Skip to main content

Cloud provisioning with Apache Zookeeper


A few months back we explored how Java apps could be provisioned on the cloud using YARN. Based on the interest that post generated, let us re-visit the topic and explore provisioning with a variant architecture leveraging MRv1, HDFS and Zookeeper.

In a paper by S.Krishnan, J.C. Counio, the authors have described an approach used at Yahoo to create an elastic web server cloud farm based on Hadoop.  The authors describe 2 primary use cases for the system which they name as ‘Pepper’:
(1)   Web Feed processing – Dynamic provisioning of web apps to process high frequency small size feeds like news, sports, finance.
(2)   Online Clustering – Extract features and assign posts to clusters determined earlier by offline clustering.

In the implemented design,
- Zookeeper is leveraged to act as central registry where the information of the tasks that run the web applications is maintained.
- A Job Manager module exposes an API to deploy web apps (WAR file) on a dedicated location in HDFS.
- A Map Web Engine runs as a map task on Hadoop nodes and starts, monitors, stop the web app on the web server installed on the node. The web server could be an embedded Jetty server or other compatible server. Each web server registers itself with Zookeeper and generates its individual log4j logs which are aggregated on a common location in the cluster.
- A Proxy router acts as a single point of entry for client requests and forwards request to one of the available web servers.


What stands out in this implementation is how Apache Zookeeper has been leveraged for cloud provisioning. We understand that Apache ZooKeeper aids in cross-node synchronization. It “acts as a centralized service for maintaining configuration information, naming, providing distributed synchronization, and providing group services”.

The key operations performed by Zookeeper in this design, include the following, but are not limited to:
- Act as a monitor for the web servers
- Maintain a list of available web servers
- Maintain information of jobs that run the web applications
- Associate each web application with a free port
- Register the web applications with the central registry
- End the session initiated by the web application if it turns faulty

Some of the key characteristics of Zookeeper come to fore here making it a good choice:
- First-In-First-Out (FIFO) order queue – utilizing a pipeline architecture for storing a list of web applications that requires to be processed
 - Caching - priority web application can be cached to improve performance
- Locking - for maintaining consistency to update the central registry sequentially

The big takeaway from Pepper’s design is that Zookeeper can be used as much more than a coordination service among distributed systems. Rather than relying on a database or a complicated state machine, the Zookeeper here plays a pivotal role in simplifying cloud provisioning.

Comments

Popular posts from this blog

Low latency SQL querying on HBase

HBase has emerged as one of the most popular NoSQL database offering distributed, versioned, non-relational tables hosted on commodity hardware. However, with a large set of users coming from a relational SQL world, it made sense to bring the SQL back in this NoSQL. With Apache Phoenix, database professionals get a convenient way to query HBase through SQL in a fast and efficient manner. Continuing our discussion with James Taylor, the founder of Apache Phoenix, we focus on the functional aspects of Phoenix in this second part of interaction.
Although Apache Phoenix started off with distinct low latency advantage, have the other options like Hive/Impala (integrated with HBase) caught up in terms of performance?
No, these other tools such as Hive and Impala have not invested in improving performance against HBase data, so if anything, Phoenix's advantage has only gotten bigger as our performance improves.  See this link for comparison of Apache Phoenix with Apache Hive and Cloudera Im…

Data deduplication tactics with HDFS and MapReduce

As the amount of data continues to grow exponentially, there has been increased focus on stored data reduction methods. Data compression, single instance store and data deduplication are among the common techniques employed for stored data reduction.
Deduplication often refers to elimination of redundant subfiles (also known as chunks, blocks, or extents). Unlike compression, data is not changed and eliminates storage capacity for identical data. Data deduplication offers significant advantage in terms of reduction in storage, network bandwidth and promises increased scalability.
From a simplistic use case perspective, we can see application in removing duplicates in Call Detail Record (CDR) for a Telecom carrier. Similarly, we may apply the technique to optimize on network traffic carrying the same data packets.
Some of the common methods for data deduplication in storage architecture include hashing, binary comparison and delta differencing. In this post, we focus on how MapReduce and…

Pricing models for Hadoop products

A look at the various pricing models adopted by the vendors in the Hadoop ecosystem. While the pricing models are evolving in this rapid and dynamic market, listed below are some of the major variations utilized by companies in the sphere.
1) Per Node:Among the most common model, the node based pricing mechanism utilizes customized rules for determining pricing per node. This may be as straight forward as pricing per name node and data node or could have complex variants of pricing based on number of core processors utilized by the nodes in the cluster or per user license in case of applications.
2) Per TB:The data based pricing mechanism charges customer for license cost per TB of data. This model usually accounts non replicated data for computation of cost.
3) Subscription Support cost only:In this model, the vendor prefers to give away software for free but charges the customer for subscription support on a specified number of nodes. The support timings and level of support further …