Compute, Storage & Network Requirements

Compute, Storage & Network Requirements
Lesson 3 · Provisioning

Compute, Storage & Network Requirements

Compute rarely means just CPU. Memory rides along with it, storage is specialized enough to deserve its own conversation, and network is what actually connects any of it to a user.

3.5

Compute Requirements

Compute resources are usually at least a combination of CPU and memory — and "compute" often gets used to cover storage and networking too. Because memory is so closely tied to CPU settings, CSPs manage the two together, the same way they allocate vCPUs. Choosing a memory-optimized instance, like Azure's Mv2-series, is the easiest path to configuring a VM for RAM-intensive work.

Resizing memory or vCPU isn't a live edit. Try adjusting the slider below with the VM powered on — then power it off first.

POWERED ON
16 GB RAM · 4 vCPU
Instance is running — resize is locked.

3.6

Storage Requirements

Storage is technically part of compute allocation, but it's specialized and extensive enough to earn its own discussion. Data lives in different structures — file, block, and object storage — each with trade-offs, and administrators need the right permissions before they can allocate or resize any of it.

Data also varies in how quickly it needs to be reachable — some has to be available with no latency, while other data is archived purely to satisfy retention requirements.

What is I/O?

An I/O request is a single read or write action performed by storage media. IOPS — I/O operations per second — is the primary measure of storage performance, and it only tells half the story without latency alongside it.

5,000
IOPS
4 ms
Latency
Balanced profile. Reasonable throughput with acceptable latency — a fit for general application storage. Both IOPS and latency are needed together to understand real storage performance.

VM instances need storage just like a physical machine, and the same fundamental choice applies: solid-state or traditional hard disk.

SSD

Speed
Cost
Capacity/$

HDD

Speed
Cost
Capacity/$

The right choice maximizes performance and capacity while minimizing cost, is subject to resource limits, and ultimately depends on what the VM is actually for. Disk redundancy is another configurable option on top of this choice.

The cloud term for a storage resource is a bucket. Buckets share a common set of characteristics regardless of provider.

Identified by a globally unique name
Stored at a specified geographic location, usually near the resources consuming it
Priced by storage type, capacity, accessibility speed, and outbound transfer, among other factors
Scalable as requirements change
Offers compliance features for retention, security, and accessibility
Access is controlled directly by cloud administrators

Anything that needs to store information in a cloud environment ends up in a bucket somewhere:

Static website files
App-generated data
Log files (load balancers, websites)
Inventory / analytics data
INFO — Storage configuration and optimization is covered in-depth in a later module.

3.7

Network Requirements

Cloud networking resembles traditional on-premises networking more than it differs from it — communication paths between services and consumers still have to be built and controlled. Virtual networks bring the same advantages virtual machines do: scalability, easy modification, and cost optimization.

On-premises network
physical infrastructure
Cloud virtual network
integrated hybrid segment
Cloud-only virtual network
exists entirely in the cloud

Cloud services allow a cloud-based virtual network to integrate with an on-premises physical network, and separately, allow one or more virtual networks to exist entirely in the cloud. To isolate network segments, administrators can build VLANs complete with their own switches and routers.

Every cloud VM connects through one or more virtual NICs — each with its own IP assignment and routing behavior. From there, administrators build two broad kinds of network environments.

INFO — Network configuration and optimization is covered in-depth in a later module.
§ 3.5–3.7 · Compute, Storage & Network Requirements