Protocols for distributed system

Views:
 
Category: Entertainment
     
 

Presentation Description

No description available.

Comments

By: manjuladdha (21 month(s) ago)

very nice presentation please post these ppts as early as possible

By: kshitijkumar23 (41 month(s) ago)

very useful presentation

By: amjadanwar (45 month(s) ago)

Very useful Presentation- Can u please me to download of this?

Presentation Transcript

Protocols for distributed system : 

Protocols for distributed system Previously mentioned protocols cannot be used by distributed systems because of the following issues: Transparency : Must be location independent because process can be migrated leading to changes in identifier address Client-server based communication: Should support simple, connectionless protocol having features like request/response behaviour Group Communication: Flexible transfer of n data to n entities should be possible Security: Encryption should be used only if the path that it takes in the network cannot be trusted, otherwise it should not be used, since it is expensive Network Management: Adding or removing a node should be done automatically Scalability: Should scale well and allow efficient communication to take place in both LAN and WAN

Continued : 

Continued Two communication protocols have been designed to achieve higher throughput and fast response in distributed systems are: Versatile Message Transport Protocol (VMTP) : provides group communication facility and implements a secure and efficient client-server-based communication protocol Fast Local Internet Protocol (FLIP) : Supports transparency, efficient client-server based communication, group communication and easy network management

VMTP : 

VMTP Connectionless protocol Has special features to support request/response behaviour between client and one or more server processes Based on concept of a message transaction that consists of a request message sent by a client to one or more servers followed by zero or more response messages sent back to the client by the server Most message transactions involve a single request message and a single response A response is used to serve as an acknowledgement for the corresponding request, and a response is usually acknowledged by the next request from the same client Using special features, a client can request for an immediate acknowledgement for its request or a server can explicitly request an acknowledgement for its response

Slide 4: 

To support transparency and to provide group communication facility, entities in VMTP are identified by 64-bit identifiers that are unique, stable and independent of the host address This allows entities to be migrated and handled independent of network layer addressing A portion of the entity identifier space is reserved for entity group identifiers that identify a group of zero or more entities VMTP provides selective retransmission mechanism Packets of a message are divided into packet groups that contain upto maximum of 16kb of segment data

Slide 5: 

Data segment is viewed as a sequence of segment blocks each of 512 bytes allowing portion of the segment in a packet group to be specified by a 32-bit mask Each packet contains a delivery mask field that indicates the portions of data segment the packet contains Maximum number of blocks per packet determined by the network maximum packet size When packet group is received, the delivery masks for the individual packet are ORed together to obtain a bitmap indicating which segment blocks are still outstanding An acknowledgement packet contains this bitmap, and the sender selectively retransmits only the missing segment blocks

Slide 6: 

VMTP uses rate based flow mechanism Packets in a packet group are spaced out with interpacket gaps to reduce the arrival rate at the receiver It allows clients and servers to explicitly communicate their desired interpacket gap times and to make adjustments based on selective retransmission If bitmap shows that every other packet has to be retransmitted, sender reasonably increases interpacket gap If bitmap shows that every fourth packet has to be retransmitted, sender again increases interpacket gap If none is missing, interpacket gap is reduced It minimizes the performance penalty arising from over flooding of packets from fast sender to slow receiver

Slide 7: 

VMTP differentiated between idempotent and nonidempotent operations Idempotent operation is one whose execution can be repeated any number of times without any side effects In VMTP, a server can label a response to indicate that a message transaction was idempotent With this, arrangements need not be made for retransmission of response when it is lost because the server can reproduce the response when the request is retransmitted When a response is nonidempotent, VMTP prevents the server from executing a request more than once

Slide 8: 

VMTP provides a rich collection of optional facilities that expand its functionality and efficiency in various situations Conditional message delivery helps a client to specify that its message should only be delivered if the server is able to process it immediately

FLIP ( Fast Local Internet Protocol) : 

FLIP ( Fast Local Internet Protocol) Connectionless protocol Main features are transparency, security, easy network management, group communication facility and efficient client-server based communication facility FLIP identifies entities , called network service access points (NSAPs) with location independent 64-bit identifiers Sites on internetwork can have more than one NSAP for more than one entity Each site is connected to internetwork by FLIP box that either can be software layer in the OS of corresponding site or can be run on separate communications processor Each FLIP box maintains a routing table mapping NSAP addresses on data link addresses Special primitives are provided to dynamically register and unregister NSAP addresses into routing table of FLIP box

Slide 10: 

An entity can register more than one address in a FLIP box FLIP uses one way mapping between the private address used to register an entity and the public address used to advertise the entity One way encryption function is used to ensure that one cannot deduce the private address from the public address Entities that know public address of an NSAP are not able to receive messages on that address, because they do not know the corresponding private address

Slide 11: 

FLIP messages are transmitted unreliably between NSAPs Flip message can be of any size less than 232-1 bytes If a message is too large for a particular network, it is fragmented into smaller chunks called fragments Fragment typically fits in a single network packet Basic function of FLIP is to route an arbitrary length message from the source NSAP to the destination NSAP Path selection policy is based on information stored in the routing tables of each FLIP box about the networks to which it is connected Two parameters: Network weight, security bit Low network weight means that the network is desirable on which to forward a message. Based on bandwidth and delay Secure bit indicates whether sensitive data can be sent unencrypted over the network or not

Slide 12: 

Types of calls provided in FLIP for sending a message to a public address: Flip_unicast Flip_multicast Flip_broadcast Provides point to point and group communication facilities Group communication protocols make heavy use of flip_multicast Group of n processes can be addressed using one flip address, even if they are located on multiple networks

Slide 13: 

FLIP does not encrypt messages but provides following two mechanism for security Sender can mark the address sensitive by using security bit. Such messages are routed only over secure networks Messages routed over untrusted network by a FLIP are marked unsafe by setting the unsafe bit. If safe route exists, sender tries to send sensitive messages in unencrypted form but with the security bit set If while routing, at some stage no further trusted path is found , it is returned to sender, with unreachable bit set Then the sender encrypts the message and retransmits it with security bit cleared Message encryption is done only when required

Slide 14: 

FLIP supports easy network management because dynamic changes in network configuration are automatically handled Only network management jobs require human intervention FLIP relies on system administrator to mark a network interface as trusted and untrusted, since it cannot determine on its own FLIP does not provide full support in WAN Does not scale well in large WANs