TU Wien:Verteilte Systeme VO (Dustdar)/Pruefung 2014-03-13

Fragenausarbeitung

Scalability Geographically - 1 Major issue bzgl. communication

Scalability is a systems ability to grow to meet increasing demands along several dimensions: Size / Geographically / Administrative First, existing distributed systems are often designed for LANs and rely on synchronous and fast communication. The client blocks until an answer has been received - what works in a reliable and fast LAN. At WAN size, which is asynchronous and way slower, this would mean higher delay. Furthermore, communication in WANs is inherently unreliable and almost always point-to-point. It is not possible to broadcast a WAN, so other location services are needed.

Scalability Administrative - 1 Problem beschreiben

A major problem are conflicting policies with respect to resource usage, management and security. e.g: A user can often trust services which reside in a a single domain where the user itself is located. Systems administration may have tested and certified applications which can be trusted by the user. This trust does not expand across domain boundaries.

Warum kann es notwendig sein, dass Datenpakete zwischen Server und Klient komprimiert werden? (oder so ähnlich)

Compressing data can ensure smaller bandwidth usage and therefore higher transmission rate. For instance, for a video stream of 30fps on a 320x240 screen and 24bit per pixel, it would need a bandwidth of 53Mbit. The re-engineered implementation of the X protocol, NX, achieves a bandwidth reduction up to a factor of 1000, which allows X to run through low bandwidth links.

Naming: Entwerfen sie einen globalen Identifier, der auf die Location von Produkten auflösen lässt. Es gibt einen globalen Naming Dienst und jede Company hat einen eigenen Name Server, der zu einem Produkt die Location hat. Jedes Produkt und jede Firma hat einen einzigartigen Identifier. Wie soll also der globale Identifier gewählt werden?

Globaler Identifier = [CompanyIdentifier].[ProductIdentifier].

Naming: Wie wird der zuvor erstellte globale Identifier aufgelöst?

CompanyIdentifier an globalen Server, der Company zurückliefert. Dann Company Name Server mit Product Identifier abfragen und fertig.

Time Sync: Warum ist time-sync wichtig -> 1 Beispiel geben!

Time sync is important because we have to maintain consistency. For example, if a bank hosts two databases for their customers accounts, one in Vienna and one in Tokyo, it could happen that a customer in Vienna deposits 1000E to his already owned 1000E, and at the same instant, the customers interests of 1% are updated in Tokyo. There are two possible outcomes without a synced time base: A) The customer has 2010E on his account or B) 2020E, depending in which order the transactions are performed

Pfeile, Labels und so zeichnen für Upload/Download Model und Remote Access Model

Siehe Folie 356

3 JA / NEIN zu Coda Version Vector -> Wozu wird der verwendet

The CODA version vector is used to detect and maybe resolve inconsistencies. Every server holds a version vector for every file in his VSG: the vector to a file contains all servers with the versions the have seen. If the VVi for file f and server j = k, then server i knows that server j has seen version k of f.

Sec: Password recovery bei Webseite schickt User sein Passwort -> Der meint das das PW unsicher gespeichert wird. Beispiel angeben, wie PW trotzdem secure ist am server

The password is not secure.

(_Either this is a trick question or the question here in VoWi has been incorrectly copied_)

Sec: Passwörter werden von Server gestohlen, sind aber gehasht -> User erklären, warum er sich keine Sorgen machen muss

ELI5 argument: "All of our passwords are hashed, which means that a complex mathematical operation is performed to create a version of the password that humans cannot read called a hash. The exact details of the operation remain secret, which means that the hash remains unreadable by the thieves and thus useless."

Communication: DS mit 1x notification server und mehreren Fileservern, alle anfragen über notification server: Was passiert wenn ein File-server ausfällt?

Once the file server recovers, it must check with the notification server what updates it is missing to be brought to an executable state. Either the notification server delivers the individual pieces of data that need to be updated, all of the data is recopied to the file server or the update logs are sent to the file server and the file server updates itself based on these logs.

Wie zuvor, was passiert wenn der einzige notification server ausfällt?

If the notification server crashes, then all operations must come to a halt. The file servers cannot know which one is the most up to date and there are no guarantees of consistency within the system.

Quelle

Link zum Forum