COE 558Lecture 02Glossary
Glossary
Every term in E2C continuum and cloud computing, defined once and used the same way in every part. Each entry links to the slides where the idea appears.
- Terms
- 61
- Letters
- 18
A
- App Engine
Google's PaaS service for developing and hosting web applications, launched in April 2008 as Google's first developer platform (Gmail, a SaaS product, came earlier in 2004).
- Availability
Percentage of time a system can serve requests: uptime divided by uptime plus downtime.
B
- Block-level storage
IaaS storage that keeps data in blocks on SSD or hard disk.
- Broad network access
Capabilities are available and accessed over the network.
C
- Capital expenditure (CapEx)
Large investments in fixed assets over a longer period, such as servers, networking and data center housing.
- Cloud bursting
Hybrid cloud technique of offloading excess load from one cloud infrastructure to another for load balancing.
- Cloud layer
Centralized layer providing compute, storage and networking at large scale: HPC, big data and multi-cloud connectivity.
- Cloud service provider (CSP)
External company that offers cloud services to which parts of the IT value chain are outsourced.
- Community cloud
Cloud infrastructure for the exclusive use of a community of organizations with shared concerns such as mission, security or compliance.
- Compound SLA
Availability of services chained in a pipeline, the product of their individual SLAs, lower than any single one.
- Container
Isolated virtual environment produced by OS-level virtualization, sharing the host OS kernel without a hypervisor.
- Content delivery network (CDN)
Caches deployed close to viewers (mini data centers) to deliver content with minimal delay, e.g. Netflix.
- Control groups (cgroups)
Linux kernel feature that limits, prioritizes, accounts for and controls resource use (CPU, memory, I/O) of process groups.
D
- Data geo-restriction
Processing data within a specific geographic region to comply with legal or privacy requirements.
E
- Economies of scale
Cost advantage of building and running computing at very large, centralized scale; the reason the cloud cannot simply be moved closer to the client.
- Edge gateway
Device providing local processing for edge devices and connecting them to fog nodes.
- Edge layer
Layer on the physical boundary of the WAN that lets clients offload compute, process small data quickly, and acts as a gateway hiding the fog.
- Edge node
A device at the edge layer that processes its own sensor data in real time, e.g. each car in a vehicular network.
- Edge-to-cloud (E2C) continuum
A continuum of latencies between client and server, layered into edge (small), fog (medium) and cloud (large latency).
F
- File-level storage
IaaS storage that keeps data in files in a file system.
- Fog layer
Layer of mini data centers providing compute, storage and networking at critical points closer to the client, serving multiple edge layers.
- Fog node
A node of the fog layer, such as a mini data center or a roadside unit, that processes local data and coordinates regionally between edge devices and the cloud.
- Full virtualization
Virtualization in which an unmodified guest OS runs on fully virtualized hardware, on a bare-metal or a hosted hypervisor; slide 20 wrongly ties it to Type 2 only.
H
- Hardware virtualization
Sharing hardware resources (CPU, RAM, networking) between isolated OS instances through a VMM that maps virtualized resources to physical ones.
- Hybrid cloud
Composition of two or more independent cloud infrastructures connected by technology that enables data and application portability.
- Hyperscaler
One of the largest public cloud providers; the slides name AWS, Google Cloud Platform and Microsoft Azure.
I
- Infrastructure as a Service (IaaS)
Provider manages data center through virtualization; customer manages OS, databases, security and applications.
L
M
- Measured service
Resources are automatically controlled and optimized with metering.
- Mini data center
A scaled-down version of a big data center bridging the gap between edge and cloud.
N
- Namespace
Linux kernel feature giving processes an isolated view of resources such as process IDs, networks and filesystems.
- NIST cloud definition
Model for ubiquitous, convenient, on-demand network access to a shared pool of configurable resources rapidly provisioned and released with minimal management effort.
O
- Object storage
Storage that keeps data as unstructured objects, as opposed to block-level or file-level storage.
- On-demand self-service
Consumers can provision computing capabilities automatically without provider interaction.
- OpenNebula
Open-source platform for managing virtual machines on distributed infrastructures, started in 2005, first released in 2008.
- OpenStack
Open-source IaaS platform started in 2010 by Rackspace and NASA, stewarded by a neutral foundation since 2012 (renamed OpenInfra in 2020, part of the Linux Foundation since 2025).
- Operational expenditure (OpEx)
Costs of day-to-day operations such as power, cooling, rent, routine maintenance, salaries and cloud subscription fees.
- OS-level virtualization
Virtualization without a hypervisor where environments share the host OS through the system call interface.
P
- Paravirtualization
Virtualization in which the guest OS kernel is modified to call the hypervisor directly; independent of where the hypervisor runs, although slide 20 ties it to Type 1.
- Platform as a Service (PaaS)
Provider supplies the runtime and development environment and all underlying layers; customer manages only applications.
- Private cloud
Cloud infrastructure operated for the exclusive use of a single organization, on or off its premises.
- Public cloud
Cloud infrastructure for open use by the general public, owned and operated by a third-party provider on its premises.
R
- Rapid elasticity
Capabilities can be elastically provisioned to scale with demand.
- Resource pooling
Computing resources are pooled to serve multiple consumers.
- Roadside unit (RSU)
Fog node deployed along the roadside that informs vehicles, e.g. about a traffic jam ahead.
S
- Service-level agreement (SLA)
Contract that commits a provider to a service level, usually availability as uptime over a period (the slides say a year; real provider SLAs use a month), with consequences such as service credits when it is missed.
- Simulation
Modelling a system's behavior rather than running its real software, used for training or testing, e.g. a flight simulator or VR training.
- Software as a Service (SaaS)
Provider manages the whole stack; user only manages application-specific configuration, typically via a browser.
- System call interface
The kernel's entry point through which user-space code, including every container, requests services; in OS-level virtualization it is the line between isolated user spaces and the shared kernel, which manages the hardware.
T
- Total cost of ownership (TCO)
Total cost of running infrastructure over a period, computed as CapEx plus OpEx.
- Type 1 hypervisor
Bare-metal hypervisor running directly on hardware; about 5% overhead and stronger isolation; used by cloud providers.
- Type 2 hypervisor
Hosted VMM running as a process on a host OS; about 10% overhead, easier to administer, less isolation.
U
- Utility computing
McCarthy's 1961 vision of computation sold like a public utility, paying only for capacity used.
V
- Vendor lock-in
Dependence on one cloud provider that makes switching costly, a reason not to drop traditional IT entirely.
- Virtual machine (VM)
An isolated environment on a physical machine that behaves like a complete computer with its own components.
- Virtual machine monitor (VMM)
Software layer (hypervisor) that intercepts and translates requests between guest OSes and physical hardware.
- Virtualization
Splitting a computer's resources so multiple independent OS instances can run on the same physical machine.