Sissejuhatus kursusesse

1
Sissejuhatus kursusesse

• Esimese loengu kava
• Korraldus
• Hajutatud süsteemid mis need on?
• Ülevaade teemadest
• Materjalid
• Algust teemadega ...
• Copmputer for the 21st century
• Kõige alus Paralleelsus

2
Hajutatud süsteemid

• Näited
• 1.
• 2.
• 3.
• 4.
• ...
• N.

3
Hajutatud süsteemid

• Mingit ühest tähendust fraasil hajutatud
  süsteemid ei ole.
• Variante
• Hardware Autonomous computers, network links
• Software Communication protocols, system and
  application software

4
Hajutatud süsteemid

• Tanenbaum
• A collection of independent computers that appear
  to the users as a single coherent system
• autonomous computers
• connected by a network
• software specifically designed to provide an
  integrated computing facility

5
Hajutatud süsteemid

• Leslie Lamport
• You know you have a distributed system when the
  crash of a computer youve never heard of stops
  you from getting any work done.
• inter-dependencies
• shared state
• independent failure of components
• partial failures

6
Hajutatud süsteemid

• Coulouris
• System of networked computers that
• communicate and coordinate their actions only by
  passing messages
• concurrent execution of programs
• no global clock
• components fail independently of one another
• Vt. figs1

7
Mis on kursuse huviorbiidis?

• Vaatame võrgurakendusi, mis töötavad korraga
  hulga
• erinevate arvutite peal igal arvutil on
  tarkvara, mis suhtleb üle
• võrgu teistes arvutites oleva tarkvaraga.
• Lihtne näide klient-server. Üldiselt vaatame aga
  süsteeme, kus on
• servereid palju ja kliendil ning serveril pole
  eriti vahet.

8
Mida hajutatakse?

• What is distributed?
• CPU Share computational resources (MPI, PVM,
  DCE, Beowulf, Mosix)
• Data Share data (sockets, remote procedure
  calls)

9
Standardnäiteid

• Automated banking systems
• Tracking roaming cellular phones
• Global positioning systems
• Retail point-of-sale terminals
• Air-traffic control
• The World Wide Web
• NFS Sun Network File System
• Napster Kazaa combined P2P server system
• Gnutella, Morpheus P2P file sharing systems
• Seti_at_home CPU sharing system for large
  computations

10
Miks hajutatud?

• Share resources
• Personalise environments
• Location independence
• People information are distributed
• Performance cost
• Modularity expandability
• Availability reliability
• Scalability

11
Probleemid I

• Naming
• Communication
• Software structure
• well-defined interfaces
• abstractions/layering support services
• Scale
• Partial failure
• detection, masking tolerance
• recovery

12
Probleemid II

• Concurrency
• No global clock
• Inconsistent states
• Independent failures
• Scale

13
Kursuse plaan

• Sissejuhatus ja ülevaade
• 2. Põhiküsimused
• - Sünkro, locking, latency jne.
• 3. Klassikalised hajutatud süsteemid
• - DNS
• - WWW-proxyd
• - RPC, hajutatud failisüsteemid (NFS)
• 4. Server-centric süsteemid ja näited
• - Napster
• - Kazaa
• - Seti_at_home
• 5. P2P süsteemid ja näited
• - Gnutella
• - Freenet

14
Kursuse plaan

• 6. XML-RPC ja SOAP
• 7. P2P ja Jxta
• 8. Intelligentsed agendid.
• 9. Bluetooth võrgud

15
Materjalid

• Kursuse koduleht
• http//cs.ttu.ee/kursused/itv0040/
• Google
• Loengud

16
The Computer for the 21st Century

• Mark Weiser XEROX PARC
• Presented By Mihail Ionescu

17
Outline

• Brief presentation of the paper
• What happened in these 10 years in this domain
  since the paper was published
• Conclusions

18
Motivation

• The best technologies are those that disappear.
• Writing is the best example ? it is almost
  invisible for us in the sense that it is used
  almost without realizing it.
• Possessing the most powerful computer is like
  having just one book, does not matter how big it
  is.
• Also, the computer screen demands full attention.

19
Contributions

• Not real technical contributions, since it is not
  a technical paper.
• One of the first (if not The first) papers that
  try argue for the need that the silicon
  technology should vanish in the background.
• Some real life scenarios of using this paradigm.
• Identity the technical requirements of an
  infrastructure for supporting the ubiquitous
  computing.

20
Ubiquitous Computing

• The computers will be everywhere pens, cans,
  pads, boards, coffee machines, alarm watches,
  etc.
• Some of the devices (like traffic lights, ovens,
  etc.) already have computers incorporated, but
  without the possibility of communicating between
  each other, which is very important.
• At least hundreds of components in each room.
• Electronic badges are already in use (I think).
• Current prototypes to build pads, tabs.

21
Technical Requirements

• Three main parts
• Cheap, low-power computers that include
  convenient displays
• A network to tie them together
• Software systems to implement ubiquitous
  applications

22
Low-power Computers

• The first requirement is not so difficult to be
  met
• Even at the time when the paper was written such
  devices existed
• Today there are such devices with processors of
  200 MHz (maybe more), RAM of 128 M and even color
  displays
• The devices should be simple, no AI or other
  complex technologies

23
Network

• Data transmission rates for both wired and
  wireless networks are increasing.
• The current systems cannot (and will not) support
  hundreds of machines per room.
• Three types of network connections tiny range
  wireless, long range wireless and very high speed
  wired ? it is a need of a single kind of network
  connection that somehow serve all three
  functions.

24
Applications

• New operating system idea that does not assume a
  relatively fixed configuration of hardware and
  software.
• New systems that have to deal with the diversity
  of inputs from the user.

25
What Happened in 10 Years

• Almost nothing that was described in the paper
  (even if Weiser predicted 20 years).
• The devices are here, maybe more powerful that
  Weiser imagined.
• However, the network does not exist in the
  generality imagined by Weiser.
• Many of the devices contain computers, but the
  computers do not communicate in a ubiquitous
  way.

26
Why?

• I think that mainly because what Weiser suggested
  is not practical nor possible.
• The second requirement (the network part) is much
  more complex than initially thought.

27
Technical Problems

• It is not clear whether this network will be
  based on the current Internet infrastructure (IP
  based), or it will require a new, completely
  different approaches.
• Even much smaller ad-hoc networks could not use
  IP as the based protocol ? new protocols like
  Blue Tooth, etc.
• New schemes of routing content based, smart
  messages, etc.

28
Other Problems

• Security would become a nightmare. The idea that
  cryptographic techniques will solve this problem
  is a joke.
• Flexibility.

29
Conclusions

• A good starting paper, that tries to present what
  ubiquitous computing is and how it can be used.
• However, the paper suggests more that this idea
  is not practical and extremely difficult to
  implement in a robust manner.
• Low power computers exist and will be used in a
  lot of devices, but not in a global network.
• Some much narrow projects might benefit from low
  power computers communicating between each other,
  like sensor networks.

30
Alused paralleeltöö

• Kui meil on süsteemis hulk arvuteid, siis
  tüüpiliselt nad teevad kõik kogu aeg midagi ei
  ole nii, et üks töötab ja teised kõik ootavad
  tema taga.
• Mõne tegevuse juures aga peab üks masin teise
  taga ootama.

31
Fine-grained and coarse-grained

• Parallelismi jaotatakse tüüpiliselt kaheks
• fine-grained (väikeseid koodijuppe tehakse
  paralleelselt hea näide on protsessori-sisene
  automaatne parallelism (pipelining, pre-fetching
  jne))
• coarse-grained (paralleelselt tehakse suuri ja
  hulk aega võtvaid koodijuppe)
• Näide maleprogramm ja otsipuu.

32
Raske probleem paralleelsuse juures

• Multiprotsessor-süsteemides (mitu protsessorit
  ühes arvutis ühe mälu peal) on suurim probleem
  selles, et kuidas protsessorid oma lokaalseid
  cache sünkroniseerivad
• Mälu on ühine
• Cached on aga igal protsessoril omad
• Tekib oht, et ühe protsessori cache sisu erineb
  teise protsessori cache sisust, võib viia
  lihtsalt valede rehkendusteni!!

33
Hajutatud võrgusüsteemides aga ...

• Tüüpiliselt on igal masinal oma protsessor(id),
  oma mälu ja oma ketas ühist infot loomu poolest
  üldse pole.
• Infot tuleb ühest masinast teise saata vastavalt
  vajadusele.
• Vältimaks info igakordset üleküsimist võib olla
  hea mõte osa infot vahel cacheda.
• Siis aga tekkivad jälle analoogilised probleemid
  multiprotsessor-masinatega.

34
Sünkroonsed ja asünkroonsed protsessid

• Sünkroonsed on sellised paralleelprotsessid, kus
  kõigil protsessoritel on ühine kell, mille abil
  neid juhtida (näide multiprotsessor-süsteemid).
• Asünkroonsetes protsessides pole ühist kella
  (näide võrguserverid, eriti olukorras, kus
  tcp/ip ühenduse aeg on teadmata ja muutuv suurus)

35

• Lõpp