A
Auto-Scaling
A cloud feature that automatically adjusts the number of running instances based on current demand. When traffic increases, new instances are launched; when it decreases, instances are terminated. This ensures applications handle traffic spikes while minimizing costs during low-demand periods. All three major providers offer auto-scaling: AWS Auto Scaling Groups, Azure Virtual Machine Scale Sets, and GCP Managed Instance Groups.
Availability Zone (AZ)
An isolated data center location within a cloud region, with independent power, cooling, and networking. Distributing workloads across multiple AZs provides resilience against data center-level failures. AWS regions typically have 3-6 AZs, Azure calls them Availability Zones, and GCP refers to them as zones within a region.
B
Burstable Instance
A VM type that provides a baseline CPU performance level with the ability to burst above it using accumulated CPU credits. When the instance is idle, it earns credits; when it needs more CPU, it spends them. Examples: AWS T3/T3a, Azure B-series, GCP e2 shared-core. Ideal for workloads with variable CPU demands like web servers and development environments.
BYOL (Bring Your Own License)
A pricing option allowing you to use your existing software licenses (typically Windows Server, SQL Server, or Oracle) on cloud VMs instead of paying the cloud provider's licensing surcharge. Azure Hybrid Benefit is the most prominent example, offering up to 40% savings on Windows VMs for customers with Software Assurance.
C
Committed Use Discount (CUD)
Google Cloud's commitment-based pricing model where you commit to using a minimum level of resources (vCPUs and memory) for 1 or 3 years in exchange for discounted pricing. CUDs are resource-based rather than instance-based, meaning they apply flexibly across different machine types within the same family.
Compute Optimized
A category of cloud instances designed for CPU-intensive workloads. These instances provide a high ratio of vCPUs to memory, making them ideal for batch processing, scientific modeling, gaming servers, and media encoding. Examples: AWS C-series (C5, C6i, C7i), Azure F-series (Fv2), GCP C2/C3.
D
Data Egress
The transfer of data out of a cloud provider's network. Most cloud providers charge for outbound data transfer while inbound data is typically free. Egress costs can be significant for data-intensive applications and are often overlooked during cost estimation. Prices vary by provider, region, and destination.
E
EBS (Elastic Block Store)
AWS's block storage service for EC2 instances. EBS volumes persist independently from the instance lifecycle and provide different performance tiers: gp3/gp2 (general purpose SSD), io2 (provisioned IOPS SSD), st1 (throughput-optimized HDD), and sc1 (cold HDD). Equivalent services: Azure Managed Disks, GCP Persistent Disk.
G
General Purpose Instance
Cloud instances that provide a balanced mix of compute, memory, and networking resources. They are the default choice for most workloads including web servers, application servers, small-to-medium databases, and development environments. Examples: AWS M-series (M5, M6i, M7i), Azure D-series (Dv5), GCP N2/N2D.
GPU Instance
Virtual machines equipped with graphics processing units (GPUs) for accelerated computing. Used for machine learning training and inference, 3D rendering, video encoding, scientific simulation, and high-performance computing. Examples: AWS P5 (NVIDIA H100), Azure NC-series, GCP A2/A3.
H
Hybrid Cloud
An architecture that combines on-premises infrastructure with public cloud services, allowing data and applications to be shared between them. This approach provides flexibility, enables gradual cloud migration, and supports compliance requirements that mandate certain data remain on-premises.
I
IaaS (Infrastructure as a Service)
A cloud computing model where the provider offers fundamental computing resources — virtual machines, storage, and networking — on a pay-as-you-go basis. You manage the operating system, applications, and data; the provider manages the physical hardware, virtualization, and data center. EC2, Azure VMs, and Compute Engine are all IaaS services.
Instance Family
A grouping of related cloud instance types that share similar hardware characteristics and are optimized for the same type of workload. For example, AWS's M-family is general purpose, C-family is compute optimized, and R-family is memory optimized. Each family has multiple generations (M5, M6i, M7i) with newer generations offering better price-performance.
IOPS (Input/Output Operations Per Second)
A storage performance metric measuring the number of read and write operations a storage device can handle per second. Higher IOPS is critical for database workloads, transaction processing, and any application with intensive random I/O patterns. SSD storage typically provides 3,000-64,000 IOPS; HDD provides 250-1,750 IOPS.
M
Memory Optimized Instance
Cloud instances designed for workloads that require large amounts of RAM relative to CPU. They provide the highest memory-to-vCPU ratios, typically 8-32 GB per vCPU. Used for in-memory databases (Redis, SAP HANA), real-time analytics, and caching systems. Examples: AWS R-series, Azure E-series, GCP M-series.
Multi-Cloud
A strategy of using services from multiple cloud providers simultaneously. Benefits include avoiding vendor lock-in, leveraging each provider's strengths, improving resilience, and optimizing costs by selecting the best-priced provider for each workload. Challenges include increased complexity, higher operational overhead, and the need for consistent tooling across providers.
O
On-Demand Pricing
The default cloud pricing model where you pay for compute capacity by the hour or second with no upfront commitment or long-term contract. On-demand provides maximum flexibility but is the most expensive per-hour rate. Ideal for unpredictable workloads, short-term projects, and auto-scaling burst capacity.
P
Preemptible VM
Google Cloud's original discounted VM type that could be reclaimed by GCP at any time (with 30-second notice) and had a maximum 24-hour runtime. Preemptible VMs have been largely succeeded by Spot VMs, which offer similar discounts without the 24-hour runtime limit.
R
Region
A geographic location containing one or more data centers (availability zones) operated by a cloud provider. Cloud pricing, service availability, and compliance requirements vary by region. Choosing the right region affects latency, cost, and data sovereignty. Major providers operate 25-60+ regions globally.
Reserved Instance (RI)
A commitment-based pricing model (primarily AWS and Azure) where you reserve a specific instance type for 1 or 3 years in exchange for significant discounts (30-72% vs on-demand). AWS offers Standard and Convertible RIs; Azure allows instance flexibility within VM families. RIs are being partially superseded by more flexible Savings Plans.
Right-Sizing
The process of analyzing actual resource utilization of cloud instances and adjusting their size to match actual needs. Studies consistently show that 30-50% of cloud instances are over-provisioned. Right-sizing is typically the highest-impact, lowest-effort cost optimization available.
S
Savings Plan
An AWS pricing model offering discounts (up to 72%) in exchange for committing to a consistent amount of compute usage (measured in $/hour) for 1 or 3 years. More flexible than Reserved Instances because Compute Savings Plans apply across instance families, regions, and even Fargate and Lambda.
Spot Instance
A cloud instance that uses spare capacity at deeply discounted prices (60-90% off on-demand). The trade-off is that the provider can reclaim the instance when capacity is needed. AWS, Azure, and GCP all offer spot pricing. Ideal for fault-tolerant workloads like batch processing, CI/CD, and distributed computing.
Storage Optimized Instance
Cloud instances designed for workloads requiring high sequential read/write access to large datasets on local storage. They include large amounts of local NVMe SSD or HDD storage. Used for NoSQL databases, data warehouses, and log processing. Examples: AWS I3/D3, Azure L-series, GCP Z2/C3D with local SSD.
Sustained Use Discount
A Google Cloud pricing feature that automatically reduces the cost of Compute Engine instances based on how long they run in a billing month. Instances running more than 25% of the month begin receiving incremental discounts, reaching approximately 30% off on-demand rates for full-month usage. No commitment or action required.
V
vCPU (Virtual CPU)
A virtual CPU allocated to a cloud VM, representing one hardware thread (hyper-thread) of a physical CPU core. A physical 8-core processor with hyper-threading provides 16 vCPUs. Performance per vCPU varies by processor generation and cloud provider. When comparing across providers, benchmark your specific workload rather than relying on vCPU counts alone.
Put your knowledge to use
Now that you understand the terminology, use CloudMetrics to compare real cloud pricing across AWS, Azure, and GCP.
Compare cloud prices →