Understand how Google and other search engines use LLMs to gain insights into the semantic meaning of the language in search queries using embedding and cosine similarity.
The pandemic response required a remarkable level of collaboration between and beyond government departments. In this blog post, I’m going to look at the Clinically Extremely Vulnerable People Service, outlining the different areas of collaboration upon which the service depended, and reflecting on the lessons that government can take forward to achieve its vision of a responsible, efficient and effective data ecosystem.
Data-literate leadership underpinned the most successful pandemic-response programmes. In this post, I explore what data-literate leadership looks like, drawing on examples from the roundtables on data sharing in government that the Institute for Government ran in partnership with Scott Logic.
A few years ago while working on a digital product in a government department, my team learnt a valuable lesson: rules can help you go faster. In this post, I explain the positive difference that regulatory and legislative frameworks can make to the design and delivery of digital services, with some examples from the government's response to the pandemic.
Over the summer, in partnership with Scott Logic, the Institute for Government (IfG) ran a series of roundtable discussions with senior civil servants and government experts on the topic of Data Sharing in Government. This is the first in a series of blog posts in which I'll share some reflections on key themes that arose.
As a keen amateur runner that somehow found themselves with a qualifying time to stand on the start line of the Great North Run with the Elite women, I take a look at the main ways data has aided my journey to that start line.
Highlights from one day training course on performance tuning with Apache Spark. Delving into the five most common reasons for poor performance.
The UK Government has an ambitious data strategy that aims to encourage and facilitate data sharing between departments and businesses. Elements of the strategy appear relatively straightforward, but how will the government fully realise the potential, and align citizens with this bold new approach?
Organisations across the globe have been on a journey to find the optimal approach for managing and leveraging analytics data. In this post, I’ll set out each of the key milestones on the journey, to arrive at the latest milestone – the Data Mesh paradigm – and ask whether it is really a thing.
This blog is about tools that help address the challenge of testing systems which handle large data volumes. We’ll see why creating a large, realistic and valid test data set is hard, how test data generators can help, and compare some of those available.
Create your own Elasticsearch cluster in cloud in next to no time. Leverage ElasticHQ and CloudWatch logging to gain transparency. Excerpts from a client project.
What started as one faulty gas reading in the summer of 2017, ended up as a series of wasted calls where my bill kept getting higher and higher until it reached £11k. How could this have been handled faster and left me without considering moving energy provider.
Distributed stream processing engines have been on the rise in the last few years, first Hadoop became popular as a batch processing engine, then focus shifted towards stream processing engines. Stream processing engines can make the job of processing data that comes in via a stream easier than ever before.
Distributed log technologies have matured in the last few years. In this article, I review the attributes of distributed log technologies, compare four of the most popular and suggest how to choose the one that's right for you.
Apache Spark is the major talking point in Big Data pipelines, boasting performance 10-100x faster than comparable tools. But how achievable are these speeds and what can you do to avoid memory errors? In this blog I will use a real example to introduce two mechanisms of data movement within Spark and demonstrate how they form the cornerstone of performance.
Spark is well known in Big Data for its incredible performance and expressive API. However, it just takes one small misstep to transform a massively parallel powerhouse into a pathetically poor performer. This post presents an example and the steps that can be taken to indentify the problem.
In this quick look at the R language and tools I'll look briefly at the syntax of the language and have a go at creating a few charts with a data set.
Apache Kafka provides distributed log store used by increasing numbers of companies and often forming the heart of systems processing huge amounts of data. This post shows how to use it for storing meteorological data and displaying this in a graphical dashboard with Graphite and Grafana
A discussion about Cassandra consistency levels and replication factor, which are frequently misunderstood. This post explains the Cassandra infrastructure and how its configuration can be tuned.
Lichess makes over 100GB of chess games from 2017 available on their website. This post shows how this data can be transformed with Apache Spark and analysed. Something for Data Engineers and Chess Enthusiasts alike!
Yesterday the Financial Times boldly declared that BP saved $7bn since 2014 by investing in Big Data technologies. I spent a couple of hours researching Big Data technologies associated with BP members of staff to try and build up a picture of exactly which technologies they are using.
Using microservices in your architecture is a very popular choice. Unfortunately it is also challenging to get it right. With the help of Twelve-Factor methodology, I will tell you how to set yourself up for a success rather than a disappointment.
A successful attempt of load testing Alteryx API with Gatling and a not-so-successful attempt with Apache JMeter
In this post we compare how Cassandra and MariaDB can be configured to operate in clusters and how this affects response time for queries. We found Cassandra to scale well and to be highly configurable. MariaDB can be used with Galera Cluster but it does not provide horizontal scaling. Also NDB can be used to scale MySQL but it was not as configurable as Cassandra.
We've been comparing Cassandra and MariaDB in single node setups, exploring the issues of each in terms of performance and ease of use from a development perspective. In this article we explore the issues at play in such a setup such as the differences in queries, speed of response and the features that seperate these two technologies.
Docker 1.13 introduces a simple way of providing secrets to containers
StreamSets Data Collector (SDC) is an open source tool for stream-based extracting, transforming and loading large quantities of data. It provides an easy to use UI on top of the underlying processing power of YARN and Spark Streaming with a large number of installable integrations with source and destination systems.
With the advent of the Internet of Things, the world of Big Data couldn't be more relevant. This post gives an overview of technologies that achieve processing at scale and in real time.
Big Data can help businesses run more efficiently. Their main challenge is getting the best value from the data they have to turn it into actionable information
The popularity of Spring Boot in the Java world is undeniable. In this post I will show you yet another reason for this. Using Spring Boot makes working with MongoDB an absolute pleasure.
In this post I describe how to use Elastic's Rally to generate benchmarks for your private Elasticsearch queries and clusters. I'll be creating a benchmark which allows comparison of an unscored query with one where scoring is enabled.
This post demonstrates how Docker 1.12 swarm mode round robins the containers in a service both for incoming connections (ingress) and DNS within the swarm.
For the last few months we've been working on a very DevOps focused project. As such we've used AWS, infrastructure as code, Docker and microservices. The different microservices were initially running all on one box, each with a different port. This solution wasn't scalable or very practical. We couldn't have all our services on one machine and it was getting tiresome and error prone having to remember/lookup which port each service was on. We needed our services to run on separate machines, and we needed a way to communicate with them without having to hard-code IP addresses or port numbers. What we needed was service discovery. As we had already been using Docker for each service, Docker Swarm was a natural candidate.
This is the second blog post orientated around Bitcoin and its inner workings. The first post took the blockchain and broke down the algorithms which create the fundamental structure of any cryptocurrency. The post was separated into two sections; the first focusing on the block header and the second focusing on the construction of a transaction. If you are not comfortable with how the blockchain works, I suggest you read the first blog post before continuing.
In most microservice architectures, there are many opportunities and temptations for sharing code. In this post I will give advice based on my experience on when it should be avoided and when code reuse is acceptable. The points will be illustrated with the help of an example Spring Boot project.
An experiment in writing a volume plugin for Docker
An insight into the ELK stack and how we used it on a big data project
This post uses Docker Compose to spin up a three container HTTP server. One container services the HTTP requests and delegates work to two other containers in a load-balanced way. Erlang is used for development to add a bit of extra challenge!
Apache Spark has quickly become the largest open source project in big data, but why has it suddenly got so much momentum behind it?
Big data is one of those buzz phrases that gets thrown round a lot, companies love saying they work with ‘Big’ data, but what is ‘Big’ data?
Welcome to my blog, Andrew's thoughts on Big Data. This page gives a little background on myself and the blog.
This post demonstrates how to create an efficient stock ticker app using HTML5 WebSockets and a Haskell server.
Sharded clusters enable the data persistence layer in MongoDB to be shared across several machines. In this post, we will look at the key considerations you should make before you use sharded clusters.
Big Data is a hot topic these days, and one aspect of that problem space is processing streams of high velocity data in near-real time. Here we're going to look at using Big Data-style techniques in Scala on a stream of data from a WebSocket.
With non-relational database implementations (key-store, graph, etc.) entering the mainstream, the necessity has arisen to synchronise relational databases to their non-relational cousins.