Press "Enter" to skip to content

Category: Hadoop

Spark Deep Learning On AWS

Published 2017-04-25 by Kevin Feasel

Joseph Spisak, et al, show how to configure and use BigDL in Amazon Web Services’s ElasticMapReduce:

Classify text using BigDL

In this tutorial, we demonstrate how to solve a text classification problem based on the example found here. This example uses a convolutional neural network to classify posts in the 20 Newsgroup dataset into 20 categories.

We’ve provided a companion Jupyter notebook example on GitHub that you can open in the Jupyter dashboard to execute the code sections.

There’s a lot to this tutorial.

Comments closed

Pipeline Architecture With Kafka

Published 2017-04-17 by Kevin Feasel

Alexandra Wang describes how Pandora Media has used Apache Kafka for real-time ad serving using Kafka Connect:

Our ad server publishes billions of messages per day to Kafka. We soon realized that writing a proprietary Kafka consumer able to handle that amount of data with the desired offset management logic would be non-trivial, especially when requiring exactly-once-delivery semantics. We found that the Kafka Connect API paired with the HDFS connector developed by Confluent would be perfect for our use case.

We’ve also found it painful not having a central authority on data structures that can share their respective schemas across all services and applications. Without a central registry for message schemas, data serialization and deserialization for a variety of applications are troublesome and the pipeline is fragile when schema evolution happens. We found Schema Registry is a great solution for this problem.

To address the above two problems, we integrated the Kafka Connect API and Schema Registry into our Kafka-centered data pipeline.

Well worth reading, especially the difficulties that they’ve had during maintenance periods and in lower environments.

Comments closed

Using h2o.ai On HDInsight

Published 2017-04-13 by Kevin Feasel

Xiaoyong Zhu shows how to set up h2o.ai on Azure HDInsight:

H2O Flow is an interactive web-based computational user interface where you can combine code execution, text, mathematics, plots and rich media into a single document, much like Jupyter Notebooks. With H2O Flow, you can capture, rerun, annotate, present, and share your workflow. H2O Flow allows you to use H2O interactively to import files, build models, and iteratively improve them. Based on your models, you can make predictions and add rich text to create vignettes of your work – all within Flow’s browser-based environment. In this blog, we will only focus on its visualization part.

H2O FLOW web service lives in the Spark driver and is routed through the HDInsight gateway, so it can only be accessed when the spark application/Notebook is running

You can click the available link in the Jupyter Notebook, or you can directly access this URL:

https://yourclustername-h2o.apps.azurehdinsight.net/flow/index.html

Setup is pretty easy.

Comments closed

HDInsight 3.6 Available

Published 2017-04-10 by Kevin Feasel

Ashish Thapliyal points out some Hive improvements in HDInsight 3.6:

2 Create a new Hive table from scratch or alter Table

Create a new table by, clicking on the ‘+’ icon, which opens the create table wizard. Enter table name, column name and choose a data type from the dropdown. You can pick folloiwng advanced hive settings directly from the UI

  • Transactional : Turn on transaction support in Hive, by checking this flag. Note that the table must be bucketed and stored using an ACID compliant format (such as ORC).

  • Location : Hive stores the table data for managed tables in the Hive warehouse directory in HDFS which is configured in hive-site.xml with property hive.metastore.warehouse.dir. The default location is /apps/hive/warehouse. The location can be changed using the Location text field.

  • File Format : The default file format for CREATE TABLE statement is ORC. choose a format from the file format dropdown.

  • Row Format : Select a row format such as Field terminator, Lines terminator, and Stored File type.

  • Table can be altered to add new columns or change the column name or column datatype.

  • Tables can also be renames and altred

Read on for more improvements, including a graphical plan viewer and improved autocomplete.

Comments closed

MERGE In Hive

Published 2017-04-10 by Kevin Feasel

Carter Shanklin notes that Hive now has the ability to run MERGE statements:

As scalable as Apache Hadoop is, many workloads don’t work well in the Hadoop environment because they need frequent or unpredictable updates. Updates using hand-written Apache Hive or Apache Spark jobs are extremely complex.  Not only are developers responsible for the update logic, they must also implement all rollback logic, detect and resolve write conflicts and find some way to isolate downstream consumers from in-progress updates. Hadoop has limited facilities for solving these problems and people who attempted it usually ended up limiting updates to a single writer and disabling all readers while updates are in progress.

This approach is too complicated and can’t meet reasonable SLAs for most applications. For many, Hadoop became just a place for analytics offload — a place to copy data and run complex analytics where they can’t interfere with the “real” work happening in the EDW.

This post mostly describes the gains rather than showing code, but it does show that Hive developers are looking at expanding beyond common Hadoop warehousing scenarios.

Comments closed

Architecting Kafka Streams

Published 2017-04-03 by Kevin Feasel

Bill Bejeck walks through a scenario in which one might use Kafka Streams:

Now, you’ve defined your source and we can start creating processors that’ll do the work on the data. The first goal is to mask the credit card numbers recorded in the incoming purchase records. The first processor is used to convert credit card numbers from 1234-5678-9123-2233 to xxxx-xxxx-xxxx-2233. The Stream.mapValues method performs the masking. The KStream.mapValues method returns a new KStream instance that changes the values, as specified by the given ValueMapper, as records flow through the stream. This particular KStream instance is the parent processor for any other processors you define. Our new parent processor provides the masked credit card numbers to any downstream processors with Purchase objects.

Unfortunately, this article seems like a mixture of high-level and low-level information that appeals more to people who already know how Kafka Streams works, but it is nevertheless interesting.

Comments closed

Encrypting Kinesis Records

Published 2017-04-03 by Kevin Feasel

Temitayo Olajide shows how to use Amazon’s Key Management Service to encrypt and decrypt Kinesis messages:

In this post you build encryption and decryption into sample Kinesis producer and consumer applications using the Amazon Kinesis Producer Library (KPL), the Amazon Kinesis Consumer Library (KCL), AWS KMS, and the aws-encryption-sdk. The methods and the techniques used in this post to encrypt and decrypt Kinesis records can be easily replicated into your architecture. Some constraints:

  • AWS charges for the use of KMS API requests for encryption and decryption, for more information see AWS KMS Pricing.

  • You cannot use Amazon Kinesis Analytics to query Amazon Kinesis Streams with records encrypted by clients in this sample application.

  • If your application requires low latency processing, note that there will be a slight hit in latency.

Check it out, especially if you’re thinking about streaming sensitive data.

Comments closed

Rolling A Log Analytics System

Published 2017-04-03 by Kevin Feasel

Michael Sun and Jeff Shmain put together a log analytics sytem using several technologies:

This is an example of tiered system design. Tiered system is a system design pattern where data is categorized and stored in different data stores that match best to each category. It can both improve performance and lower the cost of a system. One of the most famous tiered system designs is computer memory hierarchy.  In the log analytics use case, analysts mostly search for logs in recent months, but often run batch jobs to get long term trends from logs in recent years. Therefore, recent logs are indexed and stored in Solr for search, while years of logs are stored in HBase for batch processing. As such, the index in Solr is small, which both improves performance and reduces cost, among other benefits.

Although only months of logs are stored in Solr, the logs before that period are stored in HBase and can be indexed on demand for further analysis.

Now that we have covered a high level architecture of a log analytics system, we will dive into more details of individual components.

This looks like a solid architecture for a logging system and can apply to other cases as well.

Comments closed

Introduction To Amazon Kinesis

Published 2017-04-03 by Kevin Feasel

Jen Underwood describes Amazon Kinesis:

Amazon Kinesis is a fully managed service for real-time processing of streaming data at massive scale. Amazon Kinesis is ideal for Internet of Things (IoT) use cases. It can collect and process hundreds of terabytes of data per hour from hundreds of thousands of sources, allowing you to easily write applications that process information in real-time, from sources such as web site click-streams, Raspberry Pi gadgets, devices, social media, operational logs, metering data and more.

With Amazon Kinesis, you can build real-time dashboards, capture exceptions, execute algorithms, and generate alerts. With point-and-click menus, you can ingest data, query it and then send output to a variety of destinations including but not limited to Amazon S3, Amazon EMR, Amazon DynamoDB, or Amazon Redshift.

Kinesis is powerful, especially if you’re already locked into the AWS platform.  My preference is Apache Kafka, but Kinesis is definitely worth learning about.

Comments closed

Tuning Kafka And Spark Data Pipelines

Published 2017-03-29 by Kevin Feasel

Larry Murdock explains the tuning options available to Kafka and Spark Streams:

Kafka is not the Ferrari of messaging middleware, rather it is the salt flats rocket car. It is fast, but don’t expect to find an AUX jack for your iPhone. Everything is stripped down for speed.

Compared to other messaging middleware, the core is simpler and handles fewer features. It is a transaction log and its job is to take the message you sent asynchronously and write it to disk as soon as possible, returning an acknowledgement once it is committed via an optional callback. You can force a degree of synchronicity by chaining a get to the send call, but that is kind of cheating Kafka’s intention. It does not send it on to a receiver. It only does pub-sub. It does not handle back pressure for you.

I like this as a high-level overview of the different options available.  Definitely gets a More Research Is Required tag, but this post helps you figure out where to go next.

Comments closed