The Virtual World

 

I don't have any statistics to back up this claim but given the prevalence of cloud native approaches I am willing to bet that there is more virtualised hardware in the world than real physical computers.

We take for granted that we can bring up a virtual machine within a few clicks, use it for a vast variety of different workloads and then spin it back down again. Servers have gone from things that need to be carefully maintained and looked after to an ephemeral resource we can create and throw away.

The technology unpinning this ability is the hypervisor. A piece of software that provides an abstraction of physical resources such as CPU and memory to allow the creation of virtual machines. This allows a powerful host server to be utilised to provide a large number of isolated guest machines, increasing efficiency and productivity.

History of Virtualisation

Virtualisation had previously existed to the extent that multiple different software applications could run on the same hardware concurrently. In the late 1960s IBM developed a research tool called SIMMON that took this a step further allowing hardware resources to also be virtualised on mainframe computers. This was soon extended to cover operating system resources such as kernel tasks such that the idea of having virtual machines built on top of real hardware was born.

For many years this virtualisation was the preserve of mainframe systems, until in around 2005 the first attempts at virtualisation for x86 systems started to gain momentum. 

Originally the hypervisors being developed were complex and prone to relatively slow performance, but as technology advanced the level of virtualisation that could support the cloud services we now take for granted started to emerge.

Types of Hypervisor

Hypervisors can be broadly categorised into two types, type one and type two.

Type one hypervisors run directly on the host machines hardware and therefore eliminate the need for an underlying operating system. For this reason they are often referred to as native or bare metal hypervisors.

Type one hypervisors are very efficient and often more secure, they are typically used within data centres on powerful servers hosting a large number of virtual machines.

Type two hypervisors run on the host machines operating system in the same way as any other application, they provide an abstraction of the host operating system to create an isolated process the host can interact with. For this reason they are often referred to as host hypervisors.

Type two hypervisors are less efficient than type one because the host operating system prevents them from having direct control of the underlying hardware. However they are a more practical option when virtualisation is needed on a machine that isn't a server.

Benefits of Virtualisation

Aside from the ability to turn one server into multiple virtual machines, what other benefits does virtualisation bring?

As we've already touched on virtualisation can greatly increase efficiency. Rather than having multiple applications each being hosted on their own server, potentially not making full use of the available resources. All applications can be hosted on their own virtual machine on the same server.

The ability for virtual machines to be created quickly and automatically also provides an element of scalability that doesn't exist if new physical servers need to be added to a farm. The fact the underlying system is virtualised also provides an element of portability, the application can be hosted on any physical server capable of running a virtual machine to the same specification.

Coupled to this idea of portability is also the concept of snapshots and failure recovery. Since the environment an application is running in is software controlled it allows the current state of the virtualised hardware to be recorded into a snapshot which can be used to deal with failure by returning the system to a known good state.

Virtualisation also provides the ability to create crafted environments for legacy systems. Where an older application may have requirements that are no longer compatible with modern hardware, virtualisation offers a means to create such an environment whilst still utilising a cloud based architecture.

We sometimes take cloud computing for granted without thinking about the technology that underpins it. For the majority of applications it isn't really necessary to have a knowledge of how hypervisors are enabling them to be deployed into a crowd environment. But I still think it can helpful in us being well rounded engineers to at least have knowledge of the layers in our stack, what they provide and their limitations and strengths. 

Distributing Problems

 

Patterns and practices in software engineering can be very cyclical, the names and terminology applied can change but we often go back to old ideas that had previously been dismissed.

I think this is indicative of the fact that many problems in software development don't have perfect answers. It is often a balance of pro's and con's and sometimes the weighting applied to them varies over time and so we flip between different competing approaches.

One of these areas relates to whether it is better to centralise or distribute. Should our applications be monolithic in nature or a collection of loosely coupled distributed parts. For a long time the argument as been seen to be won by distributed computing and the microservices approach. However in recent times the monolith approach has started to be seen as not always being the wrong idea.

This article shouldn't be seen as an argument against a distributed approach, it should be viewed more as an argument against premature optimisation. By understanding the drawbacks of distribution you can make a better judgement on whether it's the right approach for your application right now.

Interconnected Services

Distributed computing is a relatively broad term, but within the context of this article we are taking it to mean the pattern of dividing an application up into a collection of microservices.

Usually microservices are built to align to a single business domain or process with the application being the sum of these parts communicating via a lightweight protocol such as RESTful APIs, or increasingly via an event driven architecture.

You can see from this that the term microservice is quite loosely defined, a lot of the issues that are created when applying this approach can often be traced back to the fact that defining the division between microservices and their relative size is quite a hard problem.

The best explanation I've seen for this is that a microservice should be as easy to replace as to refactor, basically meaning microservices shouldn't be so large as to negate the option of starting again with their design. 

I think this idea is much easier to apply when starting with a blank sheet of paper. When splitting up an existing application it is often more pragmatic to not subdivide too quickly, as further splitting an existing service is often easier than trying to coalesce several services back into one once they've been split.

Fallacies of Distributed Computing

In 1994 L. Peter Deutsch at Sun Microsystems devised a list of seven fallacies of distributed computing building on top of the work of Bill Joy and Dave Lyon.

The fallacies represent seven assumptions that often cause the architecture and development of a distributed system to head in the wrong direction. 

The first is that the Network is Reliable, this often leads to services not being written with network related error handling in mind, meaning when network errors do occur services often stall and become stuck consuming resources while waiting for a response that isn't forthcoming.

The second and third are related in that Latency is Zero and Bandwidth is Infinite can both cause developers to give little thought to the nature of the data that is propagating through the network. 

Number four is that the Network is Secure, which can lead to a complacency where possible intrusion from malicious insiders isn't considered.

Number five is that Network Topology Doesn't Change, which in a similar way to two and three are indicative of us not thinking about the fact that the network our applications operate in is a dynamic element in the same way as our code.

Number six is There is One Administrator, this can cause us to fail recognise inconsistent or contradictory policies around network traffic and routing.

Number seven is that Transport Cost is Zero, here we need to factor into our thinking that an API call and the resultant transfer of data has a cost in terms transmission time.

Strategies to Compensate

The fallacies described in the previous section shouldn't be seen as arguments for why distributed systems shouldn't be built, they are things that should be considered when they are.

We can often think that our software is deployed into an homogeneous environment with perfect conditions, but this is often not the case.

Errors in transport can occur so we should have an affective strategy to detect these errors, retry calls if we believe this may lead to a successful outcome, but also have a strategy for when calls continue to fail such as a circuit breaker to avoid filling the network with requests that are unlikely to received the desired response.

We must realise that as load increases on our system the size of the data we are passing between elements may start to be a factor in their performance. Even if each individual request/response is not large in sufficient quantities their impact may be felt.

We have to maintain a healthy distrust of the elements of our system we are interacting with. A zero trust approach means we do not inherently trust any element simply because it is inside the network, all elements must properly authenticate and be authorized.

We must also consider that when we subdivide our system into more elements that we are introducing a cost in the fact those elements will need to communicate, this cost must be balanced with the benefit the change in architecture would bring.

These are only some of the things we need to think about with a distributed approach. This post is too short to cover them in great detail, but the main takeaway should be that a distributed approach isn't cost free and sometimes it might not offer advantages over a monolithic approach. Getting a distributed approach right is hard and not an exact science, many things need to be considered and some missteps along the way should be expected.

As with any engineering decision its not right or wrong its a grey area where pro's and con's must be balanced.

Solid State World

 

It's often said that technology moves quickly, I'm actually of the opinion that its our learning of how to utilise technology that actually grows rapidly.

The invention of certain pieces of technology are transformative, so much so that in the initial stages we struggle to grasp its potential. Then over time the realisation of whats possible grows and more and more applications materialise.

The World Wide Web was one of those technological jumps forward, and Artificial Intelligence will undoubtedly, if it isn't already, be another notable point in history. However I believe the true giant leap forward is often not talked about as much as it should be since it underpins everything that has come since and kick started a revolution that has lasted 75+ years.

That giant leap forward was the invention of solid state electronics.

Triodes to Transistors

The early building block of electronics was the vacuum tube triode. These devices had applications across radar, radio, television and many other spheres. However they were large and power angry devices making it difficult to use them reliably in increasingly complex devices.

The key to the miniaturisation of electronics was the invention of the field effect transistor. The theory behind these devices was first formulated in the 1920s, but it wasn't until the late 1940s that the first practical examples were built.

Many different scientists independently discovered and worked on the transistor concept, however William Shockley, John Bardeen and Walter Brattain are widely credited with its invention. In the late 1950s at Bell Laboratories they first produced the Metal Oxide Semiconductor Field Effect Transistor (MOSFET).

These transistors had many applications, once of which was being able to act as a "digital" switch, with this the era of the semiconductor and solid state electronics was born.

Printing with Light

Once the MOSFET had been invented the next challenge related to developing techniques for reliably being able to manufacturer them whilst also being able to continue their miniaturisation. The ability to produce transistors at a smaller size would enable them to be more densely packed as well as reducing power consumption.

The idea of photolithography, using light to print patterns into materials, had been around for some time. But in the late 1950s Jules Andrus at Bell Laboratories looked to use similar techniques to build solid state electronics (Moe Abramson and Stanislaus Danko of the US Army Signal Corps are also credited with inventing similar techniques in the same period).

Using this technique a semiconductor substrate is covered with a material called photoresist. A mask is then placed over the top such that only certain areas of the material are exposed to a light source. The exposed areas go through a chemical change that renders them either soluble or insoluble in a developer solution depending on the type of photoresist that is used. Finally the material then goes through an etching process to leave the desired pattern of the semiconductor based components.

Typically now Ultra Violet light is used in the process, but much of the drive for ever smaller and smaller transistors was driven by using ever decreasing wavelengths of light in the photolithography process.

Moore's Law

Gordon Moore was an American engineer who originally worked at Schokley Semiconductor Laboratory, a company founded by William Schokley the co-inventor of the transistor. He would latter go on to be a founder member of both Fairchild Semiconductor and Intel.

In 1965 while working at Fairchild Semiconductor he was asked what he thought might happen to semiconductor technology over the next ten years. His answer to that question would eventually come to be known Moore's Law.

Put simply Moore's Law states that the number of transistors that can fit into a given size of integrated circuit (or "chip") will double roughly every two years.

Originally Moore predicted this would be the case for ten years, however remarkably it has continued to hold for at least 55 years with industry voices only in the last couple of years starting to question if we may now be starting to reach the limit of Moore's original prediction.

The fact the industry has been following Moore's law for so long has been a major driver in continually increasing processing power that has been a catalyst for innovation. It is now believed that the transistor is the most widely manufactured device in human history. In 2018 it was believed as many as 13 sextillion (13 followed by 21 zeros) had been manufactured. 

The lives of virtually every human on the planet has been touched by solid state electronics and the technologies it underpins. It is fair to say that the birth of solid state electronics marked the birth of our technological age and changed the world forever. Of all the transformative technologies that have come since, and will continue to be developed in the future, I don't believe any will be as impactful as this initial giant leap forward.

Network of Networks

 

When we're searching for analogies to describe the operation of the internet we often fall back on that of posting a letter. Each packet of data we need to send and receive can be compared to a parcel we need to get from A to B.

In order to achieve that you need to identify the best route to take between those two points. For parcels these routes are relatively static, for the internet that isn't the case with routes being much more dynamic.

Border Gateway Protocol (BGP) is a path vector routing protocol that makes routing decisions on how to get data packets to their intended destination. As internet users we may not directly interact with it but without it the internet wouldn't be functional across the globe as it is today.

Autonomous Systems

The internet as the name suggests is a network of networks, it isn't one cohesive systems, it is a large number of individual networks. These individual networks in the context of routing traffic are referred to as Autonomous Systems (AS).

If we continue our postal analogy then each AS is like an individual postal region covering a town or county.

Each AS consists of routers that know how to route traffic internally but need to rely on connections to neighbouring AS to route traffic outside of its region. Because the internet is a dynamic system with routes appearing and disappearing each AS needs to be kept up to date with routes each other AS can help with.

This is achieved via peering sessions where AS connect to each other in order to build up the full picture of how to route traffic to destinations outside of each others boundary. BGP is the mechanism that allows this routing information to be exchanged.

Not Just The Shortest

In order to become a AS you must register with the Internet Assigned Numbers Authority (IANA) which will assign the AS to a Regional Internet Registry (RIR) as well as allocating a 16 or 32 bit identifier. As of the late 2010s there were around 64,000 AS registered worldwide a number which will have continued to grow.

AS tend to be managed and run by large organisations, typically Internet Service Providers (ISP) but also large tech companies, governments or large institutions such as universities.

Often there are multiple possible routes for a packet to reach its destination. In order to allow the best route to be chosen BGP allows AS to apply attributes to each route that can be factored into the decision on which route to take.

These attributes may indicate hop count (how many steps are involved in getting to the destination) as well as weight where an AS can indicate which route it would prefer traffic to take. 

Because some AS are managed by commercial business this creates an interesting quirk in how traffic is routed. It would be assumed that the shortest or fastest route would be chosen. But because some companies will charge for their AS to handle traffic or may not want to help competitors then commercial relationships are sometimes factored into which route to take.

When BGP Goes Wrong

As described earlier the management of routing traffic through the internet is autonomous, it is far too complicated and dynamic to be managed by hand. An AS uses BGP to announce which routes it can provide access to, other AS then use this information to route traffic from within inside their network.

There have been several examples of AS accidentally announcing they can provide access to routes which they can't in fact handle. In 2004 a Turkish ISP accidentally announced it could handle any route on the internet, as this misinformation spread to more and more AS internet access ground to a halt across a large part of the globe. In 2008 a Pakistani ISP which was blocking traffic internally to YouTube accidentally started announcing externally it could handle connections to YouTube and caused an outage in the region to that service.

These are both examples of BGP hijacking, in these cases the announcement of false routes was accidental but this isn't always the case. This can sometimes be done maliciously in order to lure traffic meant for a legitimate site to an imposter, in 2018 hackers announced bad BGP routes for traffic being hosted in Amazon AWS and were able to steal a large amount of cryptocurrency.

In an effort to combat this kind of activity Resource Public Key Infrastructure (RPKI) allows BGP data to be cryptographically signed in order to validate that an AS is authorised to announce a route for a particular resource.

The internet is such an important part of our day to day lives its easy to consider it a slick homogeneous system that always works, and it is true to say that its a miracle of engineering that its been able to scale to the network of networks that we use today. However it is actually a very large collection of individual elements and sometime its fragility is exposed and we see it is all too easy for it to falter. 

Underpinning Kubernetes

 

Kubernetes is the de facto choice for deploying containerized applications at scale. Because of that we are all now familiar with its building blocks that allow us to build our applications such as deployments, ingresses, services and pods.

But what is it that underpins these entities and how does Kubernetes manage this infrastructure. The answer lies in the Kubernetes control plane and the nodes it deploys our applications too.

The control plane manages and makes decisions related to the management of the cluster, in this sense it acts as the clusters brain. It also provides an interface to allow us to interact with the cluster for monitoring and management.

The nodes are the work horses of the cluster where infrastructure and applications are deployed and run.

Both the control plane and the nodes comprise a number of elements each with their own role in providing us with an environment in which to run our applications.

Control Plane Components

The control plane is made up of a number of components responsible for the management of the cluster, in general these components run on dedicated infrastructure away from the pods running applications.

The kube-apiserver provides the control plane with a front end via a suite of REST APIs. These APIs are resource based allowing for interactions with the various elements of Kubernetes such as deployments, services and pods.

In order to manage the cluster the control plane needs to be able to store data related to its state, this is provided by etcd in the form of a highly available key value store.

The kube-scheduler is responsible for detecting when new pods are required and allocating a node for them to run on. Many factors are taken into account when allocating a node including resource and affinity requirements, software or hardware restrictions and data locality.

The control plane contains a number of controllers responsible for different aspects of the management of the cluster, these controllers are all managed by the kube-controller-manager. In general each controller is responsible for monitoring and managing one or more resources within the clusters, as an example the Node Controller monitors for and responds to nodes failing. 

By far the most common way of standing up a Kubernetes cluster is via a cloud provider. The cloud-controller-manager provides a bridge between the internal concepts of Kubernetes and the cloud provider specific API that is helping to implement them. An example of this would be the Route Controller responsible for configuring routes in the underlying cloud infrastructure.

Node Components

The node components run on every node in the cluster that are running pods providing the runtime environment for applications.

The kubelet is responsible for receiving PodSpecs and ensuring that the pods and containers it describes are running and healthy on the node.

An important element in being able to run containers is the Container Runtime. This runtime provides the mechanism for the node to act as a host for containers. This includes pulling the images from a container registry as well as managing their lifecycle. Kubernetes supports a number of different runtimes with this being a choice to be made when you are constructing your cluster.

An optional component is the kube-proxy that maintains network rules on the node that plays an important role in implementing the services concept within the cluster.

Add Ons

In order to allow the functionality of a cluster to be extended Kubernetes provides the ability to define Add ons.

Add ons cover many different pieces of functionality.

Some relate to networking by providing internal DNS for the cluster allowing for service discovery, or by providing load balancers to distribute traffic among the clusters nodes. Others relate to the provisioning of storage for use by the application running within the cluster. Another important aspect is security with some add ons allowing for security policies to be applied to clusters resources and applications.

Any add ons you choose to use are installed within the cluster with the above examples by no means being an exhaustive list.

As an application developer deploying code into a cluster you don't necessarily need a working knowledge of how this infrastructure is being underpinned. Bit I'm a believer that having an understanding of the environment where your code will run will help you write better code.

That isn't to say that you need to become an expert, but a working knowledge of the building blocks and the roles they play will help you become a more well rounded engineer and enable you to make better decisions. 

Compiling Knowledge

 

Any software engineer who works with a compiled language will know the almost religious concept of the build. Whether you've broken the build, whether you've declared that it builds on my machine, or whether you've ever felt like you are in a life or death struggle with the compiler. The build is a process that must happen to turn your toil into something useful for users to interact with.

But what is actually happening when your code is being complied? In this post we are certainly not going to take a deep dive into compiler theory, it takes a special breed of engineer to work in that realm, but an understanding of the various processes involved can be helpful on the road to becoming a well rounded developer.

From One to Another

To start at the beginning, why is a compiler needed? Software by the time it runs on a CPU is a set of pretty simple operations knows as the instruction set. These instructions involve simple mathematical and logical operations alongside moving data between registers and areas of memory.

Whilst it is possible to program at this level using assembly language, it would be an impossibly difficult task to write software at scale. As engineers we want to be able to code at a higher level.

Compilers give us the ability to do that by acting as translators, they take the software we've written in a high level language such as C++, C#, Java etc and turn this into a program the CPU can run.

That simple description of what a compiler does belies the complexity of what it takes to achieve that outcome, so it shouldn't be too much of a surprise that implementing it takes several different process and phases.     

Phases

The first phase of compilation is Lexical Analysis, this involves reading the input source code and breaking it up into its constituent parts, usually referred to as tokens. 

The next phase is Syntax Analysis, also know as parsing. This is where the compiler ensures that the tokens representing the input source code conform to the grammer of the programming language. The output of this stage is something called the Abstract Syntax Tree (AST), this represents the structure of the code as described by a series of interconnected nodes in a tree structure representing paths through the code.

Next comes Semantic Analysis, it is at this stage that the compiler checks that the code actually makes sense and obeys the rules of the programming language including its type system. The compiler is checking that variables are declared correctly, that functions are called correctly and any other semantic errors that may exist in the source code. 

Once these analysis phases are complete the compiler can move onto Intermediate Code Generation. At this stage the compiler generates an intermediate representation of what will become the final program that is easier to translate into the machine code the CPU can run.

The compiler will then run an Optimisation stage to apply certain optimisations to the intermediate code to improve overall performance.

Finally the compiler moves onto Code Generation in order to produce the final binary, at this stage the high level language of the input source code has been converted into an executable that can be run on the target CPU.    

Front End, Middle End and Backend

The phases described above are often segregated into front end, middle end and backend. This enables a layered approach to be taken to the architecture of the compiler and allows for a certain degree of independence. This means different teams can work on these areas of the compiler as well as making it possible for different parts of compilers to be re-used and shared.

Front end usually refers to the initial analysis phases and is specific to a particular programming language. Should any of the code fail this analysis errors and warnings will be generated to indicate to the developer which lines of source code are incorrect. In this sense the front end is the most visible part of the compiler that developers will interact with.

The middle end is generally responsible for optimisation. Many compilers will have settings for how aggressive this optimisation is and depending on the target environments may also distinguish between optimising for speed or memory footprint.

The backend represents the final stage where code that can actually run on the CPU is generated.

This layering allows for example for front ends related to different programming languages to be combined with different backends that produce code for particular families of CPUs, with the intermediary representation acting as the glue to bind them together.

As we said at the start of this post understanding exactly how compilers work is a large undertaking. But having any appreciation of the basic architecture and phases will help you deal with those battles you may have when trying to build your software. Compiler messages may sometimes seem unclear or frustrating so this knowledge may save valuable time in figuring out what you need to do to keep the compiler happy.

Terraforming Your World

 

Software Engineers are very good at managing source code. We have developed effective strategies and tools to allow us to branch, version, review, cherry pick and revert changes. It is for this reason we've been keen to try and control all aspects of our engineering environment in the same way.

One such area is the infrastructure on which our code runs.

Infrastructure as Code (IaC) is the process of managing and deploying compute, network and storage resources via files that can be part of the source control process.

Traditionally these resources may have been managed via manual interactions with a portal or front end application from your cloud provider of choice. But manual processes are prone to error and inconsistency making it difficult and time consuming to manage and operate infrastructure in this way.

Environments might not always be created in exactly the same way, and in the event of a problem there is a lack of an effective change log to enable changes to be reverted back to a known good state.

One of the tools that attempts to allow infrastructure to be developed in the same way as software is Terraform by Hashicorp.

Terraform

Terraform allows required infrastructure to be defined in configuration files where the indicated resources are created within the cloud provider via interaction with their APIs.

These interactions are encapsulated via a Provider which defines the resources that can be created and managed within that cloud. These providers are declared within the configuration files and can be pulled in from the Terraform Registry which acts like a package manager for Terraform.

Working with Terraform follows a three stage process.

Firstly in the Write phase the configuration files which describe the required infrastructure are created. These files can span multiple cloud providers and can include anything from a VPC to compute resource to networking infrastructure.

Next comes the Plan phase. Terraform is a state driven tool, it records the current state of your infrastructure and applies the necessary changes based on the updated desired state in the configuration files. As part of the plan phase Terraform creates a plan of the actions that must be taken to move the current state of the infrastructure to match the desired state, whether this be creating, changing or deleting elements. This plan can then be reviewed to ensure it matches with the intention behind the configuration changes.

Finally in the Apply phase Terraform uses the cloud provider APIs, via the associated provider, to ensure the deployed infrastructure aligns with the new state of the configuration files.

Objects

Terraform supports a number of different objects that can be described in the configuration files, presented below is not an exhaustive list but describes some of the more fundamental elements that are used to manage common infrastructure requirements.

Firstly we have Resources which are used to describe any object that should be created in the cloud provider, this could be anything from a compute instance to a DNS record.

A Data Source provides Terraform with access to data defined outside of Terraform. This might be data from pre-existing infrastructure, configuration held outside of Terraform or a database.

Input Variables allow configuration files to be customised to a specific use case increasing their reusability. Output Variables allow the Terraform process to return certain data about the infrastructure that has been created, this might be to further document the infrastructure that is now in place or act as an input to another Terraform process managing separate but connected infrastructure. 

Modules act like libraries in a software development context to allow infrastructure configuration to be packaged and re-used across many different applications.

Syntax

We have made many references in this article to configuration files but what do these actually consist of?

They are defined using the Hashicrop Configuration Language (HCL) and follow a format similar to JSON, in fact it is also possible for Terraform to work with JSON directly.

All content is defined with structures called blocks:

resource "my_provider_object_type" "my_object" {

  some_argument = "abc123"

}

Presented above is an extremely simple example of a block.

Firstly we define the type of the block, in this case a resource. Then comes a series of labels with the number of labels being dependent on the block type. For a resource block two labels are required, the first describing the type of the resource as defined by the provider, the second being a name that can be used to refer to the resource in other areas of the configuration files.

Inside the block the resource may require a series of arguments to be provided to it in order to configure and control the underlying cloud resource that will be created.

This post hasn't been intended to be a deep dive into Terraform, instead I've been trying to stoke your interest in the ways an IaC approach can help you apply the same rigour and process to your infrastructure management as you do to your source code.

Many of the concepts within Terraform have a close alignment to those in software engineering. Using an IaC approach alongside traditional source code management can help foster a DevOps mentality where the team responsible for writing the software can also be responsible for managing the infrastructure it runs on. Not only will this allow their knowledge of the software to shape the creation of the infrastructure but also in reverse knowing where and on what infrastructure their code will run may well allow them to write better software.

Being at the Helm

 

The majority of containerized applications that are being deployed at any reasonable scale will likely be using some flavour of Kubernetes.

As a container orchestration platform Kubernetes allows the deployment of multiple applications to be organised around the concepts of pods, services, ingress, deployments etc defined in YAML configuration files. 

In this post we won't go into detail around these concepts and will assume a familiarity with their purpose and operation.

Whilst Kubernetes makes this process simpler, when it's being used for multiple applications managing the large number of YAML files can come with its own challenges.

This is where Helm comes into the picture. Described as a package manager for Kubernetes Helm provides a way to manage updates to the YAML configuration files and version them to ensure consistency and allow for re-use.

I initially didn't quite understand the notation of Helm being a package manager but as I've used it more I've come to realise why this is how it's described.

Charts and Releases

The Helm architecture consists of two main elements, the client and the library.

The Helm client provides a command line interface (CLI) to indicate what needs to be updated in a cluster via a collection of standard Kubernetes YAML files, the library then contains the functionality to interact with the cluster to make this happen.

The collection of YAML files passed to the client are referred to as Helm Charts, they define the Kubernetes objects such as deployments, ingress, services etc.

The act of the library using these YAML files to update the cluster is referred to as a Release.

So far you maybe thinking that you can achieve the same outcome by applying the same YAML files to Kubernetes directly using the kubectl CLI. Whilst this is true where Helm adds value is where you need to deploy the same application into multiple environments with certain configuration or set-up differences. 

Values, Parametrisation and Repositories

It will be a common practice to need to deploy an application to multiple environments with a differing numbers of instances, servicing requests on different domains or any other  differences between testing and production environments.

Using Kubernetes directly means either maintaining multiple copies of YAML files or having some process to alter them prior to them being applied to the cluster. Both of these approaches have the potential to cause inconsistency and errors.

To avoid this Helm provides a templating engine to allow a single set of YAML files to become parameterised. The syntax of this templating when you first view it can be quite daunting, while we won't go into detail about it here, like with any language over time as you use it more it will eventually click.

Alongside these parameterised YAML files you specify a Values YAML that defines the environment specific values that should be applied to the parameterised YAML defining the Kubernetes objects.

This allows the YAML files to be consistent between all environments in terms of overall structure whilst varying where they need too. This combination of a Values YAML file and the parameterised YAML defining the Kubernetes objects are what we refer to as Helm Charts.

It maybe that your application is something that needs to be deployed in multiple clusters, for these situations your Helm charts can be provided via a repository in a similar way that we might make re-usable Docker images available.

I think it's at this point that describing Helm as a package manager starts to make sense.

When we think about code package managers we think of re-usable libraries where functionality can be shared and customised in multiple use cases. Helm is allowing us to achieve the same thing with Kubernetes. Without needing to necessarily understand all that is required to deploy the application we can pull down Helm charts, specify our custom values and start using the functionality in our cluster.   

When to Use

The benefit you will achieve by using Helm is largely tied to the scale of the Kubernetes cluster you are managing and the number of applications you are deploying.

If you have a simple cluster with a minimal applications deployed maybe the overhead of dealing with Kubernetes directly is manageable. If you have a large cluster with multiple applications or you have many clusters each with different applications deployed you will benefit more due to the ability it offers to ensure consistency.

You will also benefit from using Helm if you need to deploy a lot of 3rd party applications into your cluster, whether this might be to manage databases, ingress controllers, certificate stores, monitoring or any other cross cutting concern you need to have available in he cluster.

The package manager nature of Helm will reduce the overhead in managing all these dependencies in the same way that you manage dependencies at a code level.

As with many tools your need for Helm may grow over time as you estate increases in complexity. If like me you didn't immediately comprehend the nature and purpose of Helm then hopefully this post has helped you recognise what it can offer and how it could benefit your use case.

Avoiding Toiling

 

Site Reliability Engineering (SRE) is the practice of applying software engineering principles to the management of infrastructure and operations.

Originating at Google in the early 2000s the sorts of things an SRE team might work on include system availability, latency and performance, efficiency, monitoring and the ability to deliver change.

Optimising these kinds of system aspects covers many different topics and areas, one of which is the management of toil.

Toil in this context is not work we don't particularly enjoy doing or don't find stimulating, it has a specific meaning defined by aspects other than our enjoyment of the tasks it involves.

What is Toil?

Toil is work that exhibits some or all of the following attributes.

It is Manual in nature, even if a human isn't necessarily doing the work it requires human initiation, monitoring or any other aspect that means a team member has to oversee its operation.

Toil is Repetitive, the times work has to be done may vary and may not necessarily be at regular intervals, but the task needs to be performed multiple times and will never necessarily be deemed finished.

It is Tactical meaning it is generally reactive, it has to be undertaken either in relation to something happening within the system for example when monitoring highlights something is failing or is sub-optimal.

It has No Enduring Value, this means it leaves the system in the same state as before the work happened. It hasn't improved any aspect of the system or eliminated the need for the work to happen again in the future.

It Scales with Service Growth. Some work items need to happen regardless of how much a system is used. This tends to be viewed as overhead and is simply the cost of having the system in the first place. Toil scales with system use meaning the more users you attract the greater the impact of the toil on your team.

Finally toil can be Automated, some tasks will always require human involvement, but for a task to be toil it must be possible for it to be automated.

What is Toils Impact?

It would be wrong to suggest that toil can be totally eliminated, having a production system being used by large numbers of people is always going to incur a certain amount of toil, and it is unlikely that the whole engineering effort of your organisation can be dedicated to removing it.

Also, much like technical debt, even if you do reach a point where you feel its eliminated the chances are a future change in the system will likely re-introduce it.

But also like technical debt the first step is to acknowledge toil exists, develop ways to be able to detect it and have a strategy for managing it and trying to keep it to a reasonable minimum.

Toils impact is that it engages your engineering resource on tasks that don't add to or improve your system. It may keep it up and running but that is a low ambition to have for any system 

It's also important to recognise that large amounts of toil is likely to impact a teams morale, very few engineers will embark on their career looking to spend large amounts of time on repetitive tasks that lead to no overall value.

The Alternative to Toil

The alternative to spending time on toil is to spend time on engineering. Engineering is a broad concept but in this context it means work that improves the system itself or enables to to be managed in a more efficient way.

As we said previously completely eliminating toil is probably an unrealistic aim. But it is possible to measure how much time your team is spending on toil related tasks. Once you are able to estimate this then it is possible both to set a sensible limit on how much time is spent on these tasks but also measure the effectiveness of any engineering activities designed to reduce it.

This engineering activity might relate to software engineering, refactoring code for performance or reliability, automating testing or certain aspects of the build and deployment pipeline. It might also be more aimed at system engineering, analysing the correctness of the infrastructure the system is running on, analysing the nature of system failures or automating the management of infrastructure.

As previously stated we can view toil as a form of technical debt. In the early days of a system we may take certain shortcuts that at the time are manageable but as the system grows come with a bigger and bigger impact. Time spent trying to fix this debt will set you on a path for gradual system improvement, both for your users and the teams that work on the system.

The Language of Love

 

Software engineers are often polyglots who will learn or be exposed to multiple programming languages over the course of their career. But I think most will always hold a special affection for the first language they learn, most likely because it's the first time they realise they have the ability to write code and achieve an outcome. Once that bug bites it steers you towards a path where you continue to hone that craft.

For me that language is C and its successor C++.

Potentially my view is biased because of the things I've outlined above but I believe C is a very good language for all potential developers to start with. If you learn how to code close to the metal it will develop skills and a way of thinking that will be of benefit to you as you progress onto high level languages with greater levels of abstraction from how your code is actually running.

In The Beginning

In the late 1960s and early 1970s as the Unix operating system was being developed engineers realised that they needed a program language that could be used to write utilities and programs to run on the newly forming platform.

One of the initial protagonists in this field was Ken Thompson.

After dismissing existing programming languages such as Fortran he started to develop a variant of an existing language called BCLP. He concentrated on simplifying the language structures and making it less verbose. He called this new language B with the first version being released around 1969.

In 1971 Dennis Ritchie continued to develop B to utilise features of more modern computers as well as adding new data types. This culminated in the release of New B. Throughout 1972 the development continued adding more data types, arrays and pointers and the language was renamed C.

In 1973 Unix was re-written in C with even more data types being added as C continued to be developed through the 1970s. This eventually resulted in the release of what many consider the be the definitive book on the C programming language, written by Brian Kernighan and Dennis Ritchie The C Programming Language became known as K&R C and became the unofficial specification for the language.

C has continued to be under active development right up until the present with C23 expected to be released in 2024.

C with Classes

In 1979 Bjarne Stroustrup began work on what he deemed "C with Classes".

Adding classes to C turned into an object oriented language, where C had found a home in embedded programming running close to the metal, adding classes made it more suitable for large scale software development.

In 1982 Stroustrup began work on C++ adding new features such as inheritance, polymorphism, virtual functions and operator overloading. In 1985 he released the book The C++ Programming Language which become the unofficial specification for the language with the first commercial version being released later that year.

Much like C, C++ has continued to be developed with new versions being released up until the present day.

Usage Today

Software Engineering is often considered to be a fast moving enterprise, and while many other programming languages have been developed over the lifetime of C and C++ both are still very widely used.

Often being used when performance is critical, the fact they run close to the metal allows for highly optimised code for use cases such as gaming, network appliances and operating systems.

Usage of C and C++ can often strike fear into the heart of developers who aren't experienced in their use. However the skills that usage of C and C++ can develop will prove invaluable even when working with higher level languages so I would encourage all software engineers to spend some time expose themselves to the languages.

Good engineers can apply their skills using any programming language, the principles and practices of good software development don't vary that much between languages or paradigms. But often there are better choices of language for certain situations, and C and C++ are still the correct choice for many applications.

Vulnerable From All Sides

 

Bugs in software engineering are a fact of life, no engineer no matter what he perceives his or her skill level to be has never written a non-trivial piece of software that didn't on some level have some bugs.

These maybe logical flaws, undesirable performance characteristics or unintended consequences when given bad input. As the security of software products has grown ever more important the presence of a particular kind of bug has become something we have to be very vigilant about. 

A vulnerability is a bug that has a detrimental impact on the security of software. Whereas any other bug might cause annoyance or frustration a vulnerability may have more meaningful consequences such as data loss, loss of system access or handing over control to untrusted third parties.

If bugs, and therefore vulnerabilities, are a fact of life then what can be done about them? Well as with anything being aware of them, and making others aware, is a step in the right direction.

Common Vulnerabilities and Exposures 

Run by the Mitre Corporation with funding from the US Division of Homeland Security Common Vulnerabilities and Exposures (CVE) is a glossary that catalogues software vulnerabilities and via the Common Vulnerability Scoring System (CVSS) provides them with a score to indicate their seriousness.

 In order for a vulnerability to be added to the CVE it needs to meet the following criteria.

It must be independent of any other issues, meaning it must be possible to fix or patch the vulnerability without needing to fix issues elsewhere.

The software vendor must be aware of the issue and acknowledge that it represents a security risk. 

It must be a proven risk, when a vulnerability is submitted it must be alongside evidence of the nature of the exploit and the impact it has.

It must exist in a single code base, if multiple vendors are effected by the same or similar issues then each will receive its own CVE identifier.

Common Vulnerability Scoring System (CVSS)

Once a vulnerability is identified it is given a score via the Common Vulnerability Scoring System (CVSS). This score ranges from 0 to 10, with 10 representing the most severe.

CVSS itself is based on a reasonably complicated mathematical formula, I won't here present all the elements that go into this score but the factors outlined below give a flavour of the aspects of a vulnerability that are taken into account, they are sometime referred to as the base factors.

Firstly Access Vector, this relates to the access that an attacker needs to be able to exploit the vulnerability. Do they for example need physical access to a device, or can they exploit it remotely from inside or outside a network. Related to this is Access Complexity, does an attack need certain conditions to exist at the time of the attack or for the system to be in a certain configuration.

Authentication takes into account the level of authentication an attacker needs. This might range from none to admin level.

Confidentially assesses the impact in terms of data loss of an attacker exploiting the vulnerability. This could range from trivial data loss to the mass export of large amounts of data.

In a similar vein Integrity assesses an attackers ability to change or modify data held within the system, and Availability looks at the ability to effect the availability of the system to legitimate users.

Two other important factors are Exploitability and Remediation Level. The first relating to whether code is known to exist that enables the vulnerability to be exploited, and the latter referring to whether the software vendor has a fix or workaround to provide protection.

These and other factors are weighted within the calculation to provide the overall score.

Patching, Zero Days and Day Ones

The main defence against the exploitation of vulnerabilities is via installing patches to the affected software. Provided by the vendor these are software changes that address the underlying flaws causing the vulnerability in order to stop it being exploited.

This leads to important aspects about the lifetime of a vulnerability.

The life of the vulnerability starts when it is first inadvertently entered into the codebase. Eventually it is discovered by the vendor or a so called white hat hacker who notifies the vendor. The vendor will disclose the issue via CVE while alongside this working on a patch to address it.

This leads to a period of time where the vulnerability is known about, by the vendor, white hack hackers and possibly the more nefarious black hat hackers but where a patch isn't yet available. At this point vulnerabilities are referred to as zero days, they may or may not be being exploited and no patch exists to make the software safe again.

It may seem like once the patch is available the danger has passed. However once a patch is released the nature of the patch often provides evidence of the original vulnerability and provide ideas on how it is exploitable. At this point the vulnerability is referred to as a Day One, the circle of those who may have the ability to exploit it has increased, and vulnerable systems are not yet made safe until the patch has been installed.

CVE provides an invaluable resource in documenting vulnerabilities, forewarned is forearmed. Knowing a vulnerability exists means the defensive action can start, and making sure you stay up to date with all available patches means you are as protected as you can be.

The World Wide Internet

 

Surfing the web, getting online and hitting the net are terms that ubiquitous among the verbs that describe how we live our lives. They stopped being technical terms a long time ago and now simply describe daily activities that all generations understand and take part in.

In becoming such commonly used terms they have lost some of their meaning with the web and the internet being interchangeable in most peoples minds. However this isn't the case, they do represent two different if complementary technologies.

Internet vs Web

The Internet is the set of protocols that has allowed computers all over the world to be connected and to exchange data, the so called "network of networks". It is concerned with how data is sent and received not what the data actually represents.

Protocols such the Internet Protocol and the Transmission Control Protocol allow each computer to be addressable and routable to allow the free flowing transmission of data.

The World Wide Web (WWW or W3) is an information system that builds on top of the interconnection between devices that the Internet provides to define a system for how information should be represented, displayed and linked together.

Defined by the concepts of hypermedia and hypertext it provides the means by which we are able to view data online, how we are provided links to that data and that data is visually depicted.

The History of the Internet

As computer science emerged as an academic discipline in the 1950s access to computing resource was scarce. For this reason scientist wanted to develop a way for access to be time shared such that many different teams could take advantage of the emerging technology.

This effort culminated in the development of the first wide are network Advanced Research Projects Agency Network (ARPANET) built by US Department of Defence in 1969.

Initially this interconnected network connected a number of universities including the University of California, Stanford Research Institute and the University of Utah.

In 1973 ARPANET expanded to become international with Norwegian Seismic Array and University College London being brought onboard. Into the 1980s ARPANET continued to grow and started to be referred to as the Internet as short hand for Internetwork.

In 1982 the TCP/IP protocols were standardised and the foundations of what we now know as the Internet were starting to be put in place.

The History of the Web

The Web was the invention of Sir Tim Berners-Lee as part of his work at CERN in Switzerland. The problems he was trying to solve related to the storing, updating and discoverability of documents in large datasets being worked on by large numbers of distributed teams.

In 1989 he submitted his proposal for a system that could solve these problems and in 1990 a working prototype was completed including and HTTP server and the very first browser named after the project and called WorldWideWeb.

Building on top of the network provided by the Internet the project defined the HTTP protocol, the structure of URLs and HTML as the way that the data in documents could be represented.

In 1993 CERN made the decision to make these protocols and the code behind them available royalty free, a decision that would change the world forever and enabled the number of web sites in the world to steadily grow from tens, to hundreds, to thousands, to the vast numbers that we now take for granted.

Many technologies end up having a profound impact on our lives without its terminology becoming understood by those outside technological circles. But the Internet and the web are different, they are so embedded in our lives that URLs, hyperlinks, web address etc are normal everyday words.

In the early days of ARPANET, and probably also for the WWW project, although those teams may have realised they were working on what could be important technologies I think they wouldn't have anticipated quite where their work would lead. But when an idea is a good one, it can go in many unexpected directions.

What's In a Name

 

Surfing the web seems like straightforward undertaking, you type the website you want to go to into your browsers address bar, or you click on a result from a search engine and within no time the website you wanted to visit is in front of you.

But how did this happen simply by typing a web address into a browser? How is the connection made between you and a website out there somewhere in the world?

The answer lies in the Domain Name System (DNS).

All servers on the internet that are hosting the websites you want to access are addressable via a unique Internet Protocol (IP) address. For example as I write this article linkedin.com is currently addressable via 13.107.42.14. These addresses aren't practical for human use which is why we give websites names such as linkedin.com.

DNS is the process by which these human readable names are translated into the IP addresses that can be used to actually access the websites content.

DNS Elements 

Four main elements are involved in a DNS lookup.

A DNS Recursor is a server that receives queries from clients to resolve a websites host name into an IP address. The recursor will usually not be able to provide the answer itself but knows how to recursively navigate the phone directory of the internet in order to give the answer back to the client.

A Root Nameserver is usually the first port of call for the recursor, it can be thought as like a directory of phone directories, based on the area of the internet the websites domain points at it directs the recursor at the correct directory that can be used to continue the DNS query.

A Top Level Domain (TLD) Nameserver acts as the phone directory for a specific part of the internet based on the TLD portion of the web address. For example a TLD nameserver will exists for .com addresses, .co.uk addresses and so on.

An Authoritative Nameserver is the final link in the chain, it is the part of the phone directory that can provide the IP addresses for the website you are looking for.

DNS Resolution

To bring this process to life let's look at the path of a DNS query if you were trying to get to mywebsite.com.

The user types mywebsite.com into their browser and hits enter, the browser then asks a DNS recursor to provide the IP address for mywebsite.com.

The recursor first queries a root nameserver to find the TLD nameserver thats appropriate for this request.

In this example the root nameserver will respond with the TLD nameserver for .com addresses.

The TLD nameserver will then respond with the authoritative nameserver for the websites domain, in this example mywebsite.com, the location of this server will be related to where the website is being hosted. The authoritative nameserver then responds with the IP address for the website, the recursor returns this to the users browser and the website can be loaded.

DNS Security

DNS is one of the fundamental technologies that has its origins in the foundation of the web. At this time when the blueprint of the web was being created security was less of a concern to those solving these engineering problems, it was assumed that the authenticity of the links in the chain could be taken on trust.

Unfortunately in the modern web this level of trust in other actors can be misplaced. When a server claims to be the authoritative nameserver for a particular website how can you trust that this is the case and you aren't going to be directed to a rogue impersonation of the website you are trying to reach.

Domain Name System Security Extensions (DNSSEC) is attempting to replace the trust based system with one that is based on provable security. It introduces the signing and validation of the DNS records being returned from the various elements involved in a DNS query so that their authenticity can be determined.

DNS is one of the technologies that is now taken for granted but solves a problem without which the web as we know it wouldn't be able to exist. On the surface it sounds like a simple problem to solve but the scale of the web means even the simplest of solutions has to be able to scale to a world wide scale.

The Web of Trust

 

Everyday all of us type a web address into a browser or click on a link provided by a search engine and interact with the web sites that are presented.

Whilst we should always be vigilant, on many occasions we will simply trust that the site we are interacting with is genuine and the data flowing between us and it is secure.

Our browsers do a good job of making sure our surfing is safe, but how exactly is that being achieved. How do we create trust between a website and its users?

SSL/TLS

Netscape first attempted to solve this trust problem by introducing the Socket Secure Layer (SSL) protocol in the early 90s. Initial versions of the protocol still had many flaws but by the release of SSLv3.0 in 1996 it had matured into a technology that was able to provide a mechanism for trust on the web.

As SSL became a foundational part of the web, and because security related protocols always have to be under constant evolution to maintain safety, the Internet Engineering Task Force (IETF) developed Transport Layer Security (TLS) in 1999 as an enhancement to SSLv3.0.

TLS has continued to be developed with TLSv1.3 being released in 2018. 

Its primary purpose is to ensure data being exchanged by a server and a client is secured, but also to establish a level of trust such that the two parties can be sure who they are exchanging the data with.

Creating this functionality relies on a few different elements.

Public Key Encryption

Public key encryption is a form of asymmetric encryption that uses a pair of related keys deemed public and private.

The mathematics behind this relationship between the keys is too complex to go into in this post, but the functionality it provides is based on the fact that the public key can be used to encrypt data that only the private key can decrypt.

This means the public key can be freely distributed and used to encrypt data that only the holder of the private key can decrypt.

The keys can also be used to produce and verify digital signatures. This involves the holder of some data using a mathematical process to "sign" this data using their private key.

The receiver of the data can use the public key to verify the signature and therefore prove that the data came from someone who has the corresponding private key.

Public Key Infrastructure (PKI)

Public Key Infrastructure (PKI) builds on top of the functionality provided by public key encryption to provide a system for establishing trust between client and server.

This is achieved via the issuance of digital certificates from a Certificate Authority (CA).

The CA is at the heart of the trust relationship of the web. When two parties, the client and server, are trying to form a trust relationship they must delegate to a 3rd party that they both already trust, this is the CA.

The CA establishes the identity of the organisation the client will interact with via off line means and issues a digital certificate. This certificate establishes the identity of the organisation, its public key and is signed by the CA to prove it was the one that issued the certificate.

A client when it receives the certificate from the server can use the CA's public key to verify the signature and therefore trust the data in the certificate.

It's possible to have various levels of CA's that may delegate trust to other CA's, deemed intermediary CA's. But all certificates should ultimately be able to be traced back to a so called Root CA that all parties on the web have agreed to trust and whose public keys are available to all participants.   

Certificates and Handshakes

All of the systems previously described are combined whenever we visit a web site to establish trust and security.

• A user types a web address into the browser or clicks a link provided by a search engine.
• The user's browser issues a request to the web site to establish a secure connection.
• The server in response sends the browser it's certificate.
• The browser validates the certificate authenticity by verifying the signature of the Root CA that the certificate is issued from using the public key of the CA that has been pre-installed on the users machine.
• Once the certificate is validated, the browser creates a symmetric encryption key that will be used to secure future communication between the browser and the web site. It encrypts the symmetric key using the servers public key and sends it to the server.
• The users browser has now established the identity of the web site, based on the data contained in its validated certificate, and both parties now have a shared symmetric key that can be used to secure the rest of their communication in the session.

There are certain pieces of functionality that are fundamental to allowing the web to operate in the way it does.

Without the functionality provided by SSL/TLS it wouldn't be possible to use the web as freely as we do whilst also trusting that we can do so in a safe and secure manner.