Sudip Roy

S

Data Management Challenges in Production Machine Learning

SIGMOD

Jan 2017

A

TFX: A TensorFlow-Based Production-Scale Machine Learning Platform

ACM SIGKDD

Jan 2017

G

Goods: Organizing Google's Datasets

Jun 2016

Enterprises increasingly rely on structured datasets to run their businesses. These datasets take a variety of forms, such as structured files, databases, spreadsheets, or even services that provide access to the data. The datasets often reside in different storage systems, may vary in their formats, may change every day. In this paper, we present GOODS, a project to rethink how we organize structured datasets at scale, in a setting where teams use diverse and often idiosyncratic ways to produce the datasets and where there is no centralized system for storing and querying them. GOODS extracts metadata ranging from salient information about each dataset (owners, timestamps, schema) to relationships among datasets, such as similarity and pro

B

Managing Google’s data lake: an overview of the GOODS system

Bulletin of the IEEE Computer Society Technical Committee on Data Engineering

Jan 2016

I

PerfAugur: Robust diagnostics for performance anomalies in cloud services

IEEE st International Conference on Data Engineering Seoul

Jan 2015

Cloud platforms involve multiple independently developed components, often executing on diverse hardware configurations and across multiple data centers. This complexity makes tracking various key performance indicators (KPIs) and manual diagnosing of anomalies in system behavior both difficult and expensive. In this paper, we describe PerfAugur, an automated system for mining service logs to identify anomalies and help formulate data-driven hypotheses. PerfAugur includes a suite of efficient mining algorithms for detecting significant anomalies in system behavior, along with potential explanations for such anomalies, without the need for an explicit supervision signal. We perform extensive experimental evaluation using both synthetic and r

S

The Homeostasis Protocol: Avoiding Transaction Coordination Through Program Analysis

SIGMOD

Jan 2015

Datastores today rely on distribution and replication to achieve improved performance and fault-tolerance. But correctness of many applications depends on strong consistency properties--something that can impose substantial overheads, since it requires coordinating the behavior of multiple nodes. This paper describes a new approach to achieving strong consistency in distributed systems while minimizing communication between nodes. The key insight is to allow the state of the system to be inconsistent during execution, as long as this inconsistency is bounded and does not affect transaction correctness. In contrast to previous work, our approach uses program analysis to extract semantic information about permissible levels of inconsistency a

S

Coordination Through Querying in Youtopia System

SIGMOD

Feb 2011

In a previous paper, we laid out the vision of declarative, datadriven coordination (D3C) where users are provided with novel abstractions that enable them to communicate and coordinate through declarative specifications [3]. In this demo, we will show Youtopia, a novel database system which is our first attempt at implementing the vision. Youtopia provides coordination abstractions with within the DBMS where users can register coordination queries that coordinate through constraints with other coordination queries that are entering the system. Such “entangled” queries are evaluated together in the DBMS in such a way that the joint execution causes all their constraints to be satisfied—the queries coordinated their answers as specified in t

S

Entangled queries: enabling declarative data-driven coordination

SIGMOD

Feb 2011

Many data-driven social and Web applications involve collaboration and coordination. The vision of declarative data-driven coordination (D3C), proposed in [9], is to support coordination in the spirit of data management research â€“ to make coordination datacentric and to specify coordination scenarios through convenient declarative languages. The present paper proposes the notion of entangled queries, a coordination language that extends SQL by constraints that allow for the coordinated choice of result tuples across queries originating from dierent users or applications. A substantial challenge arises here: It is nontrivial to define a declarative coordination formalism without immediately arriving at the general (NP-complete) Constraint

S

Beyond Isolation: Research Opportunities in Declarative Data-Driven Coordination

SIGMOD Record

Sep 2010

There are many database applications that require users to coordinate and communicate. Friends want to coordinate travel plans, students want to jointly enroll in the same set of courses, and busy professionals want to coordinate their schedules. These tasks are difficult to program using existing abstractions provided by database systems because in addition to the traditional ACID properties provided by the system they all require some type of coordination between users. This is fundamentally incompatible with isolation in the classical ACID properties. In this position paper, we argue that it is time for the database community to look beyond isolation towards principled and elegant abstractions that allow for communication and coordinatio

P

A layout-aware physical design method for constructing feasible QCA circuits

Proceedings of the th ACM Great Lakes symposium on VLSI

Jan 2008

Quantum-dot Cellular Automata (QCA) is an emerging computing paradigm, in which logical operations as well as signal transmission occurs due to Coulombic charge interaction between neighbouring QCA cells, moderated by a 4-phase QCA clock potential. Thermodynamic constraints like the number of QCA cells in a clocking zone must be obeyed to obtain a logically correct and feasible QCA circuit. These constraints depend on various design factors like total wirelength in a circuit, height of a clocking zone etc. which are not available until actual circuit layout is obtained. In this paper, the various design automation problems assosciated with obtaining a feasible QCA layout are addressed. The layout generation problem is formulated as embeddin

W

Writes that Fall in the Forest and Make no Sound: Semantics-Based Adaptive Data Consistency

Datastores today rely on distribution and replication to achieve improved performance and fault-tolerance. But the correctness of many applications depends on strong consistency properties -- something that can impose substantial overheads, since it requires coordinating the behavior of multiple nodes. This paper describes a new approach to achieving strong consistency in distributed systems while minimizing communication between nodes. The key insight is to allow the state of the system to be inconsistent during execution, as long as this inconsistency is bounded and does not affect transaction correctness. In contrast to previous work, our approach uses program analysis to extract semantic information about permissible levels of inconsist

V

Entangled Transactions

VLDB

As the world becomes more interdependent and computing grows more collaborative, there is a need for new abstractions and tools to help users work together. We recently introduced entangled queries – a mechanism for information exchange between database queries [6]. In this paper, we introduce entangled transactions, units of work similar to traditional transactions that however do not run in isolation, but communicate with each other via entangled queries. Supporting entangled transactions brings about many new challenges, from an abstract model to an investigation of the unique systems issues that arise during their implementation. We ﬁrst introduce a novel semantic model for entangled transactions that comes with analogues of the classic

C

Quantum Databases

CIDR

We introduce quantum databases, a new database abstraction that allows to defer the making of choices in transactions until an ap- plication or user forces the choices by observation. Conceptually, a transaction is in a quantum state – in one of many possible worlds, which one is unknown – until fixed by observation. Practically, our abstraction enables late binding of values read from the database. This allows more transactions to succeed in environments with high contention. This is particularly important for applications in which transactions compete for scarce physical resources represented by data items in the database, such as seats in airline reservation sys- tems or meeting slots in calendaring systems. In such environ- ments, defer

About

Detail

Timeline

Résumé