Download presentation
Presentation is loading. Please wait.
Published byJasmine Pope Modified over 8 years ago
1
1 Prof. Orhan Gemikonakli Module Leader: Prof. Leonardo Mostarda Università di Camerino Distributes Systems – Consistency (2) Prof. Orhan Gemikonakli - Camerino,
2
Last lecture zState machine replication approach zPaxos protocol 2Prof. Orhan Gemikonakli - Camerino,
3
Outline zReplica management zPermanent replicas zServer initiated replicas zClient initiated replicas zPull versus push protocols zConsistency protocols 3Prof. Orhan Gemikonakli - Camerino,
4
Learning outcome zTo understand the basic concepts related to replica management yPermanent replicas yServer initiated replicas yClient initiated replicas zTo describe and discuss pull and push protocols zTo understand various consistency protocols 4Prof. Orhan Gemikonakli - Camerino,
5
REPLICA MANAGEMENT zReplica management includes yplacement ymechanisms to use for keeping the replicas consistent zThe placement problem is split into two sub problems: yplacing replica servers yplacing content 5Prof. Orhan Gemikonakli - Camerino,
6
Replica Management To decide o where o when and o by whom replicas should be placed. Placement problem* o Placing servers o Placing content * Client aware placement / client unaware placement Prof. Orhan Gemikonakli - Camerino,6
7
Replica Server Placement Prof. Orhan Gemikonakli - Camerino,7 zFChoosing a proper cell size for server placement.
8
Content Replication and Placement zThe logical organization of different kinds of copies of a data store into three concentric rings. Prof. Orhan Gemikonakli - Camerino,8
9
zPermanent replicas can be considered as the initial set of replicas that constitute a distributed data store. zWeb site distribution yfiles that constitute the Web are replicated in different servers, when a request comes in a round robin policy is applied ymirroring (the client choose which mirror to use) 9 Content Replication and Placement Prof. Orhan Gemikonakli - Camerino,
10
zServer-initiated replicas are copies of a data store that are created to enhance performance. zWeb server placed in New York. yit may happen that over a couple of days a sudden burst of requests come in from an unexpected location far from the server yit may be worthwhile to install a number of temporary replicas in regions where requests are coming from 10 Content Replication and Placement Prof. Orhan Gemikonakli - Camerino,
11
Server initiated replicas Prof. Orhan Gemikonakli - Camerino,11 zCounting access requests from different clients.
12
zClient-initiated replicas are more commonly known as (client) caches zIn principle, managing the cache is left entirely to the client. yHowever, there are many occasions in which the client can rely on participation from the data store zData are generally kept in a cache for a limited amount of time zWhenever requested data can be fetched from the local cache, a cache hit is said to have occurred zWhat is the benefit of caching? 12 Content Replication and Placement Prof. Orhan Gemikonakli - Camerino,
13
zReplica management also deals with propagation of (updated) content to the relevant replica servers. 1.Propagate only a notification of an update. 2.Transfer data from one copy to another. 3.Propagate the update operation to other copies. 13 Content Replication and Placement Prof. Orhan Gemikonakli - Camerino,
14
zPropagating a notification is what invalidation protocols do. zThe invalidation may specify which part of the data store has been updated zAdvantages ylittle bandwidth is required to transfer the notification ywork best when there are many update operations compared to read operations zFor instance for a large data store 2 updates may happen without a read in the middle. 14 Content Replication and Placement Prof. Orhan Gemikonakli - Camerino,
15
z Transferring the modified data among replicas is the second alternative ywhen the read-to-write ratio is relatively high z The third approach is not to transfer any data modifications at all, but to tell each replica which update operation it should perform (active replication) 15 Content Replication and Placement Prof. Orhan Gemikonakli - Camerino,
16
Pull versus Push Protocols zAnother design issue is whether updates are pulled or pushed zIn a push-based approach updates are propagated to other replicas without those replicas even asking for the updates. yServer-based protocols are applied when replicas need to be kept identical. yThe read-to-update ratio at each replica is relatively high 16Prof. Orhan Gemikonakli - Camerino,
17
Push Technology Prof. Orhan Gemikonakli - Camerino,17 Client registers interest for service, server pushes new content
18
Pull versus Push Protocols zIn a pull-based approach, a server or client requests another server to send it any updates it has at that moment. yare often used by client caches yis efficient when the read-to-update ratio is relatively low 18Prof. Orhan Gemikonakli - Camerino,
19
Browser cache zThe browser will first check its local cache to see whether the object requested from an origin server has already been stored locally on hard disk. z If it has, and it is up-to-date, the object will be returned by the cache to the browser. zHTTP supports the determination of whether the object is current via the If-Modified-Since header. 19Prof. Orhan Gemikonakli - Camerino, Do you know how the browser cache works?
20
CONSISTENCY PROTOCOLS z A consistency protocol describes an implementation of a specific consistency model. zprimary-based protocol yall processes see all write operations in the same order 20Prof. Orhan Gemikonakli - Camerino,
21
Local-Write Protocols zPrimary-backup protocol in which the primary migrates to the process wanting to perform an update. Prof. Orhan Gemikonakli - Camerino,21
22
Summary zReplica management zPermanent replicas zServer initiated replicas zClient initiated replicas zPull versus push protocols zConsistency protocols 22Prof. Orhan Gemikonakli - Camerino,
23
Next Lecture zFault Tolerance yBasic concepts yFault modelling yFailure masking yRPC Semantics in the Presence of Failures yReliable multicast schemes yVirtual Synchrony yTwo phase commit y|Three phase commit yCheckpointing and Message logging Prof. Orhan Gemikonakli - Camerino,23
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.