Code camp Coding Distributed Systems First programming job Getting Started in Coding JAVA LinkedIn Scala Software Engineering Spark Uncategorized Web

From Junior to Senior: Software Engineering Must-Knows

* This is a living document and will be update over time*

Why these Resources?

Along a software developer’s journey from post-grad to seasoned vet, you come across articles and literature that enlighten you, propelling your skills forward by miles rather than inches. This is a collection of those essential resources that I feel a software engineer should know to be an informed, efficient, and effective engineer.


  1. Maintaining Clean Code
  2. Database Design
  3. Lean Engineering
  4. Testing
  5. Technical Decision Making
  6. Managing Deployments
  7. Container Orchestration
  8. JVM
    1. JVM Tuning
    2. Scala
  9. Machine Learning


1. Maintaining Clean Code

Clean Code (Book by Robert Martin)

“Clean Code” is one of those books that after reading it, you come out with an immediate feeling of both excitement (You know how to write maintainable code now!), and regret (you realize the code you have been writing your whole life is smelly!). While a few chapters are pretty dated technically, it successfully outlines sound practices to maintain hygienic object oriented codebases that can be borrowed for other programming paradigms. This book is a must-know!

Dependency Injection (DI)/Inversion of Control (IoC)

2. Database Design


Normalization is easy to avoid early on, but tough to ignore its effects later down the road. When designing databases, five extra minutes spent thinking about and adhering to normalization will save days, if not weeks, later on in redesign and data integrity issue resolution. Trust me.
Short walkthrough on Normalization:

3. Lean Engineering

Implementing Lean Software Development (Book by the Mary and Tom Poppendieck)

4. Testing

Testing Quadrants

Those needing to prune, or cherry pick certain testing practices into their operations, can benefit from the diagram “Agile Testing Quadrants”. It outlines each test type’s organizational boundaries, initiation mechanism, and outcomes.

Is Unit Testing Worth it?

Chances are you eventually started work at a company whose culture had a baked-in focus on quality, where you set off following orders to test, then realized the benefits later. For some, you are one of the testing thought-leaders at your organization and have to sell the benefit! This article gives you the points that express why unit testing is more than a nicety.

Testing in a Microservices Architecture

5. Technical Decision Making

Building Consensus Before Commitment

Encroaching on the famed “How to win friends and influence people” genre, this article explains how and why you should take a holistic approach to presentations and multi-org affecting decisions.

Technology Radar

A must in every developers exposure toolkit. The Thoughtworks team hand curates languages, frameworks, and practices organizations should adopt, trial, and assess.

Site Reliability Engineering Learnings

6. Managing Deployments

Continuous Integration

Git Workflows

Terraform Up-and-Running

While not critical to know intimately, Terraform is an amazing option as a multi PAAS hosting framework and Infra as Code management tool.

7. Container Orchestration

Kubernetes vs ECS

While this article will quickly grow stale, it is a great comparison of two of the leaders in cloud container orchestration and hosting.

8. JVM


Class and Package Naming Strategies

While we all like to think we always execute the best file and class packaging practices, this naming and scoping refresher from Nikita Volkov can keep you sharp!

Scala Interview Questions

Effective Scala


Extensive Learnings from JVM Performance Tuning

Profiling with VisualVM

This tool is awesome for investigating how JAVA options affect performance, and getting a feel for your apps overall health.


9. Machine Learning

10 Algorithms Software Engineers must know

Disclaimer on References

The resources in this list are intended to be self referencing and imply the original authors are the ones that are due an immense amount of credit.

Think a resource should be added to this article? Please submit it here:

Code camp Coding Distributed Systems First programming job Getting Started in Coding JAVA LinkedIn reddit Scala Software Engineering Spark Uncategorized Web

Why Spark can’t foldLeft: Monoids and Associativity.

cover (3)
Apache Spark is the the elephant in a room full of data processing engines, yet Spark does not supply a foldLeft() or foldRight() method on its RDD class. Strange right? Such a fundamental collection method. How could it be forgotten? Or, was this not an accident?


scoreAverageByPlayer(), which would take an RDD, and return an RDD of tuples of each player with their average score. Note: foldLeft() is not an available method on the scores RDD class.

Remembering associativity

Lets dig deeper, because the realization of the answer is more useful than the answer itself.

Back to Algebra class we go! Associativity is one of the many algebraic properties defining functional mathematics and therefore functional programming. Without its understanding we cannot truly appreciate parallelism in computing and its limitations.

Mathematical Associativity:
"When the order in which the operations are performed does not matter as long as the sequence of the operands is not changed. That is, rearranging the parentheses in such an expression will not change its value."
The following expressions are associative:


Even though the parentheses were rearranged in the equation for res2, the values of res1 and res2 remained equivalent. It can then be said that the act of addition of real numbers is an associative operation.

How Spark achieves Parallelism

In order for Spark to become a leader in computational speed, it needed to incorporate operational parallelism. Parallelism will ultimately be the reason foldLeft is not found on the RDD class.


At a high level, Spark clusters computational “worker” nodes or machines, partitions the data to be computed on in the master, distributes the data partitions from the master to the worker nodes where the computations are done on each node’s respective shard of data, then aggregates the resulting dataset(s) on the master node.


You can force Spark to parallelize computation on an RDD by using parallelize() on a SparkContext.

val scores = Array(68, 71, 73)
val parScores = sc.parallelize(scores)

Below is a function f being applied to an input dataset concurrently on a spark cluster. This can be thought of as a map transformation.

Parallelization Visualized


Parallelizing reduce() in Spark

Let’s look at how spark parallelizes the reduce operation on an RDD.

reduce() from the Spark Documentation

Action Meaning
reduce(func) Aggregate the elements of the dataset using a function func (which takes two arguments and returns one). The function should be commutative and associative so that it can be computed correctly in parallel.

“The function should be commutative and associative so that it can be computed correctly in parallel.”

Signature of reduce()

def reduce[A](op: (A, A) => A): A

This reads:

Execute the function “op” on each element (type A), with the result of the previous op computation (accumulator of type A) and respective element (type A) as inputs, eventually returning the resulting accumulator value from the last iteration (type A).

Spark’s reduce() in action

Now let’s say we have a set of “score” integers and want to determine the lowest score. We can execute a reduce action on the RDD with a monoid findMin() (more on monoids later) as an operational parameter to solve this.

In code we would solve this like:

val data = Array(1, 2, 3, 4, 5)
val distData = sc.parallelize(data)
def findMin(first: Int, second: Int): Int = first.min(second)

val min = distData.reduce(findMin)

This would evaluate in Spark as:



We know what parallelism looks like in Spark, but why can’t we use foldLeft()? This will come together, but we need to understand Monoids first.

The laws surrounding Monoids are tightly coupled to associativity and state a Monoid operation:

  • Is of some type A
  • Consists of an operation, op, taking two values of type A, combining them into a single: op(op(x,y), z) == op(x, op(y,z)), where the type of x, y and z is A
  • Has an identity for the operation that maintains: op(x, zero) == x and op(zero, x) == x for any x of type A

Balanced Folds

reduce() can be categorized as a balanced fold, or a fold that allows for parallelism. Compare the following for the Sequence (a, b, c, d).

foldLeft() with the operation op would look like:

op(op(op(a, b), c), d)

While a balancedFold looks like:

op(op(a, b), op(c, d))

Can you see how this operation could be parallelized with a fork-join data structure?

Thinking of reduce() as a Balanced Fold

If we look back at the reduce of the findMin operation, its operational execution looks like:

Parallelizing reduce - Visualized

This is a balanced fold! Which can be written as either:





foldLeft/right are methods made available on many monadic collections. However, let’s focus on the List collection which provides the following signature and implementation of foldLeft():

def foldLeft[B](z: B)(op: (B, A) => B): B
  if (this.isEmpty) z
  else op(head, tail.foldRight(z)(op))

This reads:

foldLeft is of type B, takes an initial element z, performs the operation op on each element in the Traversable object, returning a Type A in each accumulator iteration, and eventually returns a type B

foldLeft()’s operation is NOT a Monoid

We know that foldLeft’s predicate operation has a non-Monoidal signature, as it breaks all three monoid laws, but why again is this not a transformation supported by Spark? Simply put, it’s because foldLeft is not sufficiently parallelizable!

Looking back at the third Monoid law, it states the following must be true:

op(op(x,y), z) == op(x, op(y,z))

This law is what drives the ability to parallelize. Spark can fork a monoidal operation across a dataset into n number of operations and join the resulting values within the master. This fork-join parallelization results in a best-case decrease in execution time by a factor of n.

foldLeft() cannot be Parallelized

If we pretend foldLeft was a transformation available on Spark collections, and visually walked through its execution, maybe we could more easily understand it’s limitations.

Given the following collection and transformation lets see it in action…

val nums = List(2.2, 3.3, 4.4)
nums.foldLeft(1)((agg, next) => (agg * next).toInt)

Parallelizing foldLeft - Visualized

We can see why foldLeft() was never implemented within Spark, as the fork-join execution model would still result in serial blocking of computation!

JAVA Scala Web Zookeeper

Zookeeper in AWS: Practices for High Availability with Exhibitor

Untitled drawing (1)


Zookeeper is a distributed sequentially consistent system developed to attack the many tough use cases surrounding distributed systems such as leader election in a cluster, configuration, and distributed locking. For more Zookeeper recipes visit: Zookeeper clusters(ensembles) can be made of of any number of nodes, but typically take the form of a three or five node ensemble with the minority of nodes able to fail and the Zookeeper service able to continue serving traffic.

How we use Zookeeper

On the Search team at Careerbuilder we use Zookeeper as a configuration service replacing static configuration file deployment. We also have plans to move to Solr Cloud, which requires a Zookeeper Ensemble for its election and configuration tasks. Below is a system diagram for the entire Zookeeper/Exhibitor service in AWS with an auto-scaling group and consuming client libraries.


Difficulties with Zookeeper

While Zookeeper provides a high level of reliability and availability through redundancy and leader election patterns, it’s critical use cases introduce a high level of risk for outages and mass failures. To mitigate a lot of the risk, the open source Exhibitor application should be used as a supervisory process to monitor and restart each Zookeeper process, as well as provide data backups, recovery, and auto ensemble configuration. Exhibitor will be explained in more detail later.



  • Deployment will take advantage of AWS resources (Currently avoiding containerization)
  • Exhibitor will be deployed alongside Zookeeper to server as it’s supervisory process and management UI (


Infrastructure as Code: Frameworks

Infrastructure automation can be achieved through many open and closed source framework providers such as Ansible, Puppet, and Chef. For deployments on the Search team at Careerbuilder we use Chef as the means for setting up environments on self hosted boxes in AWS. The Chef + Ruby combo is a robust tool in managing infrastructure as code and I highly suggest it:


To deploy a Zookeeper ensemble the following artifacts and dependencies are required to be installed on each node in the cluster. After each artifact is built they can be uploaded to an artifact repository, but in our case an S3 bucket through a continuous integration build, to be later downloaded and installed by Chef during deployment.

  • JAVA

    • Zookeeper and Exhibitor are JAVA Virtual Machine(JVM) applications therefore JAVA must be installed on all machines running them.
  • Zookeeper

  • Exhibitor

    • Exhibitor can be run as a WAR file or with an embedded Jetty Server. After finding the Tomcat/WAR based server cumbersome we chose the Jetty based Exhibitor moving forward and it has paid off through it’s ease of configuration.
    • Steps on how to build the Exhibitor artifact can be found on Github:

Cloud Formation

On a higher level of Infrastructure as code lies the ability to orchestrate resource allocation, bringup, and system integration. AWS CloudFormation is a means to perform these deployment actions internal to AWS.

Generic Templates

By generalizing CloudFormation template files for your specific use cases and parameterizing them, you can use a homegrown application or Integration processes to regularly phoenix or A-B deploy your resources with ease.

Here at Careerbuilder we have developed an in-house JAVA application known as Nimbus that pulls our generic CF Stack templates from Github, populates their parameters with values from parameter files and triggers a CloudFormation Stack creation in AWS. This abstracts a lot of CloudFormation’s unused Stack template complexity.

Cluster Discovery through Instance Tagging

In many cases it is required that you define machine instances indexes to each machine in a clustered distributed system either due to dataset sharding or static clustering. It is possible to deploy Zookeeper with exhibitor in a static ensemble that does not utilize a shared S3 configuration file(more on this later) and this would require instance tagging. Each node could be tagged with an attribute, say {“application”=”zookeeper-development”} by CloudFormation to be used as a selector during bring up. The Chef/Ansible/Puppet processes running on each node could then acquire the IP’s of each of the nodes in the ensemble by going through a query-check-sleep cycle in the local chef script until the correct number of Zookeeper boxes are up and running. However, this deployment model is clumsy and error prone, therefore the shared configuration file in S3 for exhibitor is highly suggested, as server IPs are added and removed from the config file dynamically.

Starting Exhibitor and Zookeeper

Exhibitor will start and restart the Zookeeper processes periodically during clustering tasks and rolling configuration changes. Therefore, all you need to do is start Exhibitor on each node, each node will starte Zookeeper, then register itself with the shared config file specified by the “–s3config” parameter. Below is some example ruby/chef code that can be used to start Exhibitor:

execute “Start Exhibitor” do
commandnohup java -jar #{node[:search][:exhibitor_dir]}/exhibitor.jar –hostname #{node[‘ipaddress’]} –configtype s3 –s3config #{bucket} –s3backup true > /opt/search/zookeeper/exhibitorNohup.log 2> /opt/search/zookeeper/exhibitorNohup.error.log &
action :run

Since Exhibitor manages the Zookeeper process it is important not to create any configuration files manually nor expect any manual configurations made to Zookeeper to persists after deployment or during runtime.



Securing Exhibitor/Zookeeper is an extensive topic left to the specific implementation of the reader. However, the Exhibitor Wiki lists command parameters that can be used to enable and configure security features within Exhibitor and suggest giving it a look.


A typical exhibitor bucket root will look like the following, with an config file that you have uploaded containing a default level of configuration. I suggest also maintaining a ‘last known default’ config file in a “base-config” directory for backup of the file, in the event it gets corrupted.


An example file:\=5&tickTime\=2000&initLimit\=10

Pro Tip: A great benefit we have found is the ability to modify Zookeeper Java Environment settings through Exhibitor for performance and Garbage Collection tuning.

Ensemble Registration

The following images outline the steps of the consensus process during Ensemble deployment.


Self Healing

The following images outline the event of a node failure in an auto-managed ensemble.