Adaptive Middleware for Data Replication

Abstract

Abstract. Dynamically adaptive systems sense their environment and adjust themselves to accommodate to changes in order to maximize performance. Depending on the type of change (e.g., modifications of the load, the type of workload, the available resources, the client distribution, etc.), different adjustments have to be made. Coordinating them is already difficult in a centralized system. Doing so in the currently prevalent component-based distributed systems is even more challenging. In this paper, we present an adaptive distributed middleware for data replication that is able to adjust to changes in the amount of load submitted to the different replicas and to the type of workload submitted. Its novelty lies in combining load-balancing techniques with feedback driven adjustments of multiprogramming levels (number of transactions that are allowed to execute concurrently). An extensive performance analysis shows that the proposed adaptive replication solution can provide high throughput, good scalability, and low response times for changing loads and workloads with little overhead.

Outlier Detection for Fine-grained Load Balancing in Database Clusters

“…A number of independent database replication solutions exist that provide both scaling and strong consistency in replicated database clusters [23,15,20,14,18]. With a few notable exceptions [14] [18] [13], these systems either do not investigate database replication in the context of dynamic adaptation or do not consider fine grain load balancing or allocation.…”

Autonomic Provisioning of Backend Databases in Dynamic Content Web Servers

“…Our study draws on recently proposed transparent scaling through contentaware scheduling in replicated database clusters [6,17,21,23] and in particular on our own previous work [2,3] on asynchronous replication with conflict-aware scheduling.…”

Postgres-R(SI): Combining Replica Control with Concurrency Control Based on Snapshot Isolation

“…In the last decade, a vast body of research has proposed replica control solutions based on various concurrency control mechanisms, e.g., locking [18,21,22,23,8], optimistic [27], multi-version [1], and serialization graph [3] based techniques providing serializable execution, or mechanisms that provide lower levels of isolation [21,29]. These replica control algorithms can be categorized by two parameters as described in [16].…”

“…Many proposals use, as we, group communication systems with total order multicast to help serializing transactions, e.g. [27,18,21,22,25,23]. Except of Postgres-R, they are only evaluated on a simulation basis [27,18,22] or within a middleware [23].…”

“…in 1996 [16] claimed that existing approaches scale badly, and are not suitable for modern applications. Their analysis revived research in database replication leading to many new solutions that attempt to eliminate the limitations pointed out by [16], while still providing global serializability, e.g., [3,27,7,18,21,6,22,26,25,30,1,19,8,23,14,2].…”