Názory k článku Písmenková polévka aneb Sítě ve zkratce II.

(standard vyšel ze specifikace DIX, Digital-Intel-Xerox, označená Ethernet II, ale je s ní kvůli rozdílu ve struktuře datového rámce neslučitelná)

Nová revize normy 802.3 připouští i Type Encoded rámce, takže toto tvrzení není pravdivé.

* pracují v režimu bez spojení (connectionless) - před vysláním dat nepotřebuje odesilatel navazovat spojení se zamýšleným příjemcem dat, udržovat ho po celou dobu komunikace a následně uzavřít, místo toho v LAN vysílá zdrojová stanice i bez momentální znalosti dostupnosti či existence cílové

Toto tvrzení jsem nepochopil. ATM coby klasický představitel WAN musí na PVC navazovat spojení a pak je uzavírat?

* podporují přenos rámců (tedy i dat) proměnné délky (v omezeních daných jednotlivými normami).

nechápu, proč by zrovna toto měla být charakteristika LAN sítí. K tomu není žádný důvod.

(Ve sve odpovedi pouziji citaty z normy 802.2, Internetworking with ATN A. Allese a ze svoji knihy; pro kontext - pokud to bude cist i nekdo jiny - pridavam obsaznejsi pasaze):

1) ATM je technologie pracujici v rezimu se spojenim, podobne jako ostatni v drivejsim clanku zminovane technologie WAN (pozn: ednu jsem nezminila, a to SMDS/CBDS, ktera vyjimecne nabizi sluzbu bez spojeni).

Vetsinou trochu mate rozdil mezi PVC a SVC, ale i u PVC se porad musi navazovat spojeni, i kdyz to nedela uzivatel, ale provozovatel:

"ATM networks are fundamentally con-nection
oriented. This means that a virtual circuit needs to be
set up across the ATM network prior to any data transfer.
ATM circuits are of two types: virtual paths, identified by
virtual path identifiers (VPI); and virtual channels, identified
by the combination of a VPI and a virtual channel identifier
(VCI). A virtual path is a bundle of virtual channels, all of
which are switched transparently across the ATM network on
the basis of the common VPI.

The switch operation is so simple because external mech-anisms
set up the local translation tables prior to the trans-mittal
of any data. The manner in which these tables are set
up determine the two fundamental types of ATM connections:
* Permanent Virtual Connections (PVC): A PVC is a con-nection
set up by some external mechanism, typically
network management, in which a set of switches between
an ATM source and destination ATM system are pro-grammed
with the appropriate VPI/VCI values. As is dis-cussed
later, ATM signaling can facilitate the set up of
PVCs, but, by definition, PVCs always require some
manual configuration. As such, their use can often be cum-bersome.
* Switched Virtual Connections (SVC): An SVC is a con-nection
that is set up automatically through a signaling
protocol. SVCs do not require the manual interaction
needed to set up PVCs and, as such, are likely to be much
more widely used. All higher layer protocols operating
over ATM primarily use SVCs, and it is these that are pri-marily
considered in this paper."

"In a connectionless packet service, each transfer of data between source
and destination is considered as independent, autonomous, transaction.
Therefore the addressing information must be included in every packet
and resolved at each switching device. Switching tables then associate
destination addresses with the next switching device along the path.
In connection-oriented packet service, end station pairs are associated
with a connection identifier. Each switching device constructs a table
mapping connection identifiers to outbound ports according to the path
established prior to sending the first packet into the network

PVC in the usual meaning is a VC that is not signaled by the end points.
Both of the endpoint (user) VC values are manually provisioned. The
link-by-link route through the network is also manually provisioned. If
any equipment fails, the PVC is down, unless the underlying physical
network can re-route below ATM. So a PVC is a VC which is statically
mapped at every point in the ATM network. A failure of any link that a
PVC crosses results in the failure of the PVC.

A SVC is established by UNI signaling methods. So an SVC is a demand
connection initiated by the user. If a switch in the path fails, the SVC
is broken and would have to be reconnected.

Cell switching is connection oriented, like the telephony network. In
contrast, frame switching is connectionless; a packet can take any route
across a network from its source to its destination. In cell switching,
a connection is first negotiated and established across multiple
switching points. An appropriate circuit is then set up, based on the
amount and duration of data needed to be transported through the network
in the form of ATM cells. Depending on the virtual circuit, it is torn
down when the transmission is complete (PVC), or after an idle timeout
period (SVC). The ATM switch provides dedicated circuits or paths
dynamically or "on the fly" between users at each end of the network. "

2) v pouzivani datovych jednotek na prislusne nejvyssi vrstve dane architektury LAN/WAN (paketu/ramcu/bunek) fixni delky nebo promenne delky je znacny rozdil: a) pro rychlost zpracovani v propojovacim zarizeni je lepsi fixni delka datove jednotky (zejmena, pokud je mala), minimalne je dulezita fixni delka zahlavi (presneji ridicich dat, ktere je treba kontrolovat); b) pro podporu prenosu dat ruznych charakteristik (hlasu, obrazu v realnem case, datovych souboru apod.) je idealni datova jenotka fixni, a to velmi male delky (proto napr. ATM s 53 B).

V LAN (podle IEEE) se ppouzivaji ramce promenne delky, zatimco ve WAN (ATM) se vyuzivaji bunky pevne delky.

"Packet switching is a standardized multiplexing method. Cell switching
or cell relay is a packet-like multiplexing and switching technology. It
handles high-speed (high-bandwidth), low-latency (low-delay),
mixed-media communication (voice, data, video) in support of
contemporary networking requirements, using short-data units of fixed
length. Fixed-length cells make the processing associated with switching
more efficient; in a way, they make the arrival distribution more
predictable. In particular, fixed-length cells allow pipelining of
functions, which, in turn, facilitate hardware implementations."

"...(ATM) Cells have a fixed, short length - this framework allows high-speed
switching nodes, since the logical decisions are straightforward. Cell
length is set to 53 bytes in ATM (48 for user information and 5 for
overhead, compromise between 32 and 64 bytes). The header identifies
cells belonging to the same virtual circuit. Cells are switched by means
of the label in the header. For comparison, classical packet switching
uses large packets (128 or 256 octets) and, unlike cell relay, requires
that every packet be brought entirely into each node and queued for
output. As a result, classical packet switching suffers response time
and throughput problems. "

...There is a straightforward switching goal of supporting wire-speed performance for
fixed-sized cells with fixed-length lookups..."

3) Ethernet 2 vs 802.3: pokud je takto postavime proti sobe, mame na mysli jejich odlisnost v tvorbe ramcu a zapouzdrovani bud LLC nebo primo dat sitoveho protokolu do casti "data".

Je pravda, jak pripominate, ze po ctvrtstoleti pusobeni Ethernetu na trhu se IEEE premohlo a zaclenilo (pridalo jako volitelnou moznost) ve sve verzi 802.3 z roku 2000 moznost zapouzdreni i ve forme odpovidajici puvodni firemni specifikaci, ale ja bych byla velmi opatrna a nenazyvala bych je kompatibilnimi. Pokud by IEEE zrusila puvodni zapouzdreni a priklonila by se k zapouzdreni podle Ethernetu 2, to by bylo neco jineho...

Jak jsem listovala normou 802.2, prislo mi zajimave pridat nasledujici citaci jako doplnek k LLC a komunikaci se spojenim/bez spojeni v LAN (ktere vetsinou pouzivaji LLC typu 1):

(There exist) "four distinct "classes" of LLC operation. Class I provides
data-link-connectionless-
mode service only. Class II provides data-link-connection-mode service
plus data-link-con-nectionless-
mode service. Class III provides acknowledged-connectionless-mode
service plus data-link-con-nectionless-
mode service. Class IV provides acknowledged-connectionless-mode service
plus data-link-connection-
mode service plus data-link-connectionless-mode service. Any one of
these classes of operation
may be supported."

V LAN (podle IEEE) se ppouzivaji ramce promenne delky, zatimco ve WAN (ATM) se vyuzivaji bunky pevne delky.

Takze MPLS je podle vas klasicky predstavitel LAN siti? :-) Dejte si pozor, at se nedostanete do nejakeho bludneho kruhu :-))

Mozna by stalo za to pripomenout hlavni duvody proc ATM ma fixni delku ramce: zarucuje se tim kvalita sluzby a minimalizujeme ruzne parametry typu jitter a spol.

Je pravda, jak pripominate, ze po ctvrtstoleti pusobeni Ethernetu na trhu se IEEE premohlo a zaclenilo (pridalo jako volitelnou moznost) ve sve verzi 802.3 z roku 2000 moznost zapouzdreni i ve forme odpovidajici puvodni firemni specifikaci, ale ja bych byla velmi opatrna a nenazyvala bych je kompatibilnimi. Pokud by IEEE zrusila puvodni zapouzdreni a priklonila by se k zapouzdreni podle Ethernetu 2, to by bylo neco jineho...

Ja bych naopak nazval tyto normy kompatibilni, tedy minimalne tak, ze Ethernet II je podmnozina IEEE 802.3.

Predstava, ze by IEEE zrusila svuj druh zapouzdreni je smesna, uz z toho duvodu, ze jsou na nem zalozeny vsechny ostatni normy typu Token-Ring nebo FDDI.

Navic obe verze jak Length Encapsulation, tak Type Encapsulation mohou bez problemu koexistovat a nikdy nemuze dojit ke kolizi.

> Takze MPLS je podle vas klasicky predstavitel LAN siti? :-) Dejte si pozor, at se nedostanete do nejakeho bludneho kruhu :-))

Do bludneho kruhu me dostavate vy ;-§ Nikdy jsem totiz neuvazovala o MPLS jako o "typu" WAN technologie...vedle X.25, Frame Relay nebo ATM. MPLS, jak nazev napovida, je metoda pro prepinani znacek pro ruzne (vyssi) protokoly (MultiProtocol Label Switching). S cilem zrychlit prenos siti prostrednictvim snadneho prepinani na misto komplikovaneho opakovaneho smerovani v siti.

MPLS beha na nejruznejsi infrastrukture; sice definuje svoje zapouzdreni znacky do ruznych typu ramcu (Ethernet, FR, nebo PPP pro Packet over SONET) nebo bunky (ATM), ale jinak nejnizsi vrstvy nedefinuje. Myslim, ze to je rozdil oproti architekture X.25 definujici nejnizsi tri vrstvy, Frame Relay nebo ATM definujici +- dve.

MPLS prece nedefinuje zadny svuj format datove jednotky, jenom jak ten svuj dilci "label" vclenit do stavajici datove jednotky na okraji site a uspesne ji prenest pres nejakou prenosovou sit (tu jednotku).

> Navic obe verze jak Length Encapsulation, tak Type Encapsulation mohou bez problemu koexistovat a nikdy nemuze dojit ke kolizi.

To je opravdu dulezite: sice stanice budou rozumet pouze "svemu" zapouzdreni (pripadne obema v pripade serveru?) a ty druhe ramce zahodi, ale na mediu k problemu nedojde. A ke kolizi - zpusobene nekompatibilitou ramcu - uz vubec ne.

Hm ... Na poslednim INETu bylo prezentovano, ze WorldCom provozuje Frame Relay over MPLS (tj. ze MPLS ma POD Frame Relay protoklem) a pokracuji standardizacni prace na ATM over MPLS. Tak nevim :-)

To by me zajimalo podrobneji: co je pod tim vsim? Nad cim to beha?

Zminujete normalizaci - to by musely byt RFC nebo Internet Drafts (I-D). Se zajmem jsem prohledala I-D a nasla jsem nasledujici:

1) http://search.ietf.org/internet-drafts/draft-malis-sonet-ces-mpls-05.txt (tenhle draft uz vyprchal) - vypovida o tom, jak je dnesni zapouzdrovani dat komplikovane: "MPLS labels and a
new circuit emulation header are used to encapsulate TDM signals and
provide the Circuit Emulation Service over MPLS (CEM) function. The
MPLS encapsulation may be further encapsulated in IP for carriage
across IP PSNs"

2) http://search.ietf.org/internet-drafts/draft-martini-l2circuit-trans-mpls-09.txt - zde se popisuje zapouzdreni vseho mozneho pro prenos siti s MPLS , zapouzdreni, o kterem jsem se zminovala jiz v minulem diskusnim prispevku; cilem je neco prenest prostrednictvim MPLS LSP (label switched path), ale MPLS na necem proste musi fyzicky stat.

Mozna se jen u me jedna o nejake nedorozumeni v tom, co je over a co under v tomto pripade. Vnese do toho nekdo jasnejsi pohled?

P.S. V Network Magazine vysel docela zajimavy clanek porovnavajici prepinane technologie (MPLS, FR a ATM VC) v realnem svete: http://www.networkmagazine.com/article/NMG20020603S0013

Ja prisel az na konec te prednasky, ktera byla o "anything over MPLS". Tak je treba pockat az prijdou proceedings, nebo monitorujte www.isoc.org :-)