Query optimizers in RDBMSs search for execution plans expected to be optimal for given queries. They use parameter estimates, often inaccurate, and make assumptions that may not hold in practice. Consequently, they may select plans that are suboptimal at runtime, if estimates and assumptions are not valid. Therefore, they do not sufficiently support robust query optimization. Using ML to improve data systems has shown promising results for query optimization. Inspired by this, we propose Robust Query Optimizer, (Roq), a holistic framework based on a risk-aware learning approach. Roq includes a novel formalization of the notion of robustness in the context of query optimization and a principled approach for its quantification and measurement based on approximate probabilistic ML. It also includes novel strategies and algorithms for query plan evaluation and selection. Roq includes a novel learned cost model that is designed to predict the cost of query execution and the associated risks and performs query optimization accordingly. We demonstrate that Roq provides significant improvements in robust query optimization compared with the state-of-the-art.

Join order selection is a sub-field of query optimization that aims to find the optimal join order for an SQL query with the minimum cost. The challenge lies in the exponentially growing search space as the number of tables increases, making exhaustive enumeration impractical. Traditional optimizers use static heuristics to prune the search space, but they often fail to adapt to changes or improve based on feedback from the DBMS. Recent research addresses these limitations with Deep Reinforcement Learning (DRL), allowing models to use feedback to dynamically search for better join orders and enhance performance over time. Existing research primarily focuses on capturing join order sequences and their representations at various levels, with limited comparative analysis of reinforcement learning methods. In this paper, we propose a novel framework, which integrates Graph Neural Networks (GNN), Tree-structured Long Short-Term Memory (Tree-LSTM), and dueling-DQN. We conduct a series of experiments to demonstrate the potential for improvement in DRL methods.

Query optimizers are an essential component of database management systems (DBMSs) as they search for an execution plan that is expected to be optimal for a given query. However, they commonly use parameter estimates that are often inaccurate and make assumptions that may not hold in practice. Consequently, the optimizer may select sub-optimal execution plans at runtime, when these estimates and assumptions are not valid, which may result in poor query performance. Therefore, query optimizers do not adequately support the robustness of the database system. In this tutorial, we explore the notion of robustness in the context of query optimization, as well as how it is evaluated or even further supported. Firstly, we provide a comprehensive definition for the notion of robustness in this context that accounts for risks associated with execution plans and inaccurate parameter estimates as well as the limitations of the cost models. Next, we review the approaches proposed in the literature to address the issue of robustness, including techniques that rely on query re-optimization, discovering parameters, quantifying robustness, as well as recent techniques that employ machine learning. We focus on comparing traditional cost-model-based methods with modern ML-based techniques in terms of their ability to tackle the challenge of robustness in query optimization. Finally, we discuss the limitations and gaps in the current literature and provide some recommendations for future research directions.

The generation of large amounts of healthcare data has motivated the use of Machine Learning (ML) to train robust models for clinical tasks. However, limitations of local datasets and restrictions on sharing patient data impede the use of traditional ML workflows. Consequently, Federated Learning (FL) has emerged as a potential solution for training ML models among multiple healthcare centers. In this study, we focus on the binary classification task of early ICU mortality prediction using Multivariate Time Series data and a deep neural network architecture. We evaluate the performance of two FL algorithms (FedAvg and FedProx) on this task, utilizing a real world multi-center benchmark database. Our results show that FL models outperform local ML models in a realistic scenario with non-identically distributed data, thus indicating that FL is a promising solution for analogous problems within the healthcare domain. Nevertheless, in this experimental scenario, they do not approximate the ideal performance of a centralized ML model.

Today’s analytics environments are characterized by a high degree of heterogeneity in terms of data systems, formats and types of analysis. Many occasions call for rapid, ad hoc, on demand construction of a data model that represents (parts of) the data infrastructure of an organization, including ML tasks. This data model is given to data scientists to play with (express reports, build ML models, explore, etc.) We present a novel graph-based conceptual model, the Data Virtual Machine (DVM) representing data (persistent, transient, derived) of an organization. A DVM can be built quickly and agilely, offering schema flexibility. It is amenable to visual interfaces for schema and query management. Dataframing, a frequent querying/preprocessing task in analytics applications, is usually carried out by experienced data engineers employing SQL (in the presence of a relational data warehouse) or Python/R: a procedural approach with all the known drawbacks. Dataframes over DVMs are expressed declaratively - and visually, via a simple and intuitive tool. This way, non-IT experts can be involved in dataframing. In addition, query evaluation takes place within an algebraic framework with all the known benefits. I.e. a DVM enables the delegation of data engineering tasks to simpler users. We have seen analogous cases in the past, e.g. with the introduction of SQL. Finally, a DVM offers a formalism that facilitates data sharing, data portability and a single view of any entity – because a DVM’s node is an attribute and an entity at the same time. In this respect, DVMs can excellently serve as a data virtualization technique, an emerging trend in the industry. We argue that DVMs can have a significant practical impact in today’s big data environments.

International audience

Query optimization is a challenging process of DBMSs. When tackling query optimization in the cloud, there exists a simultaneous need of providing an optimal physical query execution plan, as well as an optimal resource configuration among available ones. Cloud computing features like resource elasticity and pricing make the process of finding this optimal query plan a multi-objective problem, with the monetary cost being an equally important factor to query execution time. Apache Spark is a popular choice for managing big data in the cloud. However, query optimization in its SQL module (Spark SQL) involves a number of limitations due to the rule-based nature of its optimizer, Catalyst. We propose a multi-objective cost model for the extension of the query optimizer of Apache Spark, aiming to minimize both objectives of query execution time and monetary cost, as well as a methodology for exploring the space of Pareto-optimal query plans and selecting one. The cost model is implemented and tuned, and an experimental study is conducted to validate its accuracy.

Can we solve finite-horizon Markov decision processes (FHMDPs) while raising low memory requirements? Such models find application in many cases where a decision-making agent needs to act in a probabilistic environment, from resource management to medicine to service provisioning. However, computing optimal policies such an agent should follow by dynamic programming value iteration raises either prohibitive space complexity, or, in reverse, non-scalable time complexity requirements. This scalability question has been largely neglected. In this paper, we propose SIFTER (Space Efficient Finite Horizon MDPs), a suite of algorithms that achieve a golden middle between space and time requirements. Our former algorithm raises space complexity growing with the square root of the horizon's length without a time-complexity overhead, while the latter's space requirements depend only logarithmically in horizon length with a corresponding logarithmic time complexity overhead. A thorough experimental study under diverse settings confirms that SIFTER algorithms achieve the predicted gains, while approximation techniques do not achieve the same combination of time efficiency, space efficiency, and result quality.

The wide adaptation of the Semantic Web and the Resource Description Framework (RDF) has made available many important datasets. The SPARQL query language facilitates the exploration of this information, which is available in a semi-structured way that does not comply with relational data models, deviating from exploration techniques that most researchers are familiar with. Usually, only people with training and extensive knowledge of the RDF model can explore and understand it in depth. We present here a platform that supports the users with querying, exploring and visualizing information available in SPARQL endpoints. A dedicated visualization module, built upon a knowledge database, allows us to provide case-specific visualization solutions for SPARQL query results. The selection is based exclusively on features extracted from the result, without any knowledge about the structure, content and characteristics of the underlying dataset.

A Data Virtual Machine (DVM) is a novel graph-based conceptual model, similar to the entity-relationship model, representing existing data (persistent, transient, derived) of an organization. A DVM can be built quickly, agilely, offering schematic flexibility to data engineers. Data scientists can visually define complex dataframe queries in an intuitive and simple manner, which are evaluated within an algebraic framework. A DVM can be easily materialized in any logical data model and can be "reoriented'' around any node, offering a "single view of any entity''. In this paper we demonstrate DataMingler, a tool implementing DVMs. We argue that DVMs can have a significant practical impact in analytics environments.

The need to reduce energy consumption on a global scale has been of high importance during the last years. Research has created methods to make highly accurate forecasts on the energy consumption of buildings and there have been efforts towards the provision of automated forecasting for time series prediction problems. EnForce is a novel system that provides fully automatic forecasting on time series data, referring to the energy consumption of buildings. It uses statistical techniques and deep learning methods to make predictions on univariate or multivariate time series data, so that exogenous factors, such as outside temperature, are taken into account. Moreover, the proposed system provides automatic data preprocessing and, therefore, handles noisy data, with missing values and outliers. EnForce includes full API support and can be used both by experts and non-experts. The proposed demonstration showcases the advantages and technical features of EnForce.

The wide adoption of the RDF data model, as well as the Linked Open Data initiative, have made available large linked datasets that have the potential to offer invaluable knowledge. Accessing, evaluating and understanding these datasets as published, though, requires extensive training and experience in the field of the Semantic Web, making these valuable sources of information inaccessible to a wider audience. In the recent years, there have been many efforts to create systems that allow the visualization and exploration of this information. Some of there systems rely on techniques that allow them to limit the volume of the displayed information, by providing aggregated, filtered or summarized access to the datasets while others initialize the exploration of the dataset based on actions performed by the users, such as keyword searches and queries. The underlying technique is key for the sustainability of the system, the definition of the requirements that the input must comply with, the datasets that can be visualized as well as the visualization types provided. We present here a survey on these techniques, their strengths and weaknesses as well as the datasets that they can support. The survey will provide the reader with a deep understanding of the challenges regarding the visualization of large linked datasets, a categorization of the developed techniques to resolve them as well as an overview of the available systems and their functionalities.

There has been significant effort in recent years to explore and navigate very large linked datasets, due to the increase of their availability. Many techniques have been developed that extract the information from such datasets and present it to the user as diagrams, while others take advantage of the hierarchies of the datasets to filter and aggregate them, allowing the users to access specific information. In order to overcome the limitations regarding the volume of the presented information, we have developed a novel technique that enables the interactive visualization as one continuous graph of datasets with millions of elements. IVLG is a fully fledged prototype system that implements this technique based on a client-server architecture, enabling many users to have concurrently access to the information through a user-friendly interface. It allows the user to navigate the dataset through different levels of abstraction and locate information using innovative exploration techniques. A carefully designed storage schema along with an API that takes advantage of the appropriate indexing handles datasets with millions of elements without raising any performance issues, even when accessed from devices with limited computational resources. The proposed demonstration showcases the advantages and technical features of IVLG. A series of demonstration scenarios will show how IVLG can adapt accordingly and handle diverse real and synthetic datasets that vary on size and average node degree and how the system functionalities support the user experience and the exploration of the information.

Cloud computing provides computing resources with elasticity following a pay-as-you-go model. This raises Multi-Objective Optimization Problems (MOOP), in particular to find Query Execution Plans (QEPs) with respect to users' preferences being for example response time, money, quality, etc. In such a context, MOOP may generate Pareto-optimal front with high complexity. Pareto-dominated based Multi-objective Evolutionary Algorithms (MOEA) are often used as an alternative solution, like Non-dominated Sorting Genetic Algorithms (NSGAs) that provide better computational complexity. This paper presents NSGA-G, a NSGA based on Grid Partitioning for improving complexity and quality of current NSGAs. Experiments on DTLZ test problems using Generational Distance (GD), Inverted Generational Distance (IGD) and Maximum Pareto Front Error prove the relevance of our solution.

Modern large-scale computing deployments consist of complex applications running over machine clusters. An important issue in these is the offering of elasticity, i.e., the dynamic allocation of resources to applications to meet fluctuating workload demands. Threshold based approaches are typically employed, yet they are difficult to calibrate and optimize. Approaches based on reinforcement learning (RL) have been proposed, but they require a large number of states in order to model complex application behavior. Methods that adaptively partition the state space have been proposed, but their partitioning criteria and strategies are sub-optimal. In this work we present MDP_DT, a novel full-model based reinforcement learning algorithm for elastic resource management that employs adaptive state space partitioning. We propose two novel statistical criteria and three strategies and we experimentally prove that they correctly decide both where and when to partition, outperforming existing approaches. We experimentally evaluate MDP_DT in a real large scale cluster over variable not-encountered workloads and we show that it takes more informed decisions compared to static, model-free and threshold approaches, while requiring a minimal amount of training data. We experimentally show that this adaptation enabled MDP_DT to optimize the achieved profit while being 40% cheaper than calibrated RL and threshold approaches.

Modern large-scale computing deployments consist of complex applications running over machine clusters. An important issue there is the offering of elasticity, i.e., the dynamic allocation of resources to applications to meet fluctuating workload demands. Threshold based approaches are typically employed, yet they are difficult to configure and optimize. Approaches based on reinforcement learning have been proposed, but they require a large number of states in order to model complex application behavior. Methods that adaptively partition the state space have been proposed, but their partitioning criteria and strategies are sub-optimal. In this work we present MDP DT, a novel fullmodel based reinforcement learning algorithm for elastic resource management that employs adaptive state space partitioning. We propose two novel statistical criteria and three strategies and we experimentally prove that they correctly decide both where and when to partition, outperforming existing approaches. We experimentally evaluate MDP DT in a real large scale cluster over variable not-encountered workloads and we show that it takes more informed decisions compared to static and model-free approaches, while requiring a minimal amount of training data.

The fundamental assumption for query rewriting in heterogeneous environments is that the mappings used for the rewriting are complete, i.e., every relation and attribute mentioned in the query is associated, through mappings, to relations and attributes in the schema of the source that the query is rewritten. In reality, it is rarely the case that such complete sets of mappings exist between sources, and the presence of partial mappings is the norm rather than the exception. So, practically, existing query answering algorithms fail to generate any rewriting in the majority of cases. The question is then whether we can somehow relax queries that cannot be rewritten as such (due to insufficient mappings), and whether we can identify the interesting query relaxations, given the mappings at hand.

In a world of wide-scale information sharing, data are described in different formats, i.e. data structures, values and schemas. Querying such sources entails techniques that can bridge the data formats. Some existing techniques deal with schema mapping and view complementary aspects of the problem. Important ones, consider producing all the possible mappings for a pair of schemas, insinuating accompanying semantics in the mappings and adapting correct mappings as schemas evolve. In this work, we consider the problem of discovering mappings as schemas of autonomous sources are gradually revealed. Using as an example setting an overlay of peer databases, we present a schema mapping solution that discovers correct mappings as peer schemas are gradually revealed to remote peers. Mapping discovery is schema-centric and incorporates new semantics as they are unveiled. Mapping experience is reused and possible mappings are ranked so that the best choice is presented to the user. The experimental study confirms the suitability of the proposed solution to dynamic settings of heterogeneous sources.

Service Level Agreements (SLAs) represent service management contracts that are processed by monitoring and measurement mechanisms for the evaluation of the signatories adherence to the agreed service levels during service execution. The paper discusses SLA data management characteristics that need to be considered in the design of data models for SLA documents. The SLA anatomy is introduced with respect to the Web Service Level Agreement (WSLA) [1] language specification. Furthermore, the paper highlights current obstacles for the integration of automated SLA management in the cloud business setting. The contributed SLA data analysis maps SLA terms to data management attributes according to their operational relevance during the SLA activity. We present an SLA digraph model for the automated SLA formulation and data handling. The SLA digraph is introduced as a programming module that sits on the application layer and communicates with backend data stores for the SLA persistence.

Recently, a large number of pay-as-you-go data services are offered over cloud infrastructures. Data service providers need appropriate and flexible query charging mechanisms and query optimization that take into consideration cloud operational expenses, pricing strategies and user preferences. Yet, existing solutions are static and non-configurable. We demonstrate COCCUS, a modular system for cost-aware query execution, adaptive query charge and optimization of cloud data services. The audience can set their queries along with their execution preferences and budget constraints, while COCCUS adaptively determines query charge and manages secondary data structures according to various economic policies. We demonstrate COCCUS's operation over centralized and shared nothing CloudDBMS architectures on top of public and private IaaS clouds. The audience is enabled to set economic policies and execute various workloads through a comprehensive GUI. COCCUS's adaptability is showcased using real-time graphs depicting a number of key performance metrics.

Emerging providers of online services offer access to data collections. Such data service providers need to build data structures, e.g. materialized views and indexes, in order to offer better performance for user query execution. The cost of such structures is charged to the user as part of the overall query service cost. In order to ensure the economic viability of the provider, the building and maintenance cost of new structures has to be amortized to a set of prospective query services that will use them. This work proposes a novel stochastic model that predicts the extent of cost amortization in time and number of services. The model is completed with a novel method that regresses query traffic statistics and provides input to the prediction model. In order to demonstrate the effectiveness of the prediction model, we study its application on an extension of an existing economy model for the management of a cloud DBMS. A thorough experimental study shows that the prediction model ensures the economic viability of the cloud DBMS while enabling the offer of fast and cheap query services.

Cloud applications that offer data management services are emerging. Such clouds support caching of data in order to provide quality query services. The users can query the cloud data, paying the price for the infrastructure they use. Cloud management necessitates an economy that manages the service of multiple users in an efficient, but also, resource-economic way that allows for cloud profit. Naturally, the maximization of cloud profit given some guarantees for user satisfaction presumes an appropriate price-demand model that enables optimal pricing of query services. The model should be plausible in that it reflects the correlation of cache structures involved in the queries. Optimal pricing is achieved based on a dynamic pricing scheme that adapts to time changes. This paper proposes a novel price-demand model designed for a cloud cache and a dynamic pricing scheme for queries executed in the cloud cache. The pricing solution employs a novel method that estimates the correlations of the cache services in an time-efficient manner. The experimental study shows the efficiency of the solution.

This is a demonstration of data coordination in a peer data management system through the employment of distributed triggers. The latter express in a declarative manner individual security and consistency requirements of peers, that cannot be ensured by default in the P2P environment. Peers achieve to handle in a transparent way data changes that come from local and remote actions and events. The distributed triggers are implemented as an extension of the active functionality of a centralized commercial DBMS. The language and execution semantics of distributed triggers are integrated in the kernel of the DBMS such that the latter handles transparently and simultaneously both centralized and distributed triggers. Moreover, the management of distributed triggers is associated with a set of peer acquaintance and termination protocols which are incorporated in the centralized DBMS.

In a world of wide-scale information sharing, the decentralized coordination has to consolidate a variety of heterogeneity. Shared data are described in different formats, i.e. data structures, values and schemas. Querying manifold such sources entails techniques that can bridge the data formats. Some of these techniques deal with producing mappings for the schemas of data. The existing techniques view complementary aspects of the schema mapping problem. Important ones, consider producing all the possible mappings for a pair of schemas, insinuating any accompanying semantics in the mappings and adapting correct mappings as schemas evolve. Towards this end we have developed a solution that is fine-tuned for the discovery of mappings as schemas of autonomous sources are gradually revealed. In this demonstration we exhibit a new prototype tool that implements this solution. The tool provides a mechanism that realizes discovery of correct mappings as schemas are revealed. Mapping discovery is schema-centric and incorporates new semantics as they are unveiled. Mapping experience is reused and possible mappings are ranked so that the best choice is presented. The core mechanism collaborates with an automatic schema matching tool and the user that lightly guides the mapping process. The demonstration presents two application scenarios that prove the suitability of this prototype tool and the effectiveness of the implemented mapping solution in realistic situations of of data integration and exchange between heterogeneous autonomous sources.

As we move from a Web of data to a Web of services, enhancing the capabilities of the current Web search engines with effective and efficient techniques for Web services retrieval and selection becomes an important issue. Traditionally, the relevance of a Web service advertisement to a service request is determined by computing an overall score that aggregates individual matching scores among the various parameters in their descriptions. Two drawbacks characterize such approaches. First, there is no single matching criterion that is optimal for determining the similarity between parameters. Instead, there are numerous approaches ranging from using Information Retrieval similarity metrics up to semantic logic-based inference rules. Second, the reduction of individual scores to an overall similarity leads to significant information loss. Since there is no consensus on how to weight these scores, existing methods are typically pessimistic, adopting a worst-case scenario. As a consequence, several services, e.g., those having a single unrelated parameter, can be excluded from the result set, even though they are potentially good alternatives. In this work, we present a methodology that overcomes both deficiencies. Given a request, we introduce an objective measure that assigns a dominance score to each advertised Web service. This score takes into consideration all the available criteria for each parameter in the request. We investigate three distinct definitions of dominance score, and we devise efficient algorithms that retrieve the top-k most dominant Web services in each case. Extensive experimental evaluation on real requests and relevance sets, as well as on synthetically generated scenarios, demonstrates both the effectiveness of the proposed technique and the efficiency of the algorithms.

Cloud computing, the new trend for service infrastructures requires user multi-tenancy as well as minimal capital expenditure. In a cloud that services large amounts of data that are massively collected and queried, such as scientific data, users typically pay for query services. The cloud supports caching of data in order to provide quality query services. User payments cover query execution costs and maintenance of cloud infrastructure, and incur cloud profit. The challenge resides in providing efficient and resource-economic query services while maintaining a profitable cloud. In this work we propose an economic model for self-tuned cloud caching targeting the service of scientific data. The proposed economy is adapted to policies that encourage high-quality individual and overall query services but also brace the profit of the cloud. We propose a cost model that takes into account all possible query and infrastructure expenditure. The experimental study proves that the proposed solution is viable for a variety of workloads and data.

The peer-to-peer (P2P) paradigm has become very popular for storing and sharing information in a totally decentralized manner. At first, research focused on P2P systems that host 1D data. Nowadays, the need for P2P applications with multidimensional data has emerged, motivating research on P2P systems that manage such data. The majority of the proposed techniques are based either on the distribution of centralized indexes or on the reduction of multidimensional data to one dimension. Our goal is to create from scratch a technique that is inherently distributed and also maintains the multidimensionality of data. Our focus is on structured P2P systems that share spatial information. We present SpatialP2P, a totally decentralized indexing and searching framework that is suitable for spatial data. SpatialP2P supports P2P applications in which spatial information of various sizes can be dynamically inserted or deleted, and peers can join or leave. The proposed technique preserves well locality and directionality of space.

We consider an environment of numerous moving objects, equipped with location-sensing devices and capable of communicating with a central coordinator. In this setting, we investigate the problem of maintaining hot motion paths, i.e., routes frequently followed by multiple objects over the recent past. Motion paths approximate portions of objects' movement within a tolerance margin that depends on the uncertainty inherent in positional measurements. Discovery of hot motion paths is important to applications requiring classification/profiling based on monitored movement patterns, such as targeted advertising, resource allocation, etc. To achieve this goal, we delegate part of the path extraction process to objects, by assigning to them adaptive lightweight filters that dynamically suppress unnecessary location updates and, thus, help reducing the communication overhead. We demonstrate the benefits of our methods and their efficiency through extensive experiments on synthetic data sets.

We present a simulation environment that can be employed to study P2P mobile networks that are fast-evolving in both their topology and their content. This simulator implements a proposed P2P architecture based on Mobile Agent and Active Database technology and can be employed in order to build simulated mobile networks that are characterized by a diversity in peer needs, specifications and capabilities.

Handling spatial data in distributed environments is an intriguing issue. We consider autonomous sites of an overlay network that are bound to specific spatial information. We assume that each site has partial knowledge of the overlay. Actually, the sites are aware of and can communicate with some other sites, according to the spatial data to which they are bound to. We are interested in routing queries about spatial data in such overlays, solely by exploiting local knowledge on sites, i.e. based on locality and directionality in space. For such a system, we explore the parameters that can make search for any spatial information realizable and efficient. In this work we focus on the management of grid-partitioned space. In such a system there is a necessity for a mechanism that provides even knowledge of space to each site of the overlay. We investigate two different ways to define the directions of grid knowledge on each site and we propose a new distance metric. Moreover, we consider a new locality function that is specifically constructed in order to achieve even knowledge of space towards all directions. Furthermore, we present an experimental study that evaluates the theoretical propositions and verifies the theoretical results.

We describe a mechanism based on distributed Event-Condition-Action (ECA) rules that supports data coordination in a multidatabase setting. The proposed mechanism includes an ECA rule language and a rule execution engine that transforms rules when they are first posted, and then coordinates their execution. Like traditional ECA rules, our ECA rule language has three parts: an event language, a condition language, and an action language. The event language provides a set of operators with a formal semantics for a multidatabase environment, and which allows a wide variety of composite events. The condition language provides Boolean algebra operators that take as operands either composite or simple conditions. The action language provides a conjunction of simple or composite actions. The execution model partitions rules to more easily manageable forms, distributes them to relevant databases, monitors their execution and composes their evaluations. The mechanism has been designed in a manner that minimizes the number of messages that need to be exchanged over the network. We have also conducted an experimental evaluation to compare the implementation with a naïve centralized execution model. The paper also presents a prototype implementation as well as experimental results on its performance. This work is part of an on-going project intended to develop data coordination techniques for data sharing settings.

A challenging issue in fast-evolving pure P2P networks is the design of an appropriate mechanism for processing queries. Since both the data content of the peers as well as their acquaintances, change rapidly the typical P2P querying techniques become inappropriate. We are interested in P2P networks where peers are mobile and own a database. In this dynamic context the usage of a Mobile Agent framework appears very promising. The paper investigates the issues related to the above problem and proposes a P2P and Mobile Agent architecture based on Active Database technology. We argue that, the employment of ECA rules both for answering queries and deploying agents leads to an efficient as well as simple query processing technique. Furthermore, the proposed mobile agent system architecture offers a number of advantages due to the performance and scalability that can be achieved using Active Databases.