24 Nov 2019 •
Mesos is a framework I have had recent acquaintance with. We use it to manage resources for our Spark workloads. The other resource management framework for Spark I have prior experience with is Hadoop YARN. In this article, I revisit the concept of cluster resource-management in general, and explain higher-level Mesos abstractions & concepts. To this end, I borrow heavily the classification of cluster resource-management systems from the Omega paper.
The Omega system is considered one of the precusors to Kubernetes. There is a fine article in ACM Queue describing this history. Also, Brian Grant has some rare insights into the evolution of cluster managers in Google from Omega to Kubernetes in multiple tweet-storms, such as this and this.
.. Read More
16 Nov 2019 •
Do not attribute to malice that which can be explained by the less criminal motives of ignorance and lethargy.
An aphorism of utmost utility in my life is the Hanlon’s Razor. I find it a liberating rule of thumb to weigh a lot of unavoidably unpleasant experiences in daily life. In a less formal & more terse form that I prefer, it reads:
Stupid people abound; Malicious people, less so.
There is a neat wikipedia article on it which focuses on its origin, and also introduced me to an earlier form of the aphorism by Goethe.
Misunderstandings and lethargy perhaps produce more wrong in the world than deceit and malice do. At least the latter two are certainly rarer.
Johann Wolfgang von Goethe, in The Sorrows of Young Werther
.. Read More
11 Oct 2019 •
A common anti-pattern in Spark workloads is the use of an
or operator as part of a
join. An example of this goes as follows:
val resultDF = dataframe
.join(anotherDF, $"cID" === $"customerID" || $"cID" === $"contactID",
This looks straight-forward. The use of an
or within the join makes its semantics easy to understand. However, we should be aware of the pitfalls of such an approach.
The declarative SQL above is resolved within Spark into a physical plan which determines how this particular query gets executed. To view the query plan for the computation, we could do:
.. Read More
/* pass true if you are interested in the logical plan of the query as well */
14 Sep 2019 •
Integration of Large-Scale Data Processing Systems and Traditional Parallel Database Technology
Abouzied, A., Abadi, D.J, Bajda-Pawlikowski, K., Silberschatz, A. (2019, August). Proceedings of the VLDB Vol. 12 (12).
HadoopDB was a prototype built in 2009 as a hybrid SQL system with the features from Hadoop MapReduce framework and parallel database management systems (Greenplum, Vertica, etc). This paper revisits the design choices for HadoopDB, and investigates its legacy in existing data systems. I felt it is a great review paper for the state of modern data analysis systems.
MapReduce is the most famous example in a class of systems which partition large amounts of data over multitude of machines, and provide a straightforward language in which to express complex transformations and analyses. The key feature of these systems is how they abstract out fault-tolerance and partitioning from the user.
MapReduce, along with other large-scale data processing systems such as Microsoft’s Dryad/LINQ project, were originally designed for processing unstructured data.
The success of these systems in processing unstructured data led to a natural desire to also use them for processing structured data. However, the final result was a major step backward relative to the decades of research in parallel database systems that provide similar capabilities of parallel query processing over structured data.
The MapReduce model of
.. Read More
Map -> Shuffle -> Reduce/Aggregate -> Materialize is inefficient for parallel structured query processing.
13 Sep 2019 •
This talk is an introduction to Datomic, by its creator Rich Hickey. My notes on this talk are linked below: