The virtual love of Hadoop: Coping with Big Data challenges

The increasing volume of data and the increasing competitive pressures have allowed more and more enterprises to start thinking about how to tap the value of these data. Traditional BI systems, http://www.aliyun.com/zixun/aggregation/8302.html > Data warehouses and database systems do not handle this data well. Reasons include:

1. The data volume is too large, the traditional database can not effectively store and maintain acceptable performance;

2. The newly generated data are often unstructured, and traditional approaches are designed to deal with structured data;

3. Traditionally, the hardware required for data processing is often relatively expensive, and many businesses cannot afford to continue to deal with the cost of traditional methods as the volume of data increases. To this end, much of the internet industry-respected Apache Hadoop is increasingly attracting the attention of the business community, a large number of enterprises are thinking about how to put hadoop this beautiful bride back to their data center.

However, the traditional enterprise data center to marry this coquettish bride is not so easy. The deployment and operation of Hadoop require a lot of geeks to be fully controlled, beyond the technical capabilities of traditional enterprise data centers; In addition, Hadoop requires not only specialized hardware but also security and service levels to be challenged. How to enjoy the sweet dreams of a beautiful bride without any other consequences is a real challenge for companies to choose Hadoop.

From server virtualization to the entire data center virtualization, today we have fully felt the power of virtualization this kid! If virtualization can be in love with Hadoop, will the Enterprise data center choose Hadoop away? The answer is yes. Virtualization enables the separation of Hadoop and the underlying physical hardware, truly into the cloud ballerina, and Hadoop makes it easy to move into a cloud of fast-moving, highly available, resource-resilient scheduling, and secure multi-tenant, and the dream of large data analysis and utilization in enterprise data centers can truly become a reality.

Let's uncover the love cheats for virtualization so that we can better use Hadoop to meet the challenges of big data.

1. Rapid deployment of Hadoop: We are already familiar with virtualized passwords, including virtual machines, snapshots, templates, resource dynamics, and so on, which are good at overcoming the challenges of a large number of application deployments, and Hadoop, of course, can dramatically increase the deployment speed of Hadoop nodes. At the same time, you can quickly start and shut down the Hadoop node on demand, enabling efficient use of resources, such as VMware's Serengeti Open source project, to help push the virtualization and Hadoop love process;

2. Provide high availability and fault tolerance for Hadoop: While Hadoop improves system reliability through data distribution replication, there are still a number of components that have single points of failure, which may not be a problem in an Internet enterprise, but are definitely a challenge for traditional data centers. For example: Namenode and Jobtracker, as well as some support modules have a single point of failure, through the virtual kid's platform is highly available for these modules can easily give high reliability features, let Hadoop into the Enterprise data center, you can still sit back and relax;

3. The efficient data center embracing Hadoop: Through the virtual kid dynamic scheduling capability, you can mix various loads across the enterprise data Center cloud Platform, and Hadoop can, of course, sleep with the rest of the load, ensuring that no conflicts occur through strict security isolation. Even you can run different versions of Hadoop on the same cloud platform, coexist peacefully, share resources, reduce the overall cost of traditional Hadoop deployment, and easily achieve the goal of efficient data centers, while ensuring availability and performance.

4. Greatly enhance the utilization of Hadoop environment resources: the Hadoop and other loads deployed on the same host, through the resource control strategy to achieve efficient resource allocation and scheduling, the realization of Hadoop in the cloud of the perfect stroll, is a virtual boy to win the love of the key link;

5.Hadoop Cloud Multi-Tenant: With virtualization isolation capabilities, Hadoop ensures a perfect experience for its multi-tenant tenants, and different tenants can mix Hadoop and other loads in a cloud resource pool, with multiple tenants successfully deployed;

6. Security isolation: Virtual kid's security isolation ability, so that different organizations, users of Hadoop can run without worry, easily achieve the data and environment completely isolated target, while sharing the underlying physical resources;

7. Ease of maintenance and migration: Virtualization makes Hadoop nodes easy to replicate, migrate, and instantly realize the cloud migration between different clusters in the data center, one data center to another, and Hadoop is no longer an inconvenient mother.

Virtual Kid wins Hadoop with 7 axe, not only does Hadoop not mess with traditional enterprise data centers, but Hadoop's charm on the virtual platform has not been reduced because a lot of facts have proven that virtualized Hadoop nodes are still performing as well as physical environments, It also brings a lot of cost savings. Hadoop and virtualization are equal, their love is worthy of our common hope and wish: I wish Hadoop and virtualization, forever knot concentric, eternal!