A MongoDB White Paper

Serverless Architectures: The Evolution

of Cloud Computing
September 2016
Table of Contents
Introduction 1

The Evolution of Cloud Computing 1

What is Serverless Computing? 2

Benefits of Serverless Computing 3

Serverless Computing Principles 4

MongoDB Atlas and Serverless 4

Important Considerations Before Going Serverless 6

Summary 7

Resources 7
Introduction

Since the advent of the computer, building software has

The Evolution of Cloud
been a complicated process. Over the past decade, new
infrastructure approaches (IaaS and PaaS), software Computing
architectures (SOA and Microservices), and methodologies
(Agile, Continuous Delivery and DevOps) have emerged to
In order to understand how serverless computing fits with
mitigate the complexity of application development. While
modern application development, it is helpful to first
microservices has been the hot trend over the past couple
understand the evolution of cloud computing.
of years, serverless architectures have been gaining
momentum by providing a new way to build scalable and Initially, enterprises leveraged data centers to abstract the
cost effective applications. Serverless computing frees physical hosting environment. The hardware was the unit of
developers from the traditional cost of building applications scale, which meant that in order to scale out your
by automatically provisioning servers and storage, application you would allocate more compute, storage, and
maintaining infrastructure, upgrading software, and only networking resources to handle the required load.
charging for consumed resources.

This whitepaper discusses the background behind

serverless computing, its advantages, best practices, and
key considerations when evaluating a database with your
serverless environment.

1
 operating system, are abstracted away from the developer.
Applications are run in stateless compute containers that
are event triggered (e.g. a user uploading a photo which
triggers notifications to his/her followers). Developers
create functions and depend on the infrastructure to
allocate the proper resources to execute the function. If the
load on the function grows, the infrastructure will create
copies of the function and scale to meet demand.

Serverless computing supports multiple languages so

developers can choose the tools they are most comfortable
with. Users are only charged for runtime and resources
(e.g. RAM) that the function consumes; thus there is no
longer any concept of under or over provisioning. For
Figur
Figuree 1: Evolution of Cloud Computing
example, if a function runs for 500ms and consumes 15
MB of RAM, the user will only be charged for 500 ms of
As new technologies emerged, enterprises started moving
runtime, and the cost to use 15 MB of RAM.
to an Infrastructure-as-a-Service (IaaS) model, which
abstracted away the physical hardware and made the Serverless architectures are a natural extension of
operating system the unit of scale. Virtual machines (VMs), microservices. Similar to microservices, serverless
which are packaged operating systems that run an architecture applications are broken down into specific
application, were used to emulate dedicated hardware core components. While microservices may group similar
resources. If an application required more resources, then a functionality into one service, serverless applications
VM could be resized or other VMs provisioned to handle delineate functionality into finer grained components.
the load. Custom code is developed and executed as isolated,
autonomous, granular functions that run in a stateless
With Platform-as-a-Service (PaaS), the operating system
compute service.
was abstracted away, and the application became the unit
of scale. There was no longer any need to deal with To illustrate this point, lets look at a simple example of how
different operating system versions and middlewares, as a microservice and serverless architecture differ.
the application would be allocated more resources if the
load increased. In Figure 2, a client interacts with a User microservice. A
container is pre-provisioned, and all of the functionality of
Now with serverless computing (also referred to as the User service resides in the container. The service
Function-as-a-Service), the language runtime is abstracted consists of different functions (update_user, get_user,
away and the function is the unit of scale. The cloud create_user, delete_user) and is scaled based on the
provider takes responsibility for the vast majority of the IT overall load across the service. The service will consume
stack. If the function needs to be scaled, the cloud provider hardware resources even while idle, and the user will still
allocates the appropriate resources and creates copies of be charged for the underutilized resources.
the function to handle the required load.

What is Serverless Computing?

Serverless computing is the next layer of abstraction in

cloud computing. It does not mean that there are no
servers. Instead, the underlying infrastructure (physical and
virtual hosts, virtual machines, containers), as well as the

2
 compute resources until customers start using the
application.

Elastic Sc
Scalability:
alability: Elastic scalability is also simple with a
serverless architecture. If a function needs to scale, the
infrastructure will make copies of the function to handle
the load. An example of this could be a chatbot that
responds to weather requests. In a serverless architecture,
a chatbot function would handle the response by retrieving
the users location and responding back with the
Figur
Figuree 2: Microservices Architecture temperature. For a few requests this is not a problem, but
what happens if the chatbot service is flooded with
For a serverless architecture, the User service would be thousands of request a second. For this scenario, the
separated into more granular functions. In Figure 3, each chatbot function would automatically scale by instantiating
API endpoint corresponds to a specific function and file. thousands of copies of the function. Once the requests
When a create user request is initiated by the client, the have subsided, the environment would automatically
entire codebase of the User service does not have to run; terminate the idle instances and scale down, allowing costs
instead only create_user.js will execute. There is no need to to scale proportionally with user demand.
pre-provision containers, as standalone functions only
consume resources when needed, and users are only Rapid Development and Iteration: Serverless computing
charged on the actual runtime of their functions. is ideal for companies that need to quickly develop,
prototype, and iterate. Development is faster since there
arent any dependencies on IT Ops. Functions are typically
single threaded, which makes debugging and deploying
functions simpler. The build process is also broken down
into smaller and more manageable chunks. This increases
the number of changes that can be pushed through the
Continuous Delivery pipeline, resulting in rapid deployment
and more iterative feedback.

Less System Administration: Serverless doesnt mean

that you completely obviate the operational element of your
Figur
Figuree 3: Serverless Architecture
infrastructure, but it does mean that there is less system
administration. There are no servers to manage, provision,
Benefits of Serverless and scale, as well as no patching and upgrading. Servers
Computing are automatically deployed in multiple availability zones to
provide high availability. Support is also streamlined; if there
is an issue in the middle of the night it is the responsibility
Costs Sc
Scale
ale W
With
ith Usage: One of the biggest benefits of of the cloud provider to resolve the problem.
serverless computing is that you only pay for the runtime of
your function. There is no concept of idle resources as Developer Pr Productivity:
oductivity: By using a serverless
you are not charged if the function is not executed. This is architecture, developers can focus more on writing code
especially helpful for applications that are only used a few than having to worry about managing the operational tasks
times an hour, which means any dedicated hardware, VMs, of the application. This allows them to develop innovative
or containers would be sitting idle for the majority of the features and focus on the core business logic that matters
time, and a user would be charged for underutilized most to the business.
resources. With serverless computing, enterprises could
build out an entire infrastructure and not pay for any

3
Serverless Computing Principles queues, and applications can respond quickly to events and
scale as needed.

Embrace T Thir
hird
d Party Services: Serverless computing
St
Stateless
ateless Single-Purpose F Functions:
unctions: For a serverless
vendors may not comprise all the technologies that
architecture, it is good practice to design functions around
developers are familiar with or need. Leveraging third party
the Single Responsibility Principle (SRP). A function that is
services and applications allow developers to pick the right
developed to perform a specific task is easier to test,
tools to help accomplish the task at hand. There are many
deploy, release, and is also more robust. Functions
services that developers can leverage from Auth0 for
designed with well defined interfaces are also more likely
authentication to Braintree or Stripe for payment
to be re-used. From a deployment unit point of view,
processing. It is much more useful for a developer to spend
serverless functions are stateless, which means local
time solving a problem unique to their domain, than it is to
resources and processes will not be retained after the
recreate functionality that has already been built by
function executes. This allows the platform to quickly scale
someone else.
to handle an increasing number of events or requests.

Mor
MoreeF Frront-End F
Functionality:
unctionality: In serverless
architectures, functions that terminate faster are less MongoDB Atlas and Serverless
expensive. Creating more functionality in the front-end
Computing
helps to reduce costs by minimizing the number or
complexity of function requests. This contrasts the typical
application models where all requests flow through the With MongoDB Atlas, users can leverage the rich
back-end layer before being routed to the database. The functionality of MongoDB expressive query language,
emergence of rich client-side application frameworks, such flexible schema, always-on availability, distributed scale-out
as AngularJS, make it possible to seamlessly execute from a serverless environment. MongoDB Atlas is a
complex functionality within the browser. Back-end logic is database as a service and provides all the features of
decoupled from the server and moved to the front-end, MongoDB without the heavy lifting of operational tasks.
allowing the front-end to communicate with more services. Developers no longer need to worry about provisioning,
This reduces the number of hops required to access a configuration, patching, upgrades, backups, and failure
resource and results in better performance and a richer recovery. Atlas offers elastic scalability, either by scaling up
user experience. on a range of instance sizes, or scaling out with automatic
sharding, all with no application downtime.
Design Event Driven Pipelines: Building event driven,
push based pipelines is powerful and allows you to scale Setting up Atlas is simple.
out architectures easily. An event driven pipeline is a chain
reaction of events that propagate down the pipeline
without any user input. For example, a user could upload a
video file which kicks off a function that takes the file and
video encodes it into multiple formats: 480p, 720p, 1080p.
The newly encoded videos are saved to a storage service,
which triggers another event. A function responds to the
event and creates metadata about the videos that is
pushed to a database. Metadata is saved to the database,
which triggers another function that creates an email
notification sent to the user confirming the video has been
encoded to the different formats.These actions all happen
asynchronously without any user input. With event driven Figur
Figuree 4: Provisioning MongoDB Atlas cluster
pipelines, there is no need to manage servers or event

4
Select the instance size that fits your application needs and entry to understand the codebase, resulting in higher
click CONFIRM & DEPLOY. software performance and developer productivity.

Rapid Deployment: With MongoDB Atlas, a MongoDB

cluster can be provisioned and deployed in minutes and
even seconds. Developers no longer need to worry about
configuring or managing servers. Integrating MongoDB
Atlas into your serverless platform is as simple as passing
the connection string into your serverless application.

Figur
Figuree 5: MongoDB Atlas Cluster Monitoring

MongoDB Atlas provides many benefits for those

interested in building a serverless architecture:

Vendor Independence: Cloud providers typically only

offer databases specific to that provider, which may not fit
Figur
Figuree 6: MongoDB Atlas Connection String
with what a developer needs. MongoDB Atlas provides
independence from the underlying cloud provider and
MongoDB Atlas features extensive capabilities to defend,
empowers developers to choose the appropriate tools for
detect, and control access to MongoDB, offering among
their needs. Developers can leverage the rich functionality
the most complete security controls of any modern
of MongoDBs query language and flexible data model,
database:
without worrying about the operational tasks of managing
the database. If you decide to shift to your own User Rights Management: Control access to sensitive
infrastructure, or another cloud provider, you wont have to data using industry standard mechanisms for
repopulate your data on a different database. MongoDB authentication and authorization at the database level
Atlas is currently available only on AWS, with support for
Encryption: Protect data in motion over the network
Microsoft Azure and Google Cloud Platform (GCP) coming
and at rest in persistent storage
soon.
To ensure a secure system right out of the box,
M EAN StStac
ack:
k: Serverless architectures accelerate the
authentication and IP Address whitelisting are
trend of shifting business logic from the back-end to the
automatically enabled.
front-end. This makes the choice of front-end framework
much more important. AngularJS, is ideally suited for this IP address whitelisting is a key MongoDB Atlas security
requirement, and is a popular front-end for serverless feature, adding an extra layer to prevent 3rd parties from
architectures. AngularJS is a structural Javascript accessing your data. Clients are prevented from accessing
framework for dynamic web applications that provides the database unless their IP address has been added to
interactive functions and AJAX a technique for creating the IP whitelist for your MongoDB Atlas group.
fast and dynamic web pages rich components.
For AWS, VPC Peering for MongoDB Atlas is under
Combined with NodeJS, ExpressJS, and MongoDB, these
development and will be available soon, offering a simple,
tools form the MEAN stack (MongoDB, ExpressJS,
robust solution. It will allow the whitelisting of an entire
AngularJS, NodeJS). There are huge advantages to using
AWS Security Group within the VPC containing your
JavaScript and JSON throughout your serverless stack.
application servers.
Someone working on the front-end can easily understand
the function (back-end) code and database queries. Sc
Scalability:
alability: You should expect your serverless functions to
Additionally, using the same syntax and objects through the scale out, thus downstream setups need to be architected
whole stack frees your team from understanding multiple to keep up and scale out with your functions. Relational
language best practices, as well as reduces the barrier to databases tend to break down with this model. MongoDB

5
Atlas is designed with scalability as a core principle. When
your cluster hits a certain threshold, MongoDB Atlas will
send you an alert and with one click you can provision new
servers.

Flexible Sc
Schema:
hema: Because serverless architectures are
event driven, many use cases revolve around around the
Internet of Things (IoT) and mobile. MongoDB is ideal for
these use cases and more as its flexible document model
enables you to store and process data of any type: events,
geospatial, time series, text, binary, and anything else.
Adding new fields to the document structure is simple,
making it easy to handle changing data generated by event
driven applications. Developers spend less time modifying
schemas and more time innovating. Figur
Figuree 7: Server vs. Serverless Break Even Point

Cloud Pr Provider
ovider L Loc
ock-In:
k-In: Each cloud provider has their
Important Considerations Before own implementation of serverless computing; thus it is
Going Serverless likely that features will vary significantly from platform to
platform. This makes the risk of cloud provider lock-in high,
as porting functions between different providers is difficult
Though serverless computing offers many advantages, and expensive. Not only will you need to update your
there are several considerations to keep in mind before operational tools (deployment, monitoring), but you will
implementing a serverless project. most likely also need to update your code to be compatible
with the different interfaces. Users are more susceptible to
Underst
Understand
and YYour
our Br
Break
eak Even P Point:
oint: It is important to
changes in pricing, performance, and stability as there is
understand the break even point for your application. For
less flexibility to shift to another provider.
an application that only handles tens of requests an hour, a
serverless architecture is cost effective. On the other hand, High StStartup
artup Latency: For a function thats just been
for an application that handles thousands of requests per created, or hasnt been used recently, the startup latency
second it may be cheaper to provision more servers, rather will be high because the environment will have to spin up a
than pay for millions of processes a month. Figure 5 is new container before the function is executed. This can be
hypothetical and not based on real numbers, but illustrates problematic for certain applications that see intermittent
that beyond a specific number of requests it may not be traffic patterns or require low latencies (e.g. financial
economical to use a serverless architecture. trading applications).

Complexity Managing Pr Processes:

ocesses: In a serverless
architecture, the time that developers spend on managing
the application is not necessarily reduced, but allocated
differently. Though developers spend less time managing
the infrastructure, they spend more time managing
processes. Handling hundreds of processes may be trivial,
but handling millions or billions of processes can be
overwhelming to inexperienced teams. Before
implementing a serverless architecture, it is important to
determine if your team has the right capabilities to handle
the associated complexities.

6
Lac
Lack k of Resour
Resources
ces and Skillsets: Serverless computing
Resources
is a nascent field, which means hiring knowledgeable
developers or experts is difficult. Learning serverless
computing requires a steep learning curve and resources For more information, please visit mongodb.com or contact
are not easily accessible. There isnt a wealth of us at [email protected].
information on Stack Overflow or other forums. Books
Case Studies (mongodb.com/customers)
detailing best practices are also scarce, resulting in trial
Presentations (mongodb.com/presentations)
and error for many developers.
Free Online Training (university.mongodb.com)
Webinars and Events (mongodb.com/events)

Summary Documentation (docs.mongodb.com)

MongoDB Enterprise Download (mongodb.com/download)
MongoDB Atlas database as a service for MongoDB
Serverless architectures are relatively new and build on the (mongodb.com/cloud)
work started by microservices. MongoDB Atlas is a
database as a service that is well suited for serverless
architectures as it provides elastic scalability and all the
features of the database without managing operational
tasks. Though serverless architectures provide many
benefits, there are many considerations to keep in mind
when looking to implement a serverless architectures.
Thus, diligent planning should be taken before embarking
on that path.

Resources

Run MongoDB Atlas today.

Read the MongoDB microservices whitepaper.

New York Palo Alto Washington, D.C. London Dublin Barcelona Sydney Tel Aviv
US 866-237-8815 INTL +1-650-440-4474 [email protected]
2017 MongoDB, Inc. All rights reserved.