocols and receiver algorithms
for reliable communications.

• One of mayor goals, will be developing highly
  flexible, mobile battlefield networks.

55
The Approach
dont worry...

• The design strategy is needed that combines
  expertise in

- digital VLSI design,
- analog-RF integrated circuit design,
- communication system design,
- and antenna design.
getting jumpy?
it s true!
intruder to your brains?
ever felt someones watching U?
56
step by step...

• A highly versatile single-chip VLSI ASIC will
  provide in real-time, power to realize

- algorithms for adaptive equalization
- adaptive beam-forming
- fast frequency hopping within the 2.4 GHz ISM
band

• The chip will implement advanced signal
  processing algorithms to allow operation at any
  symbol rate within the range of 625 kBaud to 10
  MBaud as well as support for multiple QAM
  constellations.

57
and more

• The key in reducing the size and power
  consumption of the radio will be a fully
  integrated, single-chip RF/analog integrated
  circuit to realize up/down conversion between
  IF and 2.4 GHz.

• The integrated components must be

- highly linear poser amplifier
- low noise amplifier
- channel select filters
- mixer the chip

• All together, it will be a key-factors in the
  realization of small-to-medium size antenna
  arrays in a portable unit.

58
continoued

• Besides developing hardware for the project,
  system software is also being written

- developing the protocols algorithms
- achieving the highest possible
signal-to-interface ratio
- thus, increasing the system capacity.

• Novel algorithms for antenna array and element
  design are critical in the design of small
  conformal elements and arrays integrated into
  the casing of the radio unit.

• But, they can help identify the interaction of
  the antenna array with the surroundings and
  improve the dual-band antenna construction.

59
Dynamically Re-programmable Architecture

• Develop a new paradigm in adaptive computing -
  soft hardware, which can support processing
  at or near hardware speeds while achieving the
  programmability and silicon re-use .

• A new FPGA architecture is needed, the one
  that can be dynamically programmed during
  runtime along with a set of CAD tool support
  and a user-level runtime environment

• The first achievements are to be tested for the
  set of military applications. ( starting from
  1997. )

60

• The current tools for the VLSI design do not
  support re-configurable way of projecting
  circuits.

• It is believed that this concept mimics the way
  neural cells re-configure in human brain.

61
DRA CAD Tools

• A Dynamically Reprogrammable Architecture (DRA)
  will be defined and simulated at the system
  as well as the logic levels.

• Architecture introduces a new concept called
  Dynamically Programmable Memory Array (DPMA)
  where the re-configurable cells are shared
  among different memory resources.

• In order to support the above architecture, a
  new generation of CAD tools will be
  investigated. (at firs, just to map
  computations on DRA chip)

62
conclusions

• Later on, the CAD-like tools will be developed
  for - for architectural synthesis - design
  space exploration - partitioning - mapping -
  and temporal-spatial placement on the DRA .

• A runtime environment will allow the management
  of configuration memory as well as
  architectural reconfiguration to adapt to the
  existing data and computational inputs.

63
The Summary

• So every good tree brings good fruit
  and every bad tree brings bad fruit.

• Good tree cant bring bad fruit, neither
  can bad tree bring good fruit.

• And every tree, not bringing the good fruit,
  is being cut and thrown into flames.

• And so, by the fruit they bring, you shall
  recognize them.

Matthew, 16-20
64
On the day of 24.March 1999, NATO alliance
started their bombing campaign against Yugoslavia
Loads of high-sophisticated weapons were unloaded
on the heads of these children
Now, it becomes clear, what does the letter
Din DARPA mean.
65
Kick it, Belgrade ETFKick it, Belgrade ETF

• Please, dont try this athomewarranty seal does
  not support such an action.
• Democracy stops for no-one...

back