clustering

cheap server cluster

Cheap Server Cluster

Linux Server Clusters

In many branches of science complexity of the issues being studied require access to a supercomputer, with these powerful servers that can develop several billion operations per second. The traditional supercomputers used in parallel processing and contain arrangements of ultrafast microprocessors that work in synchrony to solve complex problems such as numerical forecasts of the weather, or to model complex structures of matter. Manufacturers like Cray supercomputers, IBM, Silicon Graphics, among others, models produced by special design and cost tens of millions of dollars - prices beyond the budgets of investment research groups. 

load balanced server cluster

In recent years, academics from various universities and research centers have been given the task of learning to build their own supercomputers (server cluster) by connecting personal computers and developing software to address such extraordinary problems. 

In 1994, joined the first cluster of PCs at the Goddard Space Flight Center of NASA, to solve computational problems that appear in Earth science and space. The pioneers of this project were Thomas Sterling, Donald Becker and other NASA scientists. The cluster of PCs had developed an efficiency of 70 megaflops (billion floating point operations per second). Researchers at NASA gave him the name of Beowulf in this cluster, in honor of the hero of medieval legends, who defeated the giant monster Grendel. 

In 1996, there were also two other successors of the Beowulf project of NASA. One of them is the project Hyglac developed by researchers at the California Institute of Technology (Caltech) and the Jet Propulsion Laboratory (JPL), and the other, the project Loki built in the Los Alamos National Laboratory in New Mexico. Each cluster is integrated with 16 Intel Pentium Pro microprocessors and had a sustained performance of more than a gigaflop at a cost less than $ 50,000 U.S. dollars.

Linux Clusters

In 1996, at the Oak Ridge National Laboratory in Tennessee, faced with the problem of drawing up a map of the environmental conditions of the territory of the United States. The territory was divided into 7.8 million cells 1Km. Each cell contained the information of 25 variables, from the monthly average rainfall until the content of nitrogen in the soil. Any PC or work station could with this task. Required a supercomputer for parallel processing. At present, with a cluster of 130 PCs to work on the development of eco regions of the map.

The parallel computing in high performance is achieved by dividing a huge and complex task among multiple processors. During the Second World War, before the invention of the electronic computer, used a similar technique to perform large calculations associated with the design of the atomic bomb for the Manhattan Project. To dramatically reduce the time for solving a math problem great, every part of the problem was solved by a different person. To these people were called computers. At present electronic computers can work in harmony to solve scientific problems that were not dreamed resolve more than a decade ago 

     * Parallel computing using Linux Clusters 
       Review (1997) processing in parallel with cheap Linux clusters. 

     The free availability, high reliability and efficiency of Linux, has been a great fortune for computer science since the mid-1990s and their popularity is growing every day in academic circles. Scientists used to turn PCs into Linux work stations UNIX effective for the realization of a large number of tasks, including numerical computation of complex models. The development of type Beowulf clusters of computers, has extended the utility of Linux to the field of high performance parallel computing. Additionally, the nature of open-source Linux system has allowed developers to add additional features directly to the operating system to meet the needs of server clusters.

Server Cluster

OBJECTIVE: 

To train students in the installation and administration of Cluster high availability of data and services as well as the configuration of the load balancing between servers, which complement and satisfy the needs of IT infrastructure in your company, benefiting from the advantages of free software. 

The courses are theoretical and practical, given the way attendance. The aim is that participants are able to create productive work environments through configuration and integration of different services specified in this agenda. 

OUTCOME: 

     * Using free software tools available for creating environments clusters 
     * Knowing the architecture of virtual servers in Linux and its application 
     * Apply techniques of load balancing, for better performance in applications 
     * Manage applications on Linux cluster 
     * Finding the most effective way to resolve problems in clustered environments 

VALUE ADDED SERVICE: 

     * Exercises and laboratories. 
     * Registration of our courses before Secretary of Labor. 
     * 30 days of technical support at no additional cost. 
     * 1 computer per student. 
     * Evaluation at the beginning and end of the course. 
     * Office equipment at their disposal. 
     * Service of coffee and biscuits. 

DURATION: 

     * 30 hours 
     * Offered Monday through Friday and weekends 

DELIVERABLES: 

     * Printed manual installation of Linux servers for All Clusters 
     * 1 DVD CentOS 
     * Diploma and a record of participation. 

AGENDA 
ADMINISTRATION OF SERVERS 

     * Operating System 
     * File Systems 
     * Model TCP / IP 
     E * MAC IP 
     * DNS 
     * TCP and UDP 
     * Levels of Execution 

INTRODUCTION TO CLUSTER 

     * Benefits 
     * Types of Cluster 
     * Availability and Redundancy 

LINUX CLUSTER 

     * Components and Infrastructure 
     * High availability 
     * LVS, GFS 
     * DM Multipath 
     * CLVM, GNBD 
     * Management Tool 

BY DNS load balancing 

     * DNS, and domain areas 
     * Hierarchies, TTL, rsync 
     * Setting up 3 layers 
     * Transactional database 
     * Distributed File Systems 
     * Backup Virtual Server 
     * Nestled Load Balancer 

LINUX VIRTUAL SERVER 

     * Linux Virtual Server 
     * Components LVS 
     * Setting up LVS 
     * Piranha 

ALLOCATION OF PETITIONS 

     * IPVS 
     * Features and Configuration 
     * Algorithms for Assigned Petitions 
     * Virtual Services 
     * Firewall Marks 
     * Persistence 
     * Iptables 

PROGRESS ON APPLICATIONS 

     * Methods Progress Petitions 
     * Features and Configuration 
     * NAT and ARP 
     * Direct Routing 
     * IP Tunneling 
     * Ipvsadm 
     * Disabling Servers 

MONITORING LVS 

     * Ipvsadm 
     * Process File System 
     * SNMP and MIBs 
     * SNMP and LVS 

Topology LVS 

     * Setting 2 customers 
     * Multiple customers GNBD 
     * Components GNBD 
     * Gnbd_serv, gnbd_export, gnbd.ko, gnbd_import 
     * GNBD and Multipath DM, and GFS GNBD 
     * Export and Import GNBD 

SET OF CLUSTER 

     * RedHat Cluster Manager 
     * Hardware configuration 
     * Prerequisites Hardware 
     * Set Up Cluster HW 
     * Cluster Configuration Tool 
     * Management Failover 
     * Resources and Services 
     * Conga 

ADMINISTRATION OF CLUSTER 

     * Cluster Status Tool 
     * Cluster Configuration Tool 
     * Upper and Lower Cluster 
     * Backup and Recovery 
     * Operation of the Cluster with Conga 
     * Management of nodes with Conga 
     * Monitoring Cluster 
     * Problem Solving 

STORAGE DEVICE 

     * Arrangements disk 
     * Virtualization 
     * RAID 
     * Hot Swap 
     * Hot Spare 
     * Information Management 
     * Storage Technologies 
     * DAS, NAS, SAN 

GNBD 

     * Servers and clients GNBD 
     * Block Devices 
     * Multiple customers GNBD 
     * Gnbd_serv, gnbd_export, gnbd_import, gnbd.ko 
     * GNBD Driver 
     * Export and Import GNBD 

LVM 

     * Architecture, features and benefits 
     * CLVM 
     * Creating an LVM 
     * Commands PV, VG and LV 
     * Support of Metadata 
     * Types of mapping LVM 
     * Diagnosis and Recovery 

GFS 

     * Architecture, features, benefits, requirements 
     * Cluster Volume Manager 
     * Lock Management 
     * Cluster Configuration Management 
     * Setting up and mounted a FS 
     * Scalability of a FS 
     * Suspension of activity in the FS 
     * Direct I / O 
     * Slow Start and the Cluster GFS 
     * Resolution of errors 

DM multipath 

     * Types of Configurations 
     * Handling Devices 
     * Command multipath 
     * Command dmsetup 
     * Console multipathd 

ADMINISTRATION OF APPLICATIONS 

     * High Availability Applications 
     * Stateless Components 
     * Global Objects 
     * Data in session 
     * Global File System 

MEETING OF QUESTIONS AND ANSWERS 
Assessment Review

Albert Sirvent - DomainGurus