Data Infrastructure in the Cloud

Having your data infrastructure in the cloud has become a real option for a lot of companies, especially since the big cloud providers have a lot of managed services available for a modern data architecture aside from just a database management system.

In this article I will look into what the three big cloud providers offer in tools to support a modern data architecture.

Modern Data Infrastructure Setup

A generic data infrastructure looks like the picture shown below. It contains one data entry point, where all data is collected as the general API for all producers of data. This data is pushed in single events, whenever possible and defined using a schema. All of the schemas are defined and stored in a schema registry.

Following this entry point there is the possibility to do some ETL, like collecting several messages for one event into bigger chunks. This is useful for writing data out onto a storage or inserting into an analytical database. As these do not handle single row operations as well as bulk ones. Storing raw event data on a cloud storage is also possible and then access these files using SQL from the DWH system. This potentially saves more expensive storage in the DWH system.

After the initial ETL transformations done on the events it seems sensible to do all the other transformations from the source using the compute power of the database system. As this can be done by using SQL, this offers the possibility to have people like analysts with SQL knowledge participate in the process. No other programming skills are needed from this point on.

Modern Data Infrastructure
Modern Data Infrastructure

Next we will look into the tools offered by the three largest cloud vendors that support this setup.

Google Cloud Platform

The Google Cloud Platform offers tools to implement this architecture, as seen in the image below.

Modern Data Architecture in the Cloud Googles version
Modern Data Architecture in the Cloud Googles version
  • Schema Registry: Google offers no dedicated schema registry, but it is possible to use the Confluent Schema Registry, if needed. If you want to use AVRO schemas it is possible to use this schema repo, but it is not maintained anymore.
  • PubSub is a messaging system. It only delivers once and data replay is not possible. The messages are also not ordered.
  • Cloud Dataflow is a serverless data processing platform, which can be programmed using Apache Beam. The upside of using Apache Beam is that it supports several processing platforms, like Spark, Flink. So you are flexible to move away from Dataflow, if you chose to.
  • Cloud Storage is a managed object storage form Google.
  • Bigquery is Google’s solution for a Datawarehouse. It is kind of a specialized SQL on Hadoop solution and serverless. Compute and storage are seperated, so the engine scales for each query as needed.

Amazon Web Services

Amazon Web Services also offers tools to implement the data architecture.

Modern Data Architecture in the Cloud Googles version
Modern Data Architecture in the AWS version
  • AWS Glue provides a schema registry that can be used for Kinesis streams and transformations.
  • Kinesis Data Stream provides a platform for streaming events. Data replay is possible here and messages are stored in an ordered fashion.
  • Kinesis Firehose takes the data streams and offers some in-built transformations like storing data aggregated in an Apache Parquet or Apache ORC format. Both formats are column oriented and fast to query from their Datawarehouse solution Redshift. Custom transformations are also supported using AWS Lambda. Lambda provides serverless computing for transformation code.
  • S3 offers cheap object storage for object from Amazon
  • Redshift is Amazon’s datawarehouse solution. It is a multi-parallel-processing engine (MPP) based on a PostgreSQL database. Here compute and storage are not seperated, so you need to size your cluster appropriately. Also there is some administrative effort necessary for data optimizing. This includes vacuuming, index creation, and table analysis. Sizing is also a manual task.

Microsoft Azure

Microsoft Azure offers other tools that support this data infrastructure in the cloud.

Modern Data Architecture in the Cloud Azure version
Modern Data Architecture in the Cloud Azure version
  • Data Catalog discovers data schemas and you can define them here as well.
  • EventHubs is a serverless streaming solution for events. Events can be ordered by using the partitioning function that is provided.
  • Data Factory enables you to collect and transform all data coming into the data platform.
  • Data Lake Storage object storage for data analytics that decouples compute from storage.
  • Synapse Analytics is a framework for storing and analysing all data, like Spark or an MPP engine for SQL workloads. It seemlessly integrates with long existing MS products like SSIS.


Data Infrastructure in the Cloud is possible with all three major cloud providers. Each of them offers tools to set up this solution. Which one fits your needs best depends propably most on the preference and knowledge of your data engineers, scientists and analysts or even your other infrastructure setup.

Google Cloud Data Engineer Exam Preparation

Google Cloud Data Engineer Exam

This is a little text with all the stuff that helped me prepare for the Google Cloud Data Engineer Exam. There are a lot of courses and resources, that help you in preparing for this. The following links helped me in preparation for my Google Data Engineer Exam.

On Coursera there is are several courses provided by Google to prepare for this exam, including a Specialization for the Google Cloud Data Engineer Exam.

And to complete your preparation for the exam Coursera also offers a final preparation course Preparing for the Google Cloud Professional Data Engineer Exam

Other resources I used included these:

And Google also offers a test exam for the Google Cloud Data Engineer.

And now to a little deeper description of the topics I struggled a bit with.

Big Query


When you use streaming to insert data into BigQuery, it gets saved into the streaming buffer and then is loaded into the persistent table from there. It can take up to 90 minutes until the data is persistent. But all data is queryable together, buffer and persistent table. So your results update immediately.

During streaming inserts each row gets an insertID which Bigquery uses to try, find and remove duplicates, if these occur in a time window of one minute.

If this method is not secure enough, Google Cloud Datastore is an option, since it supports transactions.

You can also stream into partitioned tables, if you add a partition decorator $YYYYMMDD to the table. Streamed data gets saved to the UNPARTITIONED partition. To check this query the system column _PARTITIONDATE IS NULL.

Typical use cases include

  • high volume input, but not transactional data
  • analysis on aggregated data and not for single row queries


  • max row size: 1 MB
  • HTTP limit: 10 MB
  • max 10,000 rows / second per project
  • max 10,000 rows / query; 500 rows recommended
  • max 100 MB / s and table inserts in streaming
  • API limit of 100 queries / s per user
  • max 300 concurrent users
  • max 500 tablesets / s

Partitioned tables

Partitioned tables are separated into smaller units to optimize query performance. There are several ways to to partition tables in BigQuery:

  • partitioned by loading time (automatically)
  • partitioned by field with the data type TIMESTAMP or DATE

You can correct loading time by updating the suffix or pseudo columns. Partitioned tables contain a system column _PARTITION_TIME, you can use to filter. If you used an explizit, not automatic partitioning, the column _PARTITION_TIME is not exiting.

There are two columns for data, that do not belong to partitions:

  • __NULL for rows with a NULL value in the date field
  • __UNPARTITIONED for rows with a date outside of the valid timerange#

Another way to optimize tables, is sharding, where you use different tables to split data, instead of a partitioning column. Join these tables using a UNION. In this case keep in mind, that there is a limit of 1,000 tables per query.

Access rights

  • viewer: right to read all tables in a dataset
  • editor: viewer plus creating, deleting and changing of tables in a dataset
  • owner: editor plus creating and deleting datasets
  • admin: everything


There are some restrictions you need to keep in mind, when using BigQuery:

  • max 100 concurrent interactive queries possible
  • max 4 concurrent interactive queries on external sources
  • max 6 UDFs in a single interactive query
  • max 1,000 updates per table per day
  • max query duration is 6 hours
  • max 1,000 tables per query
  • SQL command can be max 1 MB in size
  • max 10,000 query parameters
  • max 100 MB / row
  • max 1,000 columns per table / query / view
  • max 16 levels of nested fields

Best Practices

ETL from relational database to BigQuery

There are several ways to import data from a relational database into BigQuery. Do this by using the WebUI, in which case the schema is upload able as as JSON file. You can export data to Google Cloud Storage in JSON format and import from there.

Another way is using Google Cloud Dataflow. Automatic setup is possible, but you can customize the process and maybe join data already in Dataflow, to denormalize it by using lookup tables as side input.

Backup / Snapshots

BigQuery supports a feature for “Point in Time Snapshots”. If you you a query decorator @<time> in Legacy SQL you can get an image of the table during the specified timestamp. This decorator can contain relative or absolute values.

BiqQuery ML

BigQuery supports machine learning in SQL for structured data. This is easy to learn, as it is based on SQL and not some other programming language. Right now it does provide linear regression, binary logistic regression for classification, multi class logistic regression and k-means clustering. The advantage of using this is, that you only have to use one tool, SQL, for everything. This means, data need not be moved from one place to another.


BigQuery does support user defined functions (UDF) in the languages SQL or JavaScript. Load external libraries by storing them in GCS.

Denormalize data

BigQuery does not need normalized data for fast query speed, but rather denormalized data. This means you should use nested and repeated fields in tables. Disk usage is not an issue here and denormalized data saves you the need to join.

You can reuse nested and repeated fields by using formats like AVRO, Parquet, ORC, JSON, Datastore or Firestore.

BigQuery Transfer Service

There is an option to repeatedly load data into BigQuery from Campaign Manager, GCS, S3, AdManager, Google Ads, Google Play and Youtube. You can also load Stackdriver logs into BigQuery.


BigTable is Googles solution for a real time key / value datastore. It handles sparse tables, meaning tables, where a lot of columns in rows are null. Empty columns do not use any space. This system is designed for billions of rows, thousands of columns and data the size of TB or PB.

BigTable only has one key per row, called rowkey. It has small latency and is HBase compatible. Also it scales linearly corresponding to the number of nodes in a cluster. There is an option for replication over two clusters, in which sets up one as cold standby. It is possible to change cluster sizes while the system is running, but after that BigTable takes a few minutes to rebalance cluster load.

To make the most out of it, its use case is for high scalability, key / value data and streaming.

BigTable orders columns belonging together into groups or families. Empty columns do not use any space. To make the most of it, divide key values equally on all nodes. Data is not stored on the cluster nodes, but on Colossus. The nodes only contain references to the data. This makes it easy to recover data after a system crash and no data is lost if nodes fail.

The master node distributes data node to optimize performance. Data locality is important. It can be influenced by key design. Group data that is queried together by designing key in that way. But BigTable does not support SQL or joins and is only really recommended with over 1TB of data.


SSDread 10,000 rows in 6 ms
write 10,000 rows in 6 ms
HDDread 5,000 rows in 200 ms
write 10,000 rows in 50 ms

Reasons for underperformance

  • incorrect schema design: key does not distribute data equally, only one key is possible
  • columns are sorted by family and then alphabetical
  • reads / writes are scewed
  • all operations need to be atomic on a row
  • information is distributed over several rows
  • row is to big (only 10MB per cell and 100MB per row)
  • timestamp is beginning of rowkey
  • not using combined keys
  • key consists of domains, hash or sequential ids
  • too little data (< 300GB)
  • test was to short (needs to run at least 10 minutes)
  • cluster too small (>70% CPU; 70% space)
  • HDD is used
  • still is a development instance
  • data is distributed automatically due to usage pattern, make sure to give system time to distribute to your needs

Key Visualizer

The Key Visualizer is a tool for supporting you in finding BigTable usage patterns. It provides graphic reports to help you recognize performance problems, like spotting hotspots (rows with a lot of actions) or rows with too much data. It helps you in getting evenly distributed accesses over all rows.

The reports are generated automatically (daily / hourly) if a table has more than 300GB or at least 10,000 row read or writes per second. The reports also recognize hierarchical keys and provides them as this hierarchy. Also connected keys are summarized.

Scale BigTable

You can scale BigTable programmatically. This is helpful if you, for example get monitoring data from Stackdriver and then react on this databy scaling up or down. After scaling a cluster it can take up to 20 minutes until you can see better performance, as data needs to be redistributed.

This automatic process does not help in cases where performance peaks for a short period of time. In this case, you need to scale manually before the event. If scaling does not help performance, then probably the key design leads to imbalanced load distribution.

You can scale a running cluster. It is possible to change the following things:

  • nodes in cluster
  • cluster per instance
  • usage profiles
  • label / names
  • upgrade from development to production instance

Schema design

Schema design in BigTable is crucial for performance. There is only one index per row, called rowkey. All rows are automatically sorted due to the rowkey, in alphabetical order. And as all operations are atomical per row, rows should not be dependent on one another.

Use key design to make sure data queried together is lying close to each other. And keep in mind, that one cell should not be bigger than 10MB and one row no larger than 100MB. The max values for this is 100MB per cell and 256MB per row.

There are up to 100 column families possible. If all dependent values are in one family, it is possible to only query this family. Less columns is better and column names should be short.

Key design

There are several best pratices for key design:

  • use inverted domain names, like
  • string ids should not contain hashes, due to readability
  • integrate timestamp if time is queried often, but not only timestamp
  • use inverted timestamp if newest data is queried more often
  • always use grouping field first in key
  • try to avoid a lot of updates on a row, try to write several rows instead


Dataflow is a system to develop and run data transformation pipelines in. It has the following key concepts:

  • Pipeline: complete process in a DAG
  • PCollection: distributed dataset for a pipeline
  • PTransform: data processing operation, that takes a PCollection and performs processing. It returns again a PCollection
  • ParDo: Parallel processing runs a function on a PCollection
  • SideInputs: additional inputs for ParDo / DoFns

It is a best practice to write invalid input to its own sink and not to throw it away.


Messaging queue provided by Google. It keeps data up to seven days. All logs are automatically pushed to Stackdriver for analysis. It has the following attributes:

  • messages might not be delivered in the order they were received in, because the system is distributed.
  • if message delivery is not acknowledged after a defined time, message will be redelivered. This timespan is configurable

Data Migrations

Data Transfer Appliance

The Data Transfer Appliance is an encrypted offline data transfer up to PB in size. You get the appliance and ship it back to Google where it will be uploaded into the cloud. This is faster than internet upload and can be handled like a NAS. Only applicable if the data is not changed in between.

Storage Transfer Service

Storage Transfer Service is a transfer service online to GCS or from bucket to bucket. It can be used to backup online data, or as a sync from source to sink, including deletion of source once it is copied. If you need to decide between gsutil and Storage Transfer Service it is:

  • local data -> gsutil; this also supports rsync
  • other cloud providers -> Storage Transfer

Dedicated Interconnect

Dedicated Interconnect is a physical connection between on premise and Google Cloud. You provide your own routing equipment. It is possible at 10 or 100Gbit.


Dataproc is a managed Hadoop / Spark cluster in Google Cloud. It offers some specials:

  • Cloud Storage Connector: Hadoop or Spark jobs access GCS directly
    • data is not transferred to HDFS
    • GCS is HDFS compatible, just use gs:// instead of hdfs://
    • data still available if cluster is gone
    • high availability is possible
  • Preemptible workers:
    • same machine type is others
    • own instance group
    • lost instances will be re-added once there are free resources
    • save no data
    • no clusters with just preemptible nodes
    • local disk is not available in hdfs, it is only used for local caching
  • Scale cluster:
    • scaling is always possible, even with running jobs

Machine Learning

Machine Learning Engine

The Machine Learning Engine provides predefined scaling values, but there is also a custom tier. There you can change:

  • masterType: machine type for the master
  • workerCount: number of workers in cluster
  • parameterServerCount: number of parameter servers

Data Loss Prevention API

This API finds sensible data (PPI) and can remove it, e.g. personal data, payment data, device identification, also custom and country specific. You can define word lists and regexes. It also provides data masking, removing and encryption. Use it in streams, text, files, BigQuery, pictures

Natural Language API

Textanalysis, Sentiment Analysis, Topic Mining, People and Location identification

Vision API

Picture analysis, object recognition, etc


Translation, Speech to Text, Text to Speech, Dialogflow as Chatbot, Video API


Datastore is a NoSQL DB, that is ACID compliant and has a SQL like querying language. There is an index for rows, combined indexes are possible too. It also provides a Firestore export to a GCS bucket.


Reporting solution provided by Google. It provides caching to speed up reports or reduce costs. There are two different kind of caches:

  • responsive cache: remembers results of a query and if data is queried again, cache supplies it
  • predictive cache:
    • analyses dimensions, metrics and filters and tries to predict possible queries.
    • queries are executed in background and response is stored in predictable cache
    • if responsive cache has no data, predictive cache is asked
    • if no data in predictive cache, data is queried from source
    • works only with owner credentials

Cache is automatically refreshed, but it can be done manually. The default for refreshing is 15 minutes, but also 4 and 12 hours. You should turn of predictive cache in case of:

  • data changes frequently and data freshness is more important than speed
  • reduce costs by turning it off

After the Google Cloud Data Engineer Exam

After taking the exam, you get a preliminary result on the spot. You can see it after submission or later in your Webassessor account. Once Google checked the exam you get an email with a voucher and a link to their perk shop, where you can chose between a backpack and a hoodie. I chose the backpack.

Google Cloud Data Engineer Exam Perk
Google Cloud Data Engineer Exam Perk

Plumber: Getting R ready for production environments?

R Project and Production

Running R Project in production is a controversially discussed topic, as is everything concerning R vs Python. Lately there have been some additions to the R Project, that made me look into this again. Researching R and its usage in production environments I came across several packages / project, that can be used as a solution for this:

There are several more, but those I found the most interesting.


For reasons of ease of use and because it was not a hosted version, I took a deeper look into Plumber. This felt quite natural as it uses function decorators for defining endpoints and parameters. This is similar to Spring Boot, which I normally use for programming REST APIs.
So using Plumber is really straight forward, as the example below shows:

#' return text "Hello"
#' @get /hello
function() {
  list(msg = "hello")

The #’ @get defines the endpoint for this request. In this case /hello, so the full url on localhost is To pass in one or more parameters you can use the decorator #’ @param parameter_name parameter_description. A more complicated example using Plumber is hosted on our Gitlab. This example was built with the help of Tidy Textmining.

Production ready?

Plumber comes with Swagger, so the webserver is automatically available. As the R instance is already running, processing the R code does not take long. If your model is complicated, then, of course, this is reflected in the processing time. But as R is a single thread programming language, Plumber can only process one request at a time.
There are ways to tweak this of course. You can run several instances of the service, using a Docker image. This is decribed here. There is also the option of using a webserver to fork the request to serveral instances of R. Depending on the need of the API, single thread processing can be fast enough. If the service has to be highly available the Docker solutions seems like a good choice, as it comes with a load balancer.


After testing Plumber I am surprised by the easy of use. This package makes deploying an REST API in R really easy. Depending on your business needs, it might even be enough for a productive scenario, especially when used in combination with Docker.

Building a Productive Data Lake: How to keep three systems in sync

Three Systems for save Development

When you are building a productive Data Lake it is important to have at least three environments:

  • Development: for development, where “everything” is allowed.
  • Staging: for testing changes in a production like environment.
  • Production: Running your tested and productive data applications

With these different environments comes the need to keep data structures in sync. You need to find a way to change each system in a controlled way, when it is time to deploy some changes. There are differences in how to manage data structures, depending on where in data processing they occur. Theses differences we split into the parts “Data Entry to Staging” and “Core and Presentation”. I wi

Managing different data structures in the three systems of your Productive Data Lake

As all three systems can be have different data structures, using tools to make this easier is important. Different Apache Avro™ schema versions are managed by a publishing them to a tool called Apache Avro™ Schema Service of which there also also three instances. By publishing changes first on the development, then staging, then production system, we make sure, only tested data structures are deployed to the Production Data Lake. Each change to a schema belongs to a version number, so rolling back to an earlier state, in case of an error is possible.
Evolving our data structures not contained in Apache Avro™ schemas, we use tables managing all meta information.

Data Structures from Entry to Staging

Apache Avro™ is a good way to manage evolving data structures, as I explained here. Once the data is available in Apache Hive™ mapping the AVRO data types to those of your Productive Data Lake system is necessary. We use Apache HAWQ™ as base of our Data Lake. So we map the AVRO data types to HAWQ data types. To keep this as effortless as possible, we reuse attribute names as often as possible. If an attribute, like “user_id” is occuring in different schemas and has the same meaning, then we just reuse it. So the mapping table does not contain an entry for each column of each schema, but only for distinct ones. Considering more than 100 schema with on average 10 attributes, this helps in keeping maintanence down. In the table below is an example, of this mapping.

Schema FieldnameAVRO datatypeHAWQ datatype

Data Strucutes for Core and Presentation

A totally different challenge is it, to manage the data structures after data entry. There need to occur transformations, for technical and business user benefit. This can mean other attributes / column names or a splitting or joining of data from one schema into different tables. These tables here again can evolve. Evolving database tables is a pain, since you cannot, in most cases just drop the table and recreate the new one. Here the processing power of Apache HAWQ™ comes into play, as we can exactly do that now. In most cases, at least for transactional data, we just reload the whole table, during the daily batch process. So we manage all table definitions in tables. They contain all table and column attributes, as well as a version number. This makes rolling back possible.

Example of table definitions:

Schema NameTable NamePartitioneddistributed bydropableVersion
corerel_user_purchaseuser_id, purchase_idtrue4.0

Exmaple of column definitions:

Schema NameTable NameColumn NameData TypeData Type LengthVersion

Automation of data structure deployment.

After defining all the meta data concerning the data structures, we need a way to automate this. For this I created a Java programm using all the information from schema definitions and tables for data type mappings and table definitions. This programm runs on any of the three environments and also take in consideration the version it should create. This only works, if no one can create or change table information manually. So the programm first deploys changes to the data structure on the development system. If it works there and all changes are ready to be deployed on the staging system, run the programm in this environment. With the version numbers defined besides each change, it is possible to run version 5.0 on production, version 5.1 on staging and version 5.2 on development if neccessary.


This way is a possibility to make data structure evolution easier and keep three different systems in sync. It is most important to make sure, that there are no manual changes to the data structure. This approach works nicely, if you can drop and recreate all tables without hesitation. If that is not the case, you can still define your tables this way, just make sure they are not dropped automatically when the programm runs. Having all definitions in one place and being able to automatically deploy the tables and the changes, with version control, can be a great help. This way no one has to remember to create new tables manually in either system. So errors due to missing tables or columns are minimized which is good if you create a Productive Data Lake, where errors affect data products needed by other products of your company.

Apache HAWQ: Building an easily accessable Data Lake

Apache HAWQ for Data Lake ArchitectureData Lake vs Datawarehouse

The Data Lake Architecture is an up and coming approach to making all data accessible through several methods, be that in real-time or batch analysis. This includes unstructured data as well as structured data. In this approach the data is stored on HDFS and made accessible by several tools, including:

All of these tools have advantages and disadvantages when used to process data, but all of them combined make your data accessible. This is the first step in building a Data Lake. You have to have your data, even schemaless data accessible to your customers.
A classical Datawarehouse on the opposite only contains structured data, that is at least preproccessed and has a fixed schema. Data in a classical Datawarehouse is not the raw data entered into the system. You need a seperate staging area for tranformations. Usually this is not accessible for all consumers of your data, but only for the Datawarehouse developers.

Data Lake Architecture using Apache HAWQ

It is a challenge to build a Data Lake with Apache HAWQ, but this can be overcome on the design part. One solution to build such a system can be seen in then picture below.

Data Entry

To make utilization of Apache HAWQ possible the starting point is a controlled Data Entry. This is a compromise between schemaless and schematized data. Apache AVRO is a way to do this. Schema evolution is an integral part of AVRO and it provides structures to save unstrcutured data, like maps and arrays. A separate article about AVRO will be one of this next topics here, to explain schema evolution and how to make the most of it.
Data structured in schema can then be pushed message wise into a messaging queue. Chose a queue that fits your needs best. If you need secure transactions RabbitMQ may be the right choice. Another option is Apache Kafka.

Pre-aggregating Data

Processing and storing single message on HDFS is not an option, so there is need of another system to aggregate messages before storing them on HDFS. For this a software project called Apache Nifi is a good choice. This system comes with processors that make things like this pretty easy. It has a processor called MergeContent that merges single AVRO messages and removes all headers but one, before writing them to HDFS.
If those messages are still not above the HDFS blocksize, there is the possibility to read messages from HDFS and merge them into larger files still.

Making data available in the Data Lake

Use Apache Hive to make data accessible from AVRO format. HAWQ could read the AVRO files directly, but Hive handles schema evolution in a more effective way. For example, if there is the need to add a new optional field to an existing schema, add a default value for that field and Hive will fill entries from earlier messages with this value. So if HAWQ now accesses this Hive table it automatically reads the default value for field added later with default values. It could not do this by itself. Hive also has a more robust way of handling and extracting keys and values from map fields right now.

Data Lake with SQL Access

All data is available in Apache HAWQ now. This enables tranformations using SQL and making all of your data accessible by a broad audience in your company. SQL skills are more common than say Spark programming in Java, Scala or PySpark. From here it is possible to give analysts access to all of the data or building data marts for single subjects of concern using SQL transformations. Connectivity to reporting tools like Tableau is possible with a driver for Postgresql. Even advanced analytics are possible, if you install Apache MADlib on your HAWQ cluster.

Using Data outside of HAWQ

It is even possible to use all data outside of HAWQ, if there is a need for it. Since all data is available in AVRO format, accessing it by means of Apache Spark with Apache Zeppelin is also possible. Hive queries are possible too, since all data is registered there using external tables, which we used for integration into HAWQ.
Accessing results of such processing in HAWQ is possible too. Save the results in AVRO format for integration in the way described above or use “hawq register” to access parquet files directly from HDFS.


Using Apache HAWQ as base of a Data Lake is possible. Just take some contraints into consideration. But entering data with semi-structured with AVRO format also saves work later when you process the data. The main advantage is, that you can utilize SQL as an interface to all of you data. This enables many people in your company to access your data and will help you on your way to Data Driven decisions.

Apache Zeppelin: Use with remote Spark cluster and Yarn

Apache Zeppelin is pretty usefull for interactive programming using the web browser. It even comes with its own installation of Apache Spark. For further information you can check my earlier post.
But the real power in using Spark with Zeppelin lies in its easy way to connect it to your existing Spark cluster using YARN. The following steps are necessary:

  • Copy your Hadoop configuration files to your Zeppelin installation under $ZEPPELIN_HOME/conf
  • Restart your Zeppelin Notebook
  • Insert the value “yarn-client” into the field master in the spark interpreter, as shown in the picture below.


After these steps you can use your notebooks with spark running on a yarn cluster. So you can make use of all the resources in the queue you assigned spark on you cluster.

Apache Zeppelin: Visualization and Spark data processing

Apache Zeppelin

Apache Zeppelin is a web-based notebook for interactive data analytics. It comes will features for all the steps of data analysis:

  • Data Ingestion
  • Data Discovery
  • Data Analytics
  • Data Visualization & Collaboration

Besides that feature set it also supports multiple languages in the backend. Currently it supports languages like:

But there is also the possibility to add your own interpreter to Zeppelin. This makes this tool really flexible.
Another feature it has, is the built in integration of Apache Spark. It ships with the following features and more:

  • Automatic SparkContext and SQLContext injection
  • Runtime jar dependency loading from local filesystem or maven repository.
  • Canceling job and displaying its progress

It also has built in visualization, which is an improvemnt over using ipython notebooks I think. The visualization covers the most basic graphs, like:

  • Tables
  • BarCharts
  • Pies
  • Scatterplot
  • Lines

These visualizations can be used with all interpreters and are always the same. So you can show data from Postgres and Spark in the same notebook with the same functions used. There is no need to handle different data sources differently.
You can also use dynamic forms in your notebooks, e.g. to provide filter options to the user. This comes in handy, if you embedd a notebook in your own website.

Apache Spark 2.0

Apache Spark has release version 2.0, which is a major step forward in usability for Spark users and mostly for people, who refrained from using it, due to the costs of learning a new programming language or tool. This is in the past now, as Spark 2.0 supports improved SQL functionalities with SQL2003 support. It can now run all 99 TPC-DS. The new SQL parser supports ANSI-SQL and HiveQL and sub queries.
Another new features is native csv data source support, based on the already existing Databricks spark csv module. I personally used this module as well as the spark avro module before and they make working with data in those formats really easy.
Also there were some new features added to MLlib:

  • PySpark includes new algorithms like LDA, Gaussian Mixture Model, Generalized Linear Regression
  • SparkR now includes generalized linear models, naive Bayes, k-means clustering, and survival regression.

Spark increased its performance with the release of 2.0. The goal was to make Spark 2.0 10x faster and Databricks shows this performance tuning in a notebook.

All of these improvements make Spark a more complete tool for data processing and analysing. The added SQL2003 support even makes it available for a larger user base and more importantly makes it easier to migrate existing applications from databases to Spark.

Python vs. R for Data Science

In Data Science there are two languages that compete for users. On one side there is R, on the other Python. Both have a huge userbase, but there is some discussion, which is better to use in a Data Science context. Lets explore both a bit:

R is a language and programming environment especially developed for statistical computing and grahics. It has been around some time and several thousand packages to tackle statistical problems. With RStudio it also provides an interactive programming environment, that makes analysing data pretty easy.

Python is a full range programming language, that makes it easy to integrate into a company wide system. With the packages Numpy, Pandas and Scikit-learn, Mathplotlib in combination with IPython, it also provides a full range suite for statistical computing and interactive programming environment.

R was developed solely for the purpose of statistical computing, so it has some advantages there, since it is specialized and has been around some years. Python is coming from a programming language and moves now into the data analysis field. In combination with all the other stuff it can do, websites and easy integrations into Hadoop Streaming or Apache Spark.
And for people who want to use the best of both sides can always use the R Python integration Rpy2.

I personally am recently working with Python for my ETL processes, including MapReduce, and anlysing data, which works awesome in combination with IPython as interactive development tool.

Apache Spark: The Next Big (Data) Thing?

Since Apache Spark became a Top Level Project at Apache almost a year ago, it has seen some wide coverage and adoption in the industry. Due to its promise of being faster than Hadoop MapReduce, about 100x in memory and 10x on disk, it seems like a real alternative to doing pure MapReduce.
Written in Scala, it provides the ability to write applications fast in Java, Python and Scala, and the syntax isn’t that hard to learn. There are even tools available for using SQL (Spark SQL), Machine Learning (MLib) interoperating with Pythons Numpy, graphics and streaming. This makes Spark to a real good alternative for big data processing.
Another feature of Apache Spark is, that it runs everywhere. On top of Hadoop, standalone, in the cloud and can easily access diverse data stores, such as HDFS, Amazon S3, Cassandra, HBase.

The easy integration into Amazon Web Services is what makes it attractive to me, since I am using this already. I also like the Python integration, because latelly, that became my favourite language for data manipulation and machine learning.

Besides the official parts of Spark mentioned above, there are also some really nice external packages, that for example integrate Spark with tools such as PIG, Amazon Redshift, some machine learning algorithms, and so on.

Given the promised speed gain, the ease of use and the full range of tools available, and the integration in third party programms, such as Tableau or MicroStrategy, Spark seems to look into a bright future.

The inventors of Apache Spark also founded a company called databricks, which offers professional services around Spark.

This website is using Google Analytics. Please click here if you want to opt-out. Click here to opt-out.

By continuing to use the site, you agree to the use of cookies. more information

The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this.