L 11 - FDN

Question 1

Are traditional IT, IaaS, and Paas serverless?

Answer 1

No only FaaS.

Question 2

What is the unit of scale in traditional IT?

Answer 2

physical servers

Question 3

What is the unit of scale in IaaS?

Answer 3

VMs

Question 4

What is the unit of scale in PaaS?

Answer 4

Containers

Question 5

What is the unit of scale in FaaS?

Answer 5

Functions

Question 6

How long does the deployment of physical servers take?

Answer 6

hours/days

Question 7

How long does the deployment of VMs take?

Answer 7

minutes

Question 8

What is a cloud native application?

Answer 8

Cloud-native applications are a collection of small, independent, and loosely coupled services. They are designed to deliver well-recognized business value, like the ability to rapidly incorporate user feedback for continuous improvement.

Question 9

Is a monolithic application cloud-native?

Answer 9

No

Question 10

Are microservices or functions cloud native applications?

Answer 10

Yes

Question 11

Do you have an API gateway in monolithic applications?

Answer 11

No

Question 12

What is the difference between a monolithic application and microservices?

Answer 12

In a monolithic application you have an application with has a user-interface and a backend in one “container”.
In Microservices you have a user-interface then API gateway to the different microservices that each have their own database.

Question 13

What is the difference between FaaS-based apps and microservices?

Answer 13

In microservices you have several microservices after the API-gateway. In FaaS you can have more than one function for each service.

Question 14

What are 4 steps in FaaS?

Answer 14

You have events that
1. Trigger function invocation
2. Check if an instance is free (No –> cold start new instance (container in VM)
3. Execute Handler Method (function is run)
4. Response to events

Question 15

What is the Handler Method?

Answer 15

Internet:
In FaaS (Function as a Service), the handler method is the entry point of a serverless function. It is responsible for processing the input, performing the desired actions, and returning the output.

The handler method typically takes two arguments: an event object that contains the input data for the function and a context object that provides information about the runtime environment. The handler method should return a response object or raise an exception if an error occurs.

The signature of the handler method is specific to the FaaS platform being used and is defined in the function definition. For example, in AWS Lambda, the handler method signature is usually written as follows:

python
Copy code
def handler(event, context):
# Your function logic here
return {
“message”: “Hello World”
}

Question 16

What is the difference between a cold start and a warm start?

Answer 16

In a cold start the code has to be downloaded and then a new instance (e.g. container) is started. You bootstrap the runtime (load the dependencies etc.) and then you execute the function handler.

In a warm start the function handler method executes directly.

Question 17

What is the edge-cloud continuum?

Answer 17

You have the cloud layer, the edge layer (server closer to user), and end device layer

Question 18

Do you get less response time with an increase in memory?

Answer 18

Yes, the response time in seconds decreases if more MB of memory are allocated. You get more CPU for more memory. But the decrease is not linear and fades off with high levels of memory.

Question 19

What are 3 sources of heterogeneity in serverless deployment?

Answer 19

• Hardware-config-based
• location based
• platform-based hetero FaaSD (openwhisk-self-hosted is fastest, IBM is slowest and Google and AWS cloud functions are mid-range)

Question 20

What are heterogeneous serverless clusters about?

Answer 20

You split of functions into different “buckets” to increase speed. Like you have AWS lambda function running with 256MB or some with 1024 MB in the HW-Config-base or you split them by location or by platform base.

Question 21

Why use heterogeneous sever clusters? 3 Advantages

Answer 21

• use specialized architectures and hardware
• overcome limitations of FaaS platform and cloudbursting
• data-locality, data-privacy and not all cloud providers have data centers in all regions

Question 22

What is cloudbursting?

Answer 22

Cloud bursting is a technique for scaling computing resources on-demand by temporarily renting resources from a public cloud provider, such as Amazon Web Services or Microsoft Azure, when an application’s workload exceeds the capacity of the organization’s private cloud. This allows organizations to handle sudden spikes in traffic or processing demands without having to invest in and maintain a large amount of excess capacity.

Question 23

How to deliver function invocations? (How to send user invocations across heterogeneous compute clusters?)

Answer 23

use a Function Delivery Network

Question 24

What is Terraform?

Answer 24

Terraform is a tool for building, changing, and versioning infrastructure safely and efficiently. It can manage popular service providers as well as custom in-house solutions.

It uses a simple syntax written in the HashiCorp Configuration Language (HCL) to describe the desired state of the infrastructure and then automatically provisions and configures the infrastructure according to the specified configuration. Terraform is cloud-agnostic, which means it can be used to manage infrastructure across multiple cloud providers, including AWS, Azure, GCP, and more.

It is commonly used for automating the process of creating, updating, and maintaining infrastructure in a safe and predictable manner. It can be used to provision and manage resources such as virtual machines, DNS entries, databases, and more.

Question 25

What is Ansible?

Answer 25

Ansible is an open-source automation tool for configuring and managing systems, applications, and IT infrastructure. It uses simple, human-readable language to describe the desired state of a system, and then automatically enacts those changes.

Ansible uses a push-based model, where the control node (where Ansible is installed) pushes configurations out to the target nodes.

Question 26

Which tasks can Ansible automate?

Answer 26

Ansible can be used to automate a wide range of tasks, including:

Configuring and managing servers
Deploying and scaling applications
Managing users and permissions
Managing software packages and dependencies
Running ad-hoc commands and scripts
And more
It also have a large collection of pre-built modules, called Ansible modules, that can be used to perform specific tasks on different systems such as Linux, Windows and network devices.

Question 27

What is SSH?

Answer 27

SSH stands for Secure Shell. It is a protocol for securely connecting to and managing remote systems over an unsecured network. SSH provides a secure and encrypted connection for remote login, file transfer, and other network services.

Question 28

How to create multiple serverless clusters?

Answer 28

Question 29

Which two services are used to create multiple serverless clusters?

Answer 29

Terraform (Infrastructure as a code) and Ansible Management Node

Question 30

How to manage FDN clusters?

Answer 30

use virtual kubeletes

Question 31

What are virtual kubelets?

Answer 31

They provide the illusion to the master node that there is a node although there actually is none.

The virtual kubelet registers itself as a node and allows developers to deploy pods and containers with their own APIs. These the connect to different serverless clusters.

Question 32

How does function deployment work in FDN?

Answer 32

you write a JAML file which can be put on any platform by the FDN Management Cluster

Question 33

How to observe/monitor clusters?

Answer 33

FDN monitor sends data over virtual nodes to Grafana for visualization.

Question 34

What two options are there for the FDN-monitor?

Answer 34

K8s collector or AWS lambda collector.

Question 35

What are two methods to schedule functions?

Answer 35

1. FDN Modeler
2. Courier: Scheduling Functions

Question 36

What are the two main characteristics of the courier in scheduling functions?

Answer 36

• Function Delivering Policies
• Load Balancing Strategies

Question 37

What are Courier’s two levels?

Answer 37

1. Subset of clusters (delivery)
2. Load Balancing

Question 38

What are the two function delivering policies for the Courier?

Answer 38

• function aware
• data aware

Question 39

What are two load-balancing strategies?

Answer 39

• latency-aware
• capacity-aware

Question 40

What is data aware scheduling?

Answer 40

Data-aware scheduling is a technique used in distributed systems to schedule tasks based on the availability and location of data. The goal of data-aware scheduling is to minimize data movement and reduce the time and cost associated with data transfer over a network.

Question 41

What is a FedLess control plane?

Answer 41

A “fedless control plane” would be a control plane that eliminates the need for federation. In a federated system, multiple control planes are connected together, allowing a single control plane to manage multiple clusters or regions. A fedless control plane is a control plane that can manage multiple clusters or regions without being federated, it can be a centralized control plane that can manage multiple clusters or regions.

In the context of a Function as a Service (FaaS) and “Function Delivery Network (FDN)”, a fedless control plane would be a control plane that can handle the management of functions and resources across multiple clusters or regions without the need of federation.