A5060139136- Cloud Computing and Resource Management
- Distributed systems and fault tolerance
- Radiation Effects in Electronics
- Distributed and Parallel Computing Systems
- Software System Performance and Reliability
- Parallel Computing and Optimization Techniques
- Software Reliability and Analysis Research
- Advanced Data Storage Technologies
- IoT Networks and Protocols
- Scientific Computing and Data Management
- Logic, programming, and type systems
- VLSI and Analog Circuit Testing
- Energy Harvesting in Wireless Networks
- Software Testing and Debugging Techniques
- IoT and Edge/Fog Computing
- Software-Defined Networks and 5G
- Security and Verification in Computing
- Real-Time Systems Scheduling
- Advanced Malware Detection Techniques
- Interconnection Networks and Systems
- Caching and Content Delivery
University of California, Berkeley
2016-2024
Berkeley College
2016-2021
Universidade Nova de Lisboa
2017
Critical Software (Portugal)
1999-2005
University of Coimbra
1995-2002
Serverless cloud computing handles virtually all the system administration operations needed to make it easier for programmers use cloud. It provides an interface that greatly simplifies programming, and represents evolution parallels transition from assembly language high-level programming languages. This paper gives a quick history of computing, including accounting predictions 2009 Berkeley View Cloud Computing paper, explains motivation serverless describes applications stretch current...
Traditional datacenters are designed as a collection of servers, each which tightly couples the resources required for computing tasks. Recent industry trends suggest paradigm shift to disaggregated datacenter (DDC) architecture containing pool resources, built standalone resource blade and interconnected using network fabric.A key enabling (or blocking) factor disaggregation will be - support good application-level performance it becomes critical that fabric provide low latency...
The evolution that serverless computing represents, the economic forces shape it, why it could fail, and how might fulfill its potential.
Machine learning (ML) workflows are extremely complex. The typical workflow consists of distinct stages user interaction, such as preprocessing, training, and tuning, that repeatedly executed by users but have heterogeneous computational requirements. This complexity makes it challenging for ML to correctly provision manage resources and, in practice, constitutes a significant burden frequently causes over-provisioning impairs productivity. Serverless computing is compelling model address...
An important step in the development of dependable systems is validation their fault tolerance properties. Fault injection has been widely used for this purpose, however with rapid increase processor complexity, traditional techniques are also increasingly more difficult to apply. This paper presents a new software-implemented and monitoring environment, called Xception, which targeted at modern complex processors. Xception uses advanced debugging performance features existing most...
Serverless computing allows developers to deploy and scale stateless functions in ephemeral workers easily. As a result, serverless has been widely used for many applications, such as computer vision, video processing, HTML generation. However, we find that the nature of wastes important benefits modern language runtimes have offer. A notable example is extensive profiling Just-in-Time (JIT) compilation effort implement achieve acceptable performance popular high-level languages, Java,...
Computer-based systems are expected to be more and dependable. For that, they have operate correctly even in the presence of faults, this fault tolerance theirs must thoroughly tested by injection faults both real artificial. Users should start request reports from manufacturers on outcomes such experiments, mechanisms built into handle faults. To inject artificial physical offers a reasonably mature option today, with Swift tools being preferred for most applications because their...
The serverless computing model leverages high-level languages, such as JavaScript and Java, to raise the level of abstraction for cloud programming. However, today's design platforms based on stateless short-lived functions leads missed opportunities modern runtimes optimize through techniques JIT compilation code profiling.
In the research reported in this paper, transient faults were injected nodes and communication subsystem (by using software fault injection) of a commercial parallel machine running several real applications. The results showed that significant percentage caused system to produce wrong while application seemed terminate normally, thus demonstrating tolerance techniques are required systems, not only assure long-running applications can but also (and more important) produced correct. Of...
No abstract available.
This paper addresses the problem of injection faults in communication system disjoint memory parallel computers and presents fault results showing that 5% to 30% injected subsystem a commercial computer caused undetected errors lead application generate erroneous results. All these cases correspond situations which it would be virtually impossible detect benchmark output was erroneous, as size file plausible no had been detected. emphasizes need for tolerant techniques systems order achieve...
We introduce the design of a parallel library for MPI based on Linda programming paradigm, called Eilean. It provides scalable distribution tuple space through hierarchical (or cluster) partitioning scheme, and type specific access/distribution policies. Portability is achieved using message passing standard as underlying communication system. The static, yet general partition available processors. With such structure, run-time system, aided by programmer hints, can map tuples closely to...
Using off-the-shelf hardware and software in factory computer systems is one attractive way of increasing flexibility reducing costs development, maintenance training. However, targeting critical applications for the low-end PC market general-purpose operating such as Microsoft Windows 95/NT poses some problems. These are usually required to run 24 hours a day demand high availability data integrity. without any fault tolerance support unreliable dangerous. This article presents framework,...
The extraction of electrical energy from solar radiation using photovoltaic phenomenon depends not only on the quality modules but also level matching to load. This is an optimization process which needs dynamic, precise and embracing sensing over system. paper proposes design a smart system following cyber-physical architectural approach uses distributed computational strategy between Cloud IoT node. It shown that latter be designed as software defined hardware (SDH) better reflects need...