Professional Documents
Culture Documents
Edge Computing Patterns For Solution Architects Anonymous Full Chapter
Edge Computing Patterns For Solution Architects Anonymous Full Chapter
Architects Anonymous
Visit to download the full and correct content document:
https://ebookmass.com/product/edge-computing-patterns-for-solution-architects-anon
ymous/
Edge Computing Patterns for Solution Architects
Copyright © 2024 Packt Publishing
All rights reserved. No part of this book may be reproduced, stored in a retrieval system, or
transmitted in any form or by any means, without the prior written permission of the publisher,
except in the case of brief quotations embedded in critical articles or reviews.
Every effort has been made in the preparation of this book to ensure the accuracy of the information
presented. However, the information contained in this book is sold without warranty, either express or
implied. Neither the author(s), nor Packt Publishing or its dealers and distributors, will be held liable
for any damages caused or alleged to have been caused directly or indirectly by this book.
Packt Publishing has endeavored to provide trademark information about all of the companies and
products mentioned in this book by the appropriate use of capitals. However, Packt Publishing cannot
guarantee the accuracy of this information.
Grosvenor House
11 St Paul’s Square
Birmingham
B3 1RB, UK
ISBN 978-1-80512-406-1
www.packtpub.com
This book would not be possible without the unwavering support and understanding of Radha
and tons of encouragement from Sameer, Siddharth, and Mythili. They collectively inspire me
with their own successes. And to my newest cheerleader, Seelan, who never ceases to amaze me.
-- Ashok Iyengar
To my mentor and guide, Rob High. Thanks for the guidance, trust, and opportunity to learn
about edge computing.
– Joe Pearson
Contributors
Joseph Pearson has worked as a college instructor and in various media companies for over three
decades, most recently as Software Architect at The Weather Channel on weather.com. He has since
worked for IBM Cloud as a Strategist and then IBM Software in Edge Computing. He currently
works for IBM Software’s Networking and Edge Computing unit on their open source strategy. He
volunteers with the Linux Foundation as Chair of the LF Edge Technical Advisory Council (TAC)
and leads the Open Horizon project. And in any spare time, he enjoys geocaching.
I would like to thank my wife and children for bearing with me as I stole time from them
throughout the process of writing this book. I especially want to thank Ashok Iyengar for
supporting and encouraging me to tackle working on this book when it seemed so daunting. It’s
always helpful to have an experienced friend and author to collaborate with who can also be a
big brother and provide frank guidance when it’s most needed.
Frédéric is the author of “Building Enterprise IoT Solutions using Eclipse IoT Technologies: An
Open-Source Approach to Edge Computing,” published in December 2022 by Apress (ISBN: 978-
1484288818).
Table of Contents
Preface
Part 1:Overview of Edge Computing as a Problem
Space
7
Automate to Achieve Scale
Automating service delivery
DevOps
Infrastructure as code
Extending automation to the edge
Developing edge applications
Scalability with automation
Prepping an edge device
Prepping an edge cluster
Operational security
Limiting physical access
Limiting connectivity
Trusted hardware and provisioning
Trusted data
Trusted compute
Tactical Edge
Automation with AI
LLMs and generative AI
Using AI in automation
Summary
10
11
Index
If the name of the game is to get insights quickly from all the data, we have to ensure there are
applications able to do that, devices that can host the applications, and a network facilitating the
information flow. That is the challenge solution architects face, which is who this book is primarily
written for.
Drawing from real-world deployments in large enterprises and standards proposed by edge
computing community organizations, this book emphasizes practical and tested-at-scale patterns and
best practices used by leading companies worldwide. The book takes the reader from edge computing
terminology and concepts to simple architectures, and eventually through end-to-end industry-
specific approaches. It gives several points of view and rules of thumb and provides resources and
recommendations for deeper thinking and research.
Who this book is for
This book is intended for IT professionals who have created solution architectures and are looking to
extend their skills to the edge computing space. It should be valuable to VPs of IT infrastructure,
enterprise architects, solution architects, and SRE professionals who are familiar with cloud
computing and are interested in creating an edge reference architecture or a solution architecture for a
particular industry use case. The book provides common patterns from solutions implemented by
customers in industries ranging from retail to telcos to manufacturing.
What this book covers
Chapter 1, Our View of Edge Computing, establishes a shared vernacular, describes infrastructure
scenarios, and relies on industry conventions to ensure standard and widely compatible solutions.
Chapter 2, Edge Architectural Components, covers four roles of components in edge architectures,
benefits and limitations, and functional versus non-functional requirements.
Chapter 3, Core Edge Architecture, covers managing and enabling sensors and smart devices using
the Edge Device Hub pattern.
Chapter 5, End-to-End Edge Architecture, brings together devices, macro components, and
applications to solve industry-specific challenges.
Chapter 6, Data Has Weight and Inertia, explores the data-related considerations that edge-native
solutions will need to address.
Chapter 7, Automate to Achieve Scale, includes approaches that have worked at the extreme scales
that edge deployments may encounter.
Chapter 8, Monitoring and Observability, covers how to ensure that a deployed edge solution is
performing as designed, despite unique challenges.
Chapter 9, Connect Judiciously but Thoughtlessly, covers three connection scenarios and how
application-centered approaches can address them.
Chapter 10, Open Source Software Can Benefit You, explores strategies for ensuring that open source
dependencies are used optimally, and when and how an enterprise should open source their solutions.
Chapter 11, Recommendations and Best Practices, takes a wider view of the problem space and how
thinking deeply about what you are doing and why can yield some surprising insights.
Conventions used
There are a number of text conventions used throughout this book.
Code in text: Indicates code words in text, database table names, folder names, filenames, file
extensions, pathnames, dummy URLs, user input, and Twitter handles. Here is an example: “If you
notice closely, it is an Open Horizon environment variable, namely HZN_DEVICE_ID.”
#!/bin/sh
# Simple edge service
while true; do
echo "HZN_DEVICE_ID says: Hello from Packt!"
sleep 5
done
Bold: Indicates a new term, an important word, or words that you see onscreen. For instance, words
in menus or dialog boxes appear in bold. Here is an example: “The following code snippet is a very
simple Hello World service that outputs Hello from Packt every five seconds.”
Get in touch
Feedback from our readers is always welcome.
General feedback: If you have questions about any aspect of this book, email us at
[email protected] and mention the book title in the subject of your message.
Errata: Although we have taken every care to ensure the accuracy of our content, mistakes do
happen. If you have found a mistake in this book, we would be grateful if you would report this to us.
Please visit www.packtpub.com/support/errata and fill in the form.
Piracy: If you come across any illegal copies of our works in any form on the internet, we would be
grateful if you would provide us with the location address or website name. Please contact us at
[email protected] with a link to the material.
If you are interested in becoming an author: If there is a topic that you have expertise in and you
are interested in either writing or contributing to a book, please visit authors.packtpub.com.
Your review is important to us and the tech community and will help us make sure we’re delivering
excellent quality content.
Download a free PDF copy of this book
Thanks for purchasing this book!
Do you like to read on the go but are unable to carry your print books everywhere?
Is your eBook purchase not compatible with the device of your choice?
Don’t worry, now with every Packt book you get a DRM-free PDF version of that book at no cost.
Read anywhere, any place, on any device. Search, copy, and paste code from your favorite technical
books directly into your application.
The perks don’t stop there, you can get exclusive access to discounts, newsletters, and great free
content in your inbox daily
https://packt.link/free-ebook/9781805124061
3. That’s it! We’ll send your free PDF and other benefits to your email directly
Part 1:Overview of Edge Computing as a Problem
Space
To build a shared understanding, we lay the groundwork and survey the territory in the first two
chapters. Since the concepts and terminology in edge computing can be overloading, it’s important to
explain exactly what is meant when we elucidate our concepts, problems, and solutions. The first two
chapters in the book aim to provide clarity and a common foundation that will be built on in Parts 2
and 3. This part begins with how to think and talk about edge computing grounded in context. It then
delves into the various components, describes their purposes, and shows how they relate to others and
where they best fit.
In this chapter, we will start by describing various edge computing scenarios from an infrastructure
point of view, moving from cloud to far edge based on our experiences, research, and available de
facto standards. Along the way, we compare and contrast different points of view that will affect
architectural choices you can make, such as edge computing versus distributed computing and the
network edge versus the enterprise edge.
We will rely on conventions covered in the Suggested pre-reading material section, such as State of
the Edge annual reports and LF Edge whitepapers. By the end of the chapter, you should have the
shared vocabulary and a wider perspective needed to engage in fruitful conversations on edge
computing with software architects and other IT professionals.
(https://stateoftheedge.com/reports/state-of-the-edge-report-2023/)
From DevOps to EdgeOps: A Vision for Edge Computing (Eclipse Foundation) (https://outreach.eclipse.foundation/edge-
computing-edgeops-white-paper)
Edge computing describes computing capabilities situated at degrees of distance from a centralized
location, usually the cloud or a corporate DC. The placement of the equipment is chosen in order to
improve the performance, security, and operating cost of the applications and services that will run in
that environment. In exchange, some factors may be de-emphasized, such as resilience, availability,
and throughput. Edge computing can reduce latency and bandwidth constraints of services by not
transferring collected data to the cloud or a DC for processing and thus not needing to remotely
retrieve subsequently generated information. Most recently, the edge has also become a frequent
deployment target for control logic supporting industrial automation and machine learning (ML)
models used in visual analytics tasks.
By shortening the distance between devices and the computational resources that serve them, the
edge brings new value to existing use cases and can introduce new classes of applications. This
results in distributing workloads and ML assets southbound along the path between today’s
centralized DCs and the increasingly large number of deployed edge computing devices and clusters
in the field, on both the service provider (SP) and user sides of the last mile network – the portion of
the SP network that reaches user premises.
NOTE
When the terms “southbound” and “northbound” are used when discussing an application architecture, they refer to points
of the compass in reference to the relative location of your current point of view. So, “northbound” would refer to services,
tiers, and locations that are more physically proximate to the cloud, or, in the case of Figure 1.1, locations to the right-hand
side. Likewise, “southbound” refers to locations closer to the user edge, or locations on the left-hand side.
Notice in the preceding diagram, which we will be using as a starting point for many charts used
throughout the book, how all computing resources located to the left of the thick, black line labeled
Internet Edge would be considered the edge, and all computing resources to the right of the line
would be considered the cloud. Despite those designations, you might hear about “clouds” located in
the SP edge referred to as “regional clouds” or “enterprise clouds” and in the user edge as “edge
clouds.” Future iterations of the chart will collapse the Internet Edge and Last Mile Networks lines.
Edge computing, therefore, is described by ways in which its operating environments are different
than the physical security, hardware homogeneity, and service scalability of cloud computing. Edge
compute nodes (and by nodes, we include both standalone devices and compute clusters) can be
solitary, and many times are not rack-mounted. Edge nodes may not have reliable or consistent
power, network connectivity, or even air filtering, climate control, and controlled physical access.
Multiplicities of edge nodes may not be the same version or brand of hardware, with differing
specifications and capacities, and thus edge computing nodes are described as heterogeneous. Edge
nodes may use smaller or fewer processors, slower and/or more power-efficient processors, and fewer
specialty or co-processors to accelerate specific types of tasks or workloads. Last, edge nodes may
have a permanent or fixed placement or might have mobility or otherwise be portable.
Moving on to the contents of edge nodes … the type of chip being used, the micro-architecture, could
be mounted in a constrained device or an off-the-shelf, commodity unit, but it typically runs a Linux
distribution or similar enterprise- or consumer-class operating system. We do not speak of embedded
systems or fixed-function devices of the IoT class as being used for edge computing functions,
although they certainly are used to send data northbound to edge computing systems.
EDGE MICRO-ARCHITECTURES
Typical chip micro-architectures supported by most edge computing solutions include:
- x86_64 or amd64
- ppc64le
- risc-v
When writing and packaging applications for the edge, we no longer write an application in a high-
level language such as Python, NodeJS, or even Golang, and package it up for a package delivery
system such as pip, npm, and others. Instead, we typically containerize the application to make it self-
contained along with all its dependencies so that it doesn’t need to be installed. A container image is
downloaded from a registry and run in a container engine such as Docker or Podman. There are also
common techniques available to support multi-arch containers that will build and run on all the
common micro-architectures listed previously, which is the approach we recommend using. See the
following article for more information:
https://developers.redhat.com/articles/2021/08/26/introduction-nodejs-reference-architecture-part-5-
building-good-containers.
NOTE: Containers are not the only edge-native approach for isolating workloads. Enterprises may
use virtual machines (VMs), serverless functions, or even WebAssembly (Wasm) depending on the
code-base purpose, or execution environment. Regardless of the chosen approach, proper automation
should be employed to ensure isolation is maintained.
Figure 1.2 – The edge continuum with trade-offs shown at the bottom with gray arrows
Jeff Ready, former CEO of Scale Computing, has a pithy way of contrasting the edge with the cloud:
“The edge is really the inverse of the data center. A data center deployment is likely 1 or a small
number of physical locations, hundreds of servers at those locations. The edge, on the other hand, is
hundreds or thousands of locations, with often 1-3 servers at each one. You can’t manage them the
same way. You’re still deploying thousands of servers, but with many, many locations you obviously
can’t have a team of IT pros at each one like you would in a datacenter. You need automated
deployment, automated management, automated error recovery to make the edge work.”
(https://blocksandfiles.com/2023/02/02/dell-vxrail-edge/)
Edge computing environments are very different from, and in most cases filling requirements that are
the direct opposite of, cloud computing. As Eclipse’s Mike Milinkovich has said: “If you care about
the physical location of your devices, then you are doing edge computing.” However, edge computing
has been established on a foundation of software development processes informed by cloud-native
development best practices. In short, edge computing would not be possible if it weren’t for the
cloud.
How does edge computing bring value, and why now?
Edge computing reuses applicable cloud computing programming best practices, which give it a
standard approach to software development that is fast, flexible, and works across multiple
architectures. This methodology provides small teams with minimal cross-architecture experience in
a way to create cross-platform distributed applications comprised of multiple loosely coupled
services. This is what powers edge computing (and cheap computing).
Edge computing came about at a time when inexpensive but powerful compute became plentiful and
custom fabrication tools more widely available. The Raspberry Pi single-board computer introduced
ARM-based processors to hobbyists around the world at affordable prices while also spawning a
large ecosystem of software utilities and hardware add-on boards. Since these systems could run
many common Linux variants, they also formed the basis for proofs of concept (POCs) that could
be easily turned into commercially viable solutions.
We’re now beginning to see a similar wave of innovation with RISC-V-based systems that will
further enable low-powered and efficient solutions that could even be embedded into standard
hardware components. This would bring us to a point where computers with a Linux operating
system that are capable of running containerized workloads could be powering every household
appliance and consumer device or component. For example, Intensivate is running containers on the
RISC-V ISA-compatible SoC that controls SSD drives.
By virtue of having inexpensive but powerful compute available and placed adjacent to where data is
being generated, and being able to program that compute using existing tools and methods, you can
simultaneously reduce the cost of computation while decreasing response times and reducing latency.
Complex analytics no longer require offloading to the cloud, but ultimately, the available trade-offs
largely depend on which edge you choose for workload placement.
But, the far edge is not the only location where edge computing takes place. The Linux Foundation’s
LF Edge organization refers to all edge computing locations falling after the last mile as belonging to
the user edge, which follows a nomenclature categorizing types of computing by the owner of that
compute. The fundamental assumption is that infrastructure at these locations is typically not shared
beyond a single organization, business, or person. The Eclipse Foundation’s Edge Native Working
Group terms it Field Deployed while seeing the last mile and its associated infrastructure collectively
as Access-Transport, as shown in Figure 1.4:
Figure 1.4 – Eclipse Foundation Edge Native Working Group terms for edge
As the field of IoT computing was maturing, but before edge computing had gained widespread
adoption, Cisco and others began using the term fog computing to refer to a method of distributing
key cloud-like computing and storage infrastructure and services outside of the cloud and closer to
devices. The term never entered widespread usage and was soon supplanted by the more general term
edge computing to cover all computing outside of the cloud on programmable devices.
And lastly, you have traditional DCs. These are locations where physical access is controlled, racks
of homogenous compute hardware are provided and remotely manageable, and both Platform-as-a-
Service (PaaS) and Infrastructure-as-a-Service (IaaS) are offered. These DCs are typically owned
and used by a single party, but mostly only differ from the cloud or global compute by size and scale.
Large DCs such as these are rarely referred to as locations where edge computing happens; however,
they are technically part of the edge as long as they reside outside of the internet’s network.
Before edge computing devices made this possible, a solution would have been built to stream video
to the cloud where inferencing would have been performed. By the time the manager would have
been informed, minutes would have passed, and the individuals would likely no longer be in the area.
Edge computing removed the expense of transporting video feeds to the cloud, reduced the resulting
inferencing latency, eliminated any cloud computing costs, and ultimately ensured that the manager
was notified up to 3 or more minutes sooner.
Before the cloud, the factory would have sent closed-circuit television (CCTV) feeds to a
monitoring location where one or more persons would have viewed a bank of screens looking for
issues on low-resolution displays, and called a manager if they spotted any issues. This approach
would have been even more expensive and slow, and thus only likely to have been used to prevent
major losses or accidents, or recorded and reviewed by investigators at a later time to determine
potential causes of an accident.
With those capabilities, they are used during store hours for spill detection or traffic counting when
pointed at an aisle, dwell time when pointed at an end cap, and shelf restocking when pointed at a
row. After the store is closed, those applications are replaced by a security application that looks for
the presence of persons when the location should be unoccupied.
Edge computing makes these capabilities possible at an operating cost of pennies per day and without
needing more connectivity than a local network connection.
Hosted edge nodes are ideal for supplementing customer edge workloads temporarily while also
avoiding the vendor lock-in of proprietary cloud solutions. The drawback to using hosted nodes is
that the customer will need to provide any infrastructure and platform services and support. But it
makes for an excellent extension or overflow to existing customer DCs or for situations when a
company outgrows its existing locations but has not yet secured new facilities.
It also works well for temporary events, especially when they take place in a specific or limited
geographical area. To that end, there is even a start-up that will deliver a self-contained edge DC in a
container to your location for as long as you need it.
Develop in the most appropriate language and/or framework for the purpose
Deliver a clean separation of concerns (SoC), especially between stateless and stateful services
Abstract from underlying dependencies, both hardware and operating system
Source: https://thenewstack.io/cloud-native/10-key-attributes-of-cloud-native-applications/
Looking at the preceding summary, a good example of the cloud-out approach to application
development would be building a product using the SaaS model. SaaS is a way of building and
delivering software so that it does not need to be installed (it is hosted by someone, usually in the
cloud), and it does not need to be purchased (it is paid by subscription). Access can be provided in a
browser or over the internet via APIs.
SaaS solutions demonstrate the cloud-out approach because they are typically hosted in the cloud,
implemented with microservices, abstracted from dependencies, and deployed on elastic
infrastructure. Individual SaaS offerings may also follow many other cloud-native best practices, but
compliance with those principles is not obvious without access to the source code.
SaaS solutions clearly do not demonstrate the edge-in approach because they are not typically built to
handle dependency unavailability (hence built with highly available architectural principles), service
portability, target system constraints (since they do not need to be installed or remotely deployed),
and dependence on orchestration. Additionally, they presume an always-on network connection with
low latency and high throughput.
So the pros of the SaaS architecture approach are that the application and its constituent services do
not need to be built for wide micro-architecture compatibility and thus can be narrowly tailored to the
deployment target’s specific hardware requirements. It can rely on the hosting facility provider to
maintain and support the infrastructure, platform, and connectivity as well as the facilities
themselves. And within reason, they can scale horizontally up to the provider’s available capacity.
The cons of this approach are that the application and services may be narrowly tailored to the
environment and are thus not inherently portable, and the risk of vendor lock-in is considerable. This
may also render the application brittle when exposed to new or unanticipated conditions, thus
requiring more ongoing maintenance than if it had been built using edge-native principles. Finally, a
SaaS service will clearly be unavailable in the absence of a network connection.
Employ external application configuration and secrets per instance, including support for runtime updates
Source: https://www.ibm.com/docs/en/eam/4.5?topic=clusters-edge-native-development-best-
practices
Based on the preceding points, a good example of an application following this approach would be
one or more multi-arch containerized gardening applications running on a Raspberry Pi, which is
connected to moisture, light, humidity, and temperature sensors. The application can retrieve data
from the sensors, persist it locally, and make it available through a web application displaying a page
with the data or making it available through a REST API as a data payload.
This architecture follows edge-in principles due to its ability to collect data from a sensor while it is
connected, to provide the collected data while it maintains a network connection, and to resume
operation when rebooted or after an interruption of power. The applications are portable and can be
run on multiple architectures, providing that sensors are present on that machine. They do not require
the orchestration that Kubernetes provides. They do not require services residing on another machine;
the data stays resident on the machine itself unless manually moved.
The application architecture does not demonstrate a cloud-out approach because it may not be
abstracted from the underlying hardware dependencies (the sensors), and it does not use elastic
infrastructure to scale.
So, the benefits of this approach are that the solution is resilient to adverse conditions and can
function fine without an internet or network condition, being completely self-contained. It also runs
on inexpensive hardware and would cost pennies in electricity to operate monthly. And when the
hardware eventually fails, it is cheap to replace.
Wrapping up, you’ve seen the differences between cloud-out and edge-in architectural principles,
compared how they work in the real world and read about the pros and cons. Now, let’s discuss the
larger patterns that most applications follow on the edge … the archetypes.
Introducing archetype patterns
In this section, we introduce you to the concept of archetypes. Along the way, we cover the days
involved in the software life cycle and discuss deployment methods. By the end, you will be ready to
learn about and use archetype patterns.
What is an archetype?
An archetype is the original model or form of something that embodies all primary qualities of that
item, whether that item is an abstract concept or a physical object. In the case of architectural
patterns, our assertion is that most application architectures can be derived from an original pattern
archetype (source) or a slight variation based on local or business considerations. Therefore, in this
book, we attempt to tease out archetypes and discuss them at length based on our belief that
identifying and mastering them will give you the foundational skills needed to tackle most, if not all,
edge application scenarios. By following these patterns judiciously, you will be able to create
solutions that are not only portable but also future-proof.
In the archetype-pattern diagrams to be found in this book, we will follow certain conventions:
Show all elements of equal size and shape so as to imply that all components are of equal importance
Refer to a component by its function or role instead of a product name, which emphasizes replaceability and a vendor-neutral
approach
Indicate placement by edge category as denoted in columns ranging from the cloud on the right to the field-deployed far edge on
the left
Draw rectangles around edge nodes to indicate placement of components, applications, and services
Connect edge devices, nodes, systems, platforms, and infrastructure with arrows indicating the direction of data flow
Once you settle on an architecture pattern on day 0, based on the best fit to the requirements and
business needs, the next consideration you need to address is how to deploy that architecture … both
on day 1 and maintaining it on days 2-N. Let’s delve into the idea of days 0-N, which will also be
used in Chapter 7 when we discuss applying automation to these deployments.
Day 1 denotes programming, provisioning, and configuration of environments and pipelines. That would also include dependency
installation, finalizing automation, and all unit and end-to-end (E2E) testing. At the end of this period, the application should be
documented, approved by all relevant parties, and ready for launch into a production environment.
Day 2 marks the support period where time is spent working on issues, optimization, A/B testing of new features, blue/green
testing of differing versions, and so on. This time should also be used to build up a support database, any FAQs, and possibly
training support chatbots or other automated response mechanisms.
Day N (end) would then be for activities related to the retirement of an application and its provisioned environment, assets, and
supporting infrastructure. By the end of this period, all traces of the application should be removed except for any items of
historical value or required to be kept intact for legal retention purposes:
Figure 1.5 – Depiction of days 0-N categories with major activities shown in each
But when it comes to field-deployed solutions at the far edge, potential connectivity issues and
heterogeneous systems, combined with a lack of infrastructure services, will require a different
approach. We’ll go into greater detail in Chapters 9 and 11, but using deployment and application
life-cycle management tools that can operate autonomously and pulling configuration from a central
control plane rather than pushing to the edge tend to resolve most of those deployment issues,
including scalability and security.
Summary
In this chapter, you learned about what capabilities the edge and the cloud have in common, and what
distinguishes one from the other when it comes to the tiers of compute: infrastructure, platforms,
services, and applications.
We covered the names and characteristics of edge categories and described the benefits and
drawbacks of each. We also discussed various ways that edge nodes can be shared between tasks. We
described how an architecture can be scaled based on its scope and responsibilities with cloud-out
and edge-in paradigms. Also, we introduced you to the concept of archetypes and the “days”
involved in the software development life cycle.
Now, you should be ready to learn about the basic components and building blocks that go into
archetype patterns. Just as importantly, we’ll also discuss how to approach solutions from the right
perspective in order to build a future-proof application, follow best practices, and use long-term
thinking.
2
These solution architectures are unique because there are limitations at each layer, from device to
compute to storage. Architects designing them often must think about the limitations, especially
when it comes to the far-edge aspect. One must keep in mind the intrinsic benefits of edge computing
such as low latency, high performance, less power consumption, high bandwidth, and multiple
dispersed locations. Edge computing has given rise to a new paradigm of application architecture
specifically designed to run in the distributed edge domain, which we call edge-native applications.
This chapter describes the four major roles of the components in an edge architecture. We then talk
about the common functional and non-functional requirements (NFRs) and discuss the software
and hardware components that commonly go into the creation of edge architectures. It concludes with
a discussion of device architectures, data transmission protocols, and architectural decisions. The
main topics are as follows:
Edge components
Functional requirements
Architectural decisions
Edge components
There are four major roles for the edge components in an enterprise’s edge computing architecture:
edge devices, the edge gateway, or server in the enterprise edge (part of the user edge’s field
deployed compute), the micro data center in the Service Provider Edge’s Regional Compute, and the
enterprise cloud. The edge server not only acts as a gateway to connect all edge devices in a secure
manner but also allows for the management of all those devices. See Figure 2.1:
Figure 2.1 – Common representation of the edges in edge computing
The enterprise cloud is shown on the far right. This could be a public, private, or hybrid cloud, which
is the domain of hyperscalers. To the left of it is the realm of regional compute, where the telcos or
communications service providers (CSPs) operate. Next to it is the user edge (in this case, the user
could be an enterprise) where the edge clusters and gateways are deployed. On the far left, IoT and
edge devices are shown, including sensors, gauges, cameras, robots, and the like.
At a macro level, there are four aspects to an edge computing solution and you will see them
reflected in the architecture diagrams that follow in this chapter. They are as follows:
Edge devices: While some IoT devices such as sensors, gauges, and actuators cannot run any software, many edge devices have
some computing power and some storage. That lets them store some data and run some simple analytics. Depending on their form
factor, certain edge devices have enough compute, memory ranging from 128 MB to 256 MB, and almost 1 GB of storage, which
is enough to analyze the data and perform real-time inferencing without needing to send the data to a backend server or the cloud.
That is what edge-native computing is all about. If they are not using ARM architecture, the devices could be x86 class CPUs
equipped with one or two cores.
Note that now it is possible to deploy commodity AI accelerators connected to a USB port to
supplement their inferencing and analytics capabilities.
Edge servers: The other components in the user edge space are edge servers or edge nodes. There is a one-to-many relationship
between an edge server and edge devices. Edge servers or gateways are constantly in touch with edge devices by way of agents
running on the devices and are used to deploy applications onto those devices. These are typically commercial off-the-shelf
(COTS) computers that could be located in a distributed facility such as a factory floor, store backroom, warehouse, or remote
office. These could be ruggedized or placed in a protective enclosure. The small-sized machines have 8 cores, the medium-sized
machines have 16 cores, and anything with more compute capacity would constitute a large machine. The memory in these
machines starts at 16 GB RAM and they could have hundreds of gigabytes of storage. If inferencing at the device is insufficient,
then data from the far-edge devices is sent to the edge server or even to the cloud for further analysis and deeper insights.
Regional compute or service provider edge: This edge is sometimes also referred to as the network edge or micro data center.
CSPs are taking advantage of newer networking technologies to create these regional clouds or local clouds that provide software-
based infrastructure services for devices to communicate with at the far edge of the network. The major selling point of the telcos
is that data from the edge devices does not have to be sent to the cloud but can reside in this regional cloud, thereby reducing the
distance and time that data must travel. For the end user, it means decreased latency, better bandwidth, and more security. This is
especially true with the use of 5G.
Enterprise cloud: This is the centralized cloud that could be a public or private cloud or an on-premises data center. As is
common to clouds, enterprises get unlimited compute and storage along with management capabilities, plus access to a growing
portfolio of other cloud services. From an edge computing perspective, this is home to four facets: storage of most of the device
data, device management at a global level, AI model building and training, and enterprise-level analytics.
Now that we have seen the major components that one finds in edge computing solutions, the next
thing is to dive into some of the functional requirements that enterprises ask for.
Functional requirements
In Chapter 1, we talked about cloud-out and edge-in paradigms. Cloud-out is where computing is
taken out of the data center and brought to the far edges of the network. Conversely, the movement of
the generated data from the source or the edge to a location with more computational resources for
analysis is the edge-in part. Those facets drive functional requirements in an edge computing
solution. We will discuss the common functional requirements of an edge computing solution in the
subsequent sections.
Sensing
This is where we still deal with traditional sensors (e.g., IoT/accelerometers, thermometers, or
actuators) as architectural components that acquire data and help create a signal. When combining
technologies such as edge and artificial intelligence (AI), you introduce new ways of designing and
deploying technology, thus improving and automating situational awareness with sense-making
systems. These are deployed in stores, shop floors, industrial equipment, mines, and even in vehicles.
They are often referred to as systems that have situational awareness. These sense-making systems
fuse human-like thinking with sensing technologies so they can take actionable insights to augment
humans in their work or help humans. Additionally, such systems can support augmented reality
(AR) deployments by providing data that will be “visible” to humans working nearby. That is the
core idea of these edge computing patterns. Businesses that want to create edge/IoT solutions should
have a good understanding of their assets, especially their people, because that helps them in making
better decisions while incorporating this newly enabled sensing technology.
Inferencing
Inferencing, by definition, is the act of reasoning from factual knowledge. Inferencing at the edge
means providing actionable intelligence using AI-powered techniques based on the data gathered by
different types of devices, such as sensors, cameras, microphones, and so on. One of the outcomes of
Another random document with
no related content on Scribd:
would altogether go from us. It is through you that we are still
favoured with his countenance occasionally.”
Alice’s tearful eyes had besought mercy for her parent long before
the queen seemed disposed to yield it. While the adherence of the
noblesse to the royal cause was regarded as a matter of course, and
therefore not rewarded with extraordinary gratitude, all symptoms of
halting or defection were observed with scorn, and commented on
without reserve by the haughty woman who regarded her rank and
empire as natural, instead of conventional, and would as soon have
dreamed of being denied the use of her limbs and senses as the
privileges of royalty.
It was through her influence that the king refused to sign the
declaration till the last moment,—when he was compelled to do so at
a tremendous sacrifice of regal dignity;—at the bidding, namely, of
twelve poissardes who forced their way into the presence with the
deputies from the Assembly, and under the compulsion of threats of
what might be expected from the army of eighteen thousand men
who had marched from Paris during the afternoon, under the
enforced command of Lafayette.
Never was anything beheld more dreary than the aspect, more
disgusting than the incidents of this day and night. The skies frowned
upon the scene, and wind and rain added to the difficulty of what
was achieved, and the horror of what was witnessed. The deputies
and their attendants, the poissardes, appeared in the king’s
presence, covered with mud and drenched with rain; the House of
Assembly was crowded with women, who came in for shelter, taking
their seats among the members, now eating and drinking, and now
lifting up an outcry to drown the voice of an unpopular deputy; the
fires of the bivouacs in the streets were quenched with torrents of
rain, again and again, and the peaceable inhabitants were in fear of
being compelled at length to throw open their gates to the rabble.
The leading figure of the mob, however, had a peculiar reason for
disliking the weather, as he took care to show everybody. He was a
gaunt-looking ruffian, with a high pointed cap, and grotesque garb,
well armed, but especially proud of an axe which he carried, ready
for immediate use at the slightest hint from the leaders of the mob.
With all his fear,—the only fear he seemed capable of,—that it
should be rusted with the wet, and he thus delayed in his vocation,
he could not refrain from brandishing it over his head, and displaying
it in sight of the sentinels, and such of the body-guards as looked out
now and then from the palace. This ruffian took his stand
immediately under the king’s window, prepared a cannon as a
convenient block, and waited impatiently for victims. He could not be
persuaded to quit his post for shelter; but he did once step aside for
brandy. On his return, he found two poissardes sitting astride on his
cannon, face to face, tossing off their drams, and devouring the
rations which their prompting demon had taken care to provide. The
executioner warned them off, and prevailed by the offer of a better
seat within five minutes. A hint was enough to show them his
meaning. He just pointed towards an approaching group, consisting
of an unfortunate soldier with whom some of the mob had picked a
quarrel as he was going to his post for the night,—and his captors.
The victim looked dogged. He saw the cold metal block on which the
axe was presently to ring his death-stroke: he saw the fidgetty
executioner, and the fierce women, gathering round, munching their
suppers as if his life-blood was the draught they looked for to wash
down the last mouthful: he saw that no help was within reach or call.
He saw all this, and seemed disposed to take quietly, though
sullenly, what was inevitable. He stood firm while they pulled off his
stock; he moved forwards when they pushed him; he kneeled when
they pressed upon his shoulders; but some impatience in their
manner of doing so excited his passions in a moment to their utmost
strength. Before they could keep him down, he was not only on his
feet again, but bounding high in the air, grappling with the
executioner for the axe, kicking, trampling, buffeting all who laid
hands on him, and creating a hubbub which brought the king to the
window above, and conveyed to the senses of the ladies a
knowledge of what was passing. It was a short struggle; but a
struggle it was to the last, and force alone could subdue the victim.
One virago clutched the hair of his head, and others held down his
feet. When his blood flowed on the ground, and mixed with the
puddles of rain, one or two stooped down to see how the eyes rolled
and the nostrils yet quivered, while, on the other side the block, the
executioner, mindful of his promise, tossed the headless body to a
little distance, so that his friends might sit on it to finish their meal.
What are the invisible issues of life there was no one present to
think, during the whole scene, unless the victim himself might have
been conscious of his thoughts darting that way; but such was the
visible issue of a life which a stupendous and delicate natural
apparatus had been appointed to create, sustain, and develop. It had
originated in the deepest passions of human nature; been
maintained by appliances, both natural and moral, which the keenest
powers can barely recognize, and not estimate; and developed for
objects of which man has only the remotest ken. Such was the
visible issue of this mighty series of operations. That the handy work
of Providence should ever have been thus crushed, and its
mysteries thus boldly made sport of, may in time appear as
incredible as it would now seem that children had ever been
encouraged to pull planets from their spheres in mockery, and
quench the milky way,—supposing such power to have been left in
their hands. In the latter case, who would be answerable for the
profanation? Surely those who taught mockery in the place of
reverence. Who then was answerable in the former case? Those
who made the perpetrators ignorant through oppression, and savage
by misrule. The responsibilities of a certain order through many
centuries were called to judgment during the brief period before us;
and the sentence of condemnation not only went forth on the four
winds to the farthest corners of the globe, but shall be repeated
down to those remote ages when it shall be forgotten on earth,
though recorded in heaven, that man ever shed the blood of man.
One or two more such murders on the cannon and at the palace
gates had not the effect of alarming the court or the really patriotic
leaders of the people so far as to keep them on the watch through
the night. The king believed that all was safe when he had given the
signature which it was the professed object of the expedition to
obtain. The queen was assured by Lafayette that the people were
wearied, and that nothing was to be apprehended till morning; and
the general himself reposed in his hotel in full confidence of the
security of all parties. All were not, however, thus satisfied. Some of
the deputies refused to withdraw from their chamber; and while all
was sleep and silence in the palace, except where some watchful
ear caught the soft tread of the sentinels in the corridors, and the
pattering of the rain without, and at intervals, some tidings from the
passing gust, of revelry in the streets,—while armed ruffians sang
their songs, or snored in their dreams round the watch-fires where
the shrill-voiced poissardes were broiling their rations, or heating
their strong liquors,—a few of the wiser deputies sat, each in his
place, with folded arms, and in perfect silence, while the light of a
single lamp fell on their uncovered heads and thoughtful
countenances, and foresight was invisibly presenting to each
pictures of that which was about to befall their monarchy and
themselves. Revellers, legislators, and sentinels were not the only
ones who watched. One or two, who did not partake the general’s
confidence in the people thus strangely congregated, wandered from
watch-fire to watch-fire, and about the precincts of the palace, to be
in readiness to warn Lafayette of the first symptoms of movement.
Among these was Charles, whose anxiety had been awakened by
the aspect of Paris after the departure of the army for Versailles. It
was well known that Lafayette’s generalship on this occasion was
enforced; and not all the apparatus of triumph amidst which the
troops marched out,—not all the drumming, and military music, and
display of flags amidst the rain, and echo of shouting heard above
the strong winds, could remove the impression of the hollowness of
all this rejoicing,—the desperation of this defiance. When the sights
and sounds were gone, a deep gloom settled down upon Paris. The
shops were shut, the streets were silent, except where the waggons,
laden with meat, bread, and brandy, converged towards the
Versailles road, or where groups of two or three observers whispered
their anticipations to each other, mindful of none but political storms,
and questioning only whether the sun of royalty would not this night
have a crimson setting, to rise upon their state no more.
Charles had been among these observers, and the tidings he
brought home made his wife anxious to depart from this
revolutionary city, and take refuge in their country possessions. She
would be ready to go at any moment, she declared, and when would
there be so favourable a time as when the place was half emptied of
its inhabitants, the police otherwise engaged than in watching the
proceedings of private individuals, and all fear at an end of any
attack upon the wine-stores? Charles was half disposed to listen to
the scheme, though his views of what was likely to happen differed
as widely from his wife’s as the prevailing tone of mind by which they
were influenced. Marguerite feared the worst: her husband hoped all
might yet be well, and thought it, at all events, a good thing that
something decided must arise out of the present crisis. He
determined to follow the march to Versailles, and to return as soon
as he could anticipate the event, to bid Marguerite stay and make
herself easy, or to carry her, her father and children into Guienne.
While she was packing up the few necessaries she meant to take
with her, and persuading all the household but herself to go early to
rest, Charles was reconnoitring the proceedings of such as were
preparing a terrible retribution for those under whose tyranny they
had suffered.
He was no spy; being devoted to no party, and acting for his own
honest purposes; and he therefore used no concealment. He
conversed with the riotous poissardes on public injuries, conferred
with the deputies on public order, and exchanged a few words with
the sentinels on the probability of an attack on the palace in the
morning. The horrible threats breathed over the fires against the
queen, the brutal exultation which appeared through mysterious and
slang expressions respecting the royal household, made him wonder
at the apparent defencelessness of the palace. He was by no means
satisfied that all was safe till morning, and said so to a little muffled
up man whom he found standing in the shadow, close by the great
iron gates. He could not make out whether this man was a mere
looker on, like himself, or a watch appointed by either party.
“Is it your own choice to be out to-night, sir; or are you occupying a
post? Because, in the first case I would direct you where you might
see more of the state of things than here; and, in the other, I would
strongly recommend your appealing to the general for support.”
“Alas! yes. I am sent hither,” replied the quavering voice of the
muffled up person. “None would willingly be abroad this night, and all
my desire is to be left unobserved in this shadow at present;—
unless, indeed, some friend should pass who might protect me, and
from whom I might learn that which I am sent to ascertain.—You
seem, sir, to be an orderly, honest man. Can you tell me whether the
duke,—whether Orleans is at hand?”
“Orleans being the most honest and orderly of men, hey,
marquis?” said Charles, laughing. “So you are sent out by lady Alice
for tidings, and you wait here for them till Orleans passes by—Is it
not so?”
“Ah! what can I do? These canaille will smother me again with
flour, or drag me to the cold cannon;” and here the little man
shivered, and his teeth chattered. “Do but bring me to Orleans, my
good sir, or get me a re-entrance into the palace, and I will——I will
——This morning air is so raw! and I am——I am——not fit for
enterprise.”
Charles fully agreed with him; but having no interest to get a royal
spy housed before his errand was done, he could only tell him that,
to the best of his belief, Orleans was lingering about the road
between Versailles and Paris, or hanging somewhere on the
outskirts of the encampment to witness the issue, without being
implicated.
“Ah! how he is happy in comparison with me!” cried the poor
marquis. “I have never, sir, meddled with politics——”
“Further than as all the noblesse have operated,” interrupted
Charles. “I mean in stimulating the people to meddle with politics.
You have wrought at second hand, marquis, hitherto. Now is your
time for taking your part finally, and acting in it.”
“Alas! what evils come of any one interfering in such affairs but
ministers and deputies! Let them act, and let us be neutral. This is all
I ask.”
“Aye, but, marquis, it is too late to ask this; because there has
been great mistake about what is, and what is not, being neutral. I
dare say you believed yourself neutral when you lay sleeping in bed,
while your peasantry were keeping the frogs quiet in your ponds. I
dare say you had no thought of politics in your boar hunts, or when
three fathers of families were sent to the galleys for alarming the lady
Alice’s brooding doves. Yet you were all the while——”
The marquis’s light blue eyes were now seen by the lamp light to
be opened upon Charles with such an expression of vacant
wonderment that it was plain there was no use in proceeding. He
evidently had yet to learn the true province of politics; and, for his
part, he thought the merchant must have drunk a little too deep in his
own wine, to be talking of peasants and pigeons in connexion with
an insurrection in Paris.—He would never have had courage to leave
his nook by himself; but now that he had met a face that he knew, it
required more courage to remain there by himself, and he therefore
hooked Charles by the arm, and said he would be wholly guided by
him. Charles would rather have dispensed with his attachment, but
could not shake off the old man into darkness and helplessness, if
he himself preferred venturing into the light of the watch-fires, and
upon the threshold of Lafayette’s lodging, whither he was warned he
would be conveyed.
If the marquis had carried a bold front, nothing would have
happened to him, any more than to his companion; but his slouched
hat, halting gait, and shrinking deportment at once drew attention
upon him. The consequence was that he heard double the number
of threats, and imprecations ten times more horrid than had met
Charles’s ears before. If he had now regained entrance into the
palace, he could have told that which would have made even the
queen’s fiery blood run cold, and have given the whole household a
foretaste of worse horrors than even those of the ensuing day.
When they had arrived at the last of the line of fires, the marquis
believed his purgatory to be nearly over, and indulged himself in a
few ejaculations of thankfulness on the occasion. He was overheard,
seized, dragged to the light, his coat torn open, and his hat pushed
back. The queue looked suspicious; the manner of speech, mixed
up, as even these people could perceive, of high breeding and
imbecility, gave assurance that he was a court adherent; to which
there was to be opposed only his own and Charles’s assurance that
he was a companion and friend of Orleans. The knot of drinkers
hesitated whether to cut off his head or let him go, and the marquis
stood panting with open lips and closed teeth, when an amiable
creature, partly masculine in her attire, and wholly so in her address,
proposed a half measure.
“If he is one of them,” she observed, “we shall find him again in the
palace presently; so let us mark him.”
With the word, she seized the poor man’s nose with the left hand,
a burning stick with the right, and branded his forehead with a cross;
then pushed him away, and turned to Charles, offering to drink to
him in his own liquor, the choicest in Paris, if Orleans said true. She
pointed at the same time to a waggon near, on which, to his
amazement, Charles saw piled wine-casks with his own mark, and
brandy-bottles sealed with his own seal.
Perceiving at a glance that his cellars must have been forced
since he left home, and that all further resistance would be useless,
he determined to yield to his wife’s desire to quit Paris; and he
hastened to discharge his duty of rousing and warning the general,
before turning his back on this scene of disorder.
Lafayette was up in a moment, and, though still trusting in the
peaceable disposition of the people, dressed himself hastily, that he
might be among them by daybreak. Before he could leave his hotel,
however, warning sounds came from the direction of the palace, and
messengers succeeded one another rapidly, stating that an attack
was being made on the great iron gates, that blood had already been
shed, and that the lives of the whole royal family seemed to be at the
people’s mercy. The general threw himself upon a horse which
happened to be standing saddled below, and galloped off, before
Charles could recommend the marquis de Thou to his protection,
should he happen to find him in the hands of the populace. His own
anxiety to get home was such as ill to brook any delay, and to admit
little other interest of any kind; but chance threw him once more in
the path of the old man.
As he was making the best of his way towards the Paris road,
stemming the tide of people that was rushing towards the palace, he
was suddenly jostled and thrown down by an impulse in the contrary
direction. Nor was he the only one. Many were bruised, some
trampled, while a fugitive burst through the throng, followed by a knot
of pursuers, who overthrew all that came in their way, while their
mingled curses and laughter contrasted strangely with the panting
cry of the pursued. Some cried out that it was the king; others
uttered imprecations against him as one of the hated guards; while
Charles saw, amidst his tattered, scared, and helpless condition, that
it was no other than the poor marquis. His desperation gave the
hunted man strength to clear the mob, and to fly some way beyond,
till he reached the trees of the avenue, where there was an end of
his safety unless some better aid was brought him than his own
failing strength. His enemies dogged him, surrounded him;—some
brandishing pitchforks, others large knives, and not a few firing off
their muskets to give a new impulse to his terror. This sight was
intolerable to Charles, who saw in such cruelty none of the
palliations which he had admitted in the case of some former acts of
violence. Forgetting all but what was before his eyes, he snatched a
pike, threw himself in front of the pursuit, reached the victim just as
he fell exhausted at the foot of a tree, and stood astride over him,
with one hand in an attitude of defence, while the other beckoned to
the people to listen. He shouted amid the din, and the few words
which were heard by those nearest to him served his purpose of
diverting their thoughts from immediate murder. He told them that, in
the name of the marquis’s tenantry, he demanded that the marquis
should be placed in the custody of the Assembly of deputies, to
answer for an infringement of the new laws by which the property of
the peasantry was protected. He told them that the general was
gone to the palace, to mediate between the queen and the
poissardes, and as it would be a pity that those who heard him
should be absent from so interesting a spectacle, he and one or two
more would take charge of the criminal, and convey him before the
sitting deputies. A well-timed roll of the drums and discharge of
musketry confirmed his appeal, and drew away his auditors, so that
in a few moments, when the last lingerers had gratified themselves
with pricking their victim a little with the points of their various
weapons, Charles found himself alone with the almost lifeless old
man.
On hearing that his further existence probably depended on his
reaching the assembly while the mob was engaged elsewhere, the
marquis made an effort to rise and walk, and found himself so much
less hurt than frightened that he accomplished the transit with small
difficulty. Such a deplorable object was never before presented to
the Assembly, at least under the title of a marquis. He had scarcely a
shred of clothing under the soldier’s cloak which Charles had
borrowed from a sentinel at the door. His powdered hair was dripping
with rain, and his face smeared with blood. He wept bitterly;
murmuring, in the tones of a woman, his wonder as to what he could
have ever done to offend the people, and how the world could have
grown so cruel and ungrateful. The Assembly had little leisure at this
time, and were glad to accept Charles’s offer of conveying the
prisoner away, and his guarantee that the marquis should set out for
his estate in the provinces without delay, and not return till the
troubles of the capital were at an end. The marquis was little
disposed to make opposition.
“Take me away,” he said, “though I only fly from one doom to
another. You say my tenants are enraged against me; and I say that
they will drink my blood. The vile are sovereigns in these days, and
the noble have the knife at their throats from day to day. O, if they
had killed me under the tree, it would have been over; but now it is
still to come. O save me! Do not leave me! Make me your servant.
Employ me as you will; but do not let them kill me!”
Charles recommended that the old man should in fact travel into
Guienne as his servant, and take possession of his chateau or not,
according to the apparent disposition of the peasantry when they
should arrive.—Not a moment was to be lost in proceeding to Paris,
if the departure of the family was to take place while the populace
and the troops were engaged at Versailles, and the whole attention
of the magistracy was directed upon what was passing there.
An empty cart was found in which to stow the marquis, while his
protector walked by its side. They left behind them the most fearful
spectacles of that day,—the murder of some of the guards, the
narrow escape of the queen, the brutal joy of the mob at the
enforced consent of the royal family to be conveyed to Paris, and the
beginning of that dreadful march itself, as anomalous, as disgusting,
as any spectacle that was ever presented as a pageant. But, one
circumstance which signalized that march, they were also witnesses
to. Half-way between Versailles and Paris, on a mound planted with
trees, a figure was seen, moving behind the stems, and peeping
forth at every sound of wheels or footsteps. It was Orleans, who had
stationed himself here to watch the issue of his plot,—the return of
the expedition, with the bodies of his royal cousins, dead or alive.
With some difficulty, he was persuaded to come down and speak
to his humble servant the marquis; and when he did greet him, it was
with something very like a smile at his crest-fallen appearance, and
querulous complainings.
“My good friend, these are strange times,” he observed. “I should
think your valet has hardly had time to attend to you this morning.
However, you will find plenty unoccupied at Paris to renew your
powder.—O, you wish to go at once, and shoot on your own territory.
Well; perhaps you are wise, since our kind of shooting here is not
exactly to your taste. You must take care, however; for I hear that
more bullets fly from behind the hedges there than in the open fields.
Farewell, my dear sir, for I see your companion is impatient. He
wants to be keeping guard over his wine-cellars. I wish him an ample
fortune out of the wines therein contained at this moment.”
Charles’s impatience was not only on account of his own affairs.
He distrusted Orleans so far as to be vexed that the marquis
whispered to him their plan of escape. There was no particular sign
of interest in the duke’s countenance at the relation; and it only
remained to be hoped that no harm would come of this unnecessary
confidence. The marquis was far from thinking it unnecessary, as a
word from the duke would procure passports for the whole family.
This word Orleans was prevailed upon to write, and furnished with it,
the marquis poured out his gratitude more vehemently than, but a
few months before, he would have supposed possible; and then
bade his vehicle proceed, watching from a distance how the duke
once more passed the enclosure, and took his station among the
trees as before.
The cellars were found to be indeed more than half emptied; and
of the casks that remained, one or two were staved, to drown the
gunpowder and other combustibles. No attack had been made upon
the house, and Marguerite had sufficiently got the better of her
terrors, to be ready for immediate departure. No obstacle arose, and
Steele, with Pierre under him, consented to remain in charge of the
property till Charles could return, after having deposited his family in
security.
The marquis made a rather singular-looking valet, with a manner
alternating between superciliousness and awkward deference,—a
strutting gait when he forgot what he was about, and a cringing one
when he happened to cast a glance upon his dress. He passed
muster very well, however, as a battered old soldier turned valet; his
strut passing for regimental paces, and his cringe being ascribed to
the honourable wounds he was supposed to bear. M. Raucourt took
off the attention of all who might be disposed to make remarks, by
telling everybody that he was going to see his olive groves. The
party travelled with more speed than the dismal procession from
Versailles; so that before the royal family was mournfully ushered
into the Hotel de Ville at dusk, Charles and his household were some
leagues on their journey southwards.
Chapter VIII.
UPSHOT OF FEUDALISM.