The Next Generation Data Center ה-3 לדצמבר 2007, בית חיל האוויר, הרצליה 7:30 - 18:00 להרשמה
The IGT 2007 annual conference deals with the challenges of the Grid infrastructure virtualization in the next generation data center: Service Oriented Infrastructure, that enables Capacity on Demand for SLA based applications.
Steve Rubinow, CTO, NYSE Euronext CTO challenges with regards to Scalability & Latency
Chris Swan, Director - Credit Suisse Trusted virtual appliances This presentation will begin by examining the virtual appliance deployment model, why it is relevant to 'grid' and what the potential security issues are. It will then look at how a chain of trust might be established from the underlying hardware platform to a virtual appliance. Finally, the presentation will describe some of the issues of trustworthiness in the process of creating virtual appliances.
Chris works in the Credit Suisse IT Research and Development group looking for innovative solutions to financial services problems. His primary areas of coverage are networks and security, having previously looked after infrastructure software (including virtualisation, performance management, provisioning and configuration management). Before joining the R&D group he headed engineering efforts around enterprise grid computing, SOA and J2EE infrastructure. He holds a BEng in Electronics Engineering from the University of York, and an MBA from OUBS.
Prof. Yossi Matias, Head of Google's R&D Center in Tel Aviv, Google Towards a world of Cloud Computing
Meir Shor, VP Technologies, Bank Leumi
Nati Shalom, Founder & CTO, GigaSpaces GigaSpaces’ Enterprise Data Grid and Microsoft’s Excel server The new architecture for high-speed processing, transforms Excel into a fully scalable, enterprise-ready application running on powerful grid servers with full version and auditing controls. Intensive computing is moved off the user's machine, processing speeds increase sharply, and a firm can maintain consistency among users. One bank that implemented GigaSpaces' in-memory data grid (IMDG) with Excel Services saw a sixfold performance improvement.
Michael Kagan, VP of Architecture, Mellanox Converged IO Solutions for the Heterogeneous DataCenter Convergence and heterogeneity is an oxymoron. Yet, this is a reality in the data center. IO convergence on the server side is becoming a critical need, especially with blade server and green initiatives where space, power and cost savings are critical. But, when it comes to fabrics in the Server-to-Server (IPC), LAN and SAN infrastructure, heterogeneity is the norm. Fibre Channel SAN infrastructures are here to stay – there is too much investment by OEMs and end users, and significant reliability, software and manageability advantages. iSCSI SANs hold the promise of converging the LAN and SAN fabrics to Ethernet only, but is limited to smaller scale deployments only. InfiniBand SAN is the new kid in the block that is seeing traction in server-to-server low latency, clustering and capacity driven data center deployments. Emerging technologies like Fibre Channel over Ethernet (FCoE) promises to further the convergence and heterogeneity story in the data center by providing a solution where the IO adapter in the server can be converged to Ethernet-based only (as in iSCSI), but Fibre Channel SAN connectivity is maintained. However, since Ethernet by nature is not reliable, enhancements to Ethernet are needed (as defined in the Converged Enhanced Ethernet or CEE related industry initiatives) to carry Fibre Channel SAN traffic. This will result in yet another form of fabric in the infrastructure – the CEE fabric, aside from existing Ethernet or classical Ethernet.
This session explores different SANLANIPC solution options, challenges to IO convergence in the server and proposes solutions that address the datacenter challenges in an incremental way, keeping legacy connectivity in mind, while providing for an upgrade path where connectivity with all possible fabrics in the LAN and SAN infrastructures can be achieved. The goal is to provide end users flexibility with their deployment options and provide room to adopt newer technologies.
Yaron Haviv, CTO, Voltaire Managing Network and Storage Infrastructure as a Grid Resource The expansion of grid technologies brings benefits such as greater capacity, lower costs and the support of more diverse applications. This trend has also brought additional infrastructure requirements and associated challenges. Next generation grid architectures will address those complexities by treating infrastructure as a managed resource that can be centrally scheduled and provisioned to provide optimal configuration and resources to meet application requirements. Servers, network topologies and storage elements will be created and torn down to meet application requirements much like compute jobs and processes are scheduled today. This session will examine how grids can be built around a service oriented infrastructure (SOI) and by leveraging unified fabrics like InfiniBand, network, cluster communication, and storage can be dynamically optimized to meet application requirements. This architecture will be integrated with schedulers and resource management to maximize the value and enable rapid infrastructure deployment, treating hardware as a managed resource pool.
Victoria Livschitz, Founder and CEO, Grid Dynamics Bridging The Paradigms: Convergence of Compute Grids with In-Memory Data Grids As compute grids are becoming more wide spread in commercial data centers, the bottlenecks in application performance move from raw processing to searching, storing and retrieving the data. In-Memory Data Grid (IMDG) technology solve this fundamental problem by acting as super-efficient application accelerator, taking advantage of unused resources readily available on the grid – disk, memory, IO – to put the data in memory of the same computer that performs the calculations.The talk will explore how IMDG can be easily integrated with existing enterprise grids to create data-aware grid applications and provide application performance acceleration while improving application scalability and reliability.
Assaf Marron, PhD, Corporate Architect, BMC Software Service Management and Automation in the New Data Center As datacenters become increasingly dynamic and virtualized, IT management becomes a growing challenge. In this talk, Assaf will discuss the new management challenges faced by IT professionals - from CIOs to system administrators, and present new solutions and future trends for full-life-cycle data center management and service automation.
Assaf Marron is a Corporate Architect in the CTO Office of BMC Software. In this role he is responsible for various technology projects company wide, among them virtualization and grid computing. Assaf has over 25 years of technical and management experience in software development and in implementation of IT, working both at leading software companies and in large enterprises in the United States and Israel. He is the inventor or co-inventor of several patents, and holds a PhD degree in Computer Science.
Ofir Zamir, Senior Systems Engineer, VMware Transforming the Data Center - Automation of the Virtual Data Center Virtualization technology is changing the way that people do computing today - it transforms the Data Center. The technology is based on 4 key features: Partitioning, Isolation, Encapsulation, and Hardware independence. All those key features enable VMware to deliver full-automated Data Center. It is based on Virtual infrastructure which is dynamic pool of resources that is available to run the virtual machines. By using the virtualization technology the customers get adaptive data center that gives the best resources available to each application.
Edward Aronovich, Tel Aviv University Grid infrastructure - How can it work for you?
We all expect "The Grid" to do a set of actions and have a certain behavior. Some of them can be well defined services and some are general characteristics. Grid is implemented as set of services and features called middleware. In this talk we will present the building block of the grid and the features it can provide. We will refer to the architect ion, security, usability and application profile and adjustments needed in order to achieve the grid. The talk will survey the latest grid architecture and concepts.
Philippe Trautmann, HPC BD Manager, SUN
Philip Olenick, Dell
DELL – Enabling the Cloud
Dr. Gilad Zlotkin, President & CTO, Xeround Systems
Xeround real-time data management solutions demonstrate massive scalability on a 320-core grid In this session we will present recent benchmark results of Xeround innovative real-time data management solutions for large-scale, high performance, next-generation services, demonstrating unmatched real-time performance and online scale-out abilities with the Nokia Database Benchmark (NDB). Designed and published by Nokia, the NDB benchmark simulates real-time transaction processing and other data operations that are typical in large-scale, write-intensive and mission critical service provider environments. The benchmark was run on off-the-shelf industry-standard and commodity servers interconnected by commodity 1G Ethernet switches, running Linux.
Ronen Yochpaz, CTO, Venotion Context driven discovery and classification of services (Business Governance)
The first phase of any Grid project implementation endorse a stage of mapping the current existing IT software components (services) and classify them in order to allow search, reuse and scale capabilities over the grid network. Most of the ways today to enable this option is based on manual work of service by service cataloguing in a sequential and opportunistic manner. More over, most of the classification of services is done without any ordered or context methodology.Those two work processes causing a delay in the implementation of the grid and slowing its adoption and use. In the lecture we will introduce an automatic discovery and classification tools allowing a fast service discovery, automated classification tool and method, engaging a new work process for service search and reuse giving a context classification for each service. Those abilities allow faster Grid implementation and new dimension for management by business context.
Yaron Moradi, VP ALP Business, Crescendo Networks
Application Performance in Next Generation Data Centers
The migration to next generation data centers substantially increases the dependencies between the data center applications and the network. Convergence between network and IT applications can prevent the negative effects of poor application performance on the organization.
The complexity of multi-tiered architectures and the potential risks involved with application bottlenecks within data centers, place additional burdens on IT managers in their challenge to control mission critical applications.
Each of the many tiers in the data center (web, application servers, databases, legacy systems, etc) may tackle a number of performance bottlenecks – from network level problems to application bottlenecks - all of which may unpredictably effect end-user response times of transactions and business processes.
This session will discuss the consequences of datacenter application performance bottlenecks on the organization and the focus that needs to be placed on how applications and network interact to deliver constant and optimal end-user response time per transaction and business process in the organizations.
Ronen Hamias, CTO, PrimaGrid
Grid and SOA: the road to optimized Business Services
Service Oriented Architecture (SOA) offers the most effective way to overcome IT complexities involved in businesses that require IT systems to be flexible and adaptive enough to address the changing business needs.
However, Business Services must be available for consumption and feature optimized performance, in order to avoid customers' dissatisfaction.
Grid technologies offer the ability to best optimize your Business Services layer. This session explores the road to optimized SOA Business Services, which feature fast and reliable service consumption through advanced grid infrastructure. We will discuss best practices in solution implementations, ranging from the architecture level to technology standards.
RanGilad-bachrach, Intel
Workstation Capacity Tuning using Reinforcement Learning
Computer grids are complex, heterogeneous, and dynamic systems, whose behavior is governed by hundreds of manually tuned parameters. As the
complexity of these systems grows, automating the procedure of parameter tuning becomes indispensable. In this talk, we present the problem of
auto-tuning server capacity, i.e. the number of jobs a server runs in parallel. We present three different reinforcement learning algorithms,
which generate a dynamic policy by changing the number of concurrent running jobs according to the job types and machine state. These
algorithms were tested on synthetic environments as well as on a production environment and were proven to be very efficient, especially
when multi-core servers are used.
Dr. Avi Mendelson – Computer Architect Intel Israel
New HW and SW challenges in modern computer architectures
In order to keep the traditional performance growth of computer systems, mainly two new trends are being developed; the first trend, termed many-core architecture calls to trade single thread performance with multithreaded performance while the second trend, termed multi-core architecture calls to increase both the single thread performance and the multithreaded performance and so to divide the "transistor budget" of the processor between relatively small number of "large cores".
In this talk I will discuss the root cause of the need to change the traditional computer architectures, what are the SW and HW implications of each of these new trends and extend the discussion to what is needed for each of them in order to win the market. I will conclude the talk with a discussion other alternatives such as the use of special cores.
Ilona Kifer, TAU
Protein Structure Prediction becomes feasible with parallel computing.
The task of Protein Structure Prediction is considered the 'Holy Grail' of structural biology. It is necessary for understanding the connection
between the genetic code encoded in the DNA and the 3-dimensional structure of the operating units of our body - the proteins. This in
turn is crucial for the purpose of designing drugs that can correct for mis-structured proteins, and thus allowing to find a cure for diseases.
The current methods are very expensive and time consuming and are not feasible for large scale in practice. Efficient computational methods
for predicting the structure are very challenging and researched by many leading scientific groups around the world.
Considering the speed of existing methods and the amount of available data this task is expected to take 17 years. Using the condor framework,
we have managed to perform this task in 3 months on university computers.
