15.4 Local, Network, and Cloud Computing and Storage; How Binary Data Are Stored
Key Takeaways
- Local computing and storage run on the user's own device: fast and usable offline, but limited to that device's capacity and vulnerable to its loss.
- Network (on-premises) computing uses an organization's own servers over its local network, centralizing files, backups, and control, but it requires the organization to buy and maintain the equipment.
- Cloud computing delivers servers, storage, and software over the Internet on demand; it scales easily and supports collaboration but depends on connectivity and on the provider's security and policies.
- Cloud service models: IaaS provides virtual machines and storage, PaaS provides a managed platform for deploying code, and SaaS provides complete applications used through a browser or app.
- Hard drives and tape store bits magnetically, SSDs and USB drives store them as trapped charge in flash memory cells, and CDs, DVDs, and Blu-ray discs store them as reflective marks read by a laser.
What this competency asks
ETS asks you to be familiar with trade-offs between local, network, and cloud computing and storage:
- Identify advantages and disadvantages in terms of performance, cost, security, reliability, and collaboration.
- Identify means of storing binary data.
Section 3.3 introduced the privacy side of local versus cloud storage. This section compares the options as computing systems.
Three places to compute and store
| Model | Where it runs | Example |
|---|---|---|
| Local | On the user's own device | Editing a video with software on a laptop; saving files to its SSD |
| Network (on-premises) | On servers the organization owns, reached over its local network | A school file server; a district's student-information server in its own data room |
| Cloud | On a provider's data centers, reached over the Internet | Google Workspace or Microsoft 365; cloud storage; renting virtual servers |
Comparing them on ETS's five criteria
| Criterion | Local | Network (on-premises) | Cloud |
|---|---|---|---|
| Performance | Fastest for large local files; no network delay; limited by one device's hardware | Fast on the local network; limited by the organization's servers | Huge capacity on demand; each access limited by Internet bandwidth and latency |
| Cost | Up-front device cost; no subscription | High up-front cost for servers, plus ongoing staff, power, and maintenance | Pay as you go or by subscription; little up-front cost; can grow expensive at scale |
| Security | Full control; risk of theft or loss and weak personal practices | Organization controls policies; needs skilled IT staff | Providers invest heavily in security; data sit with a third party; shared responsibility; account compromise is a major risk |
| Reliability | A single device fails; backups are up to the user | Depends on the organization's redundancy and backups | Providers replicate data across data centers; outages still happen; requires Internet access |
| Collaboration | Hard; files must be copied and sent | Good within the organization | Excellent: real-time co-editing from anywhere |
Hybrid designs are common: work locally, synchronize to the cloud, and keep sensitive systems on-premises.
Worked decisions
| Scenario | Best fit | Deciding criterion |
|---|---|---|
| A photographer edits 200 GB of raw images while traveling with unreliable Wi-Fi | Local, with a cloud backup whenever a connection is available | Performance and offline access |
| District policy requires student records to stay on servers the district controls, and it has an IT staff | Network (on-premises) | Security and control |
| A new app might have 100 users or 100,000 users next month | Cloud (IaaS or PaaS) | Cost and scalability: pay only for what is used |
| Students in three schools co-write one lab report | Cloud (SaaS) | Collaboration |
Cost: up front versus ongoing
Local and on-premises systems are mostly capital expenses: hardware is bought up front. The cloud is mostly an operating expense: the customer pays each month for what it uses. With illustrative numbers (not real prices), suppose an on-premises server costs $9,000 up front plus $2,400 a year for power, space, and maintenance, and a comparable cloud server costs $500 a month, or $6,000 a year.
| After | On-premises total | Cloud total |
|---|---|---|
| 1 year | $11,400 | $6,000 |
| 2 years | $13,800 | $12,000 |
| 3 years | $16,200 | $18,000 |
The costs break even when 9,000 + 2,400t = 6,000t, so t = 2.5 years. The cloud wins for short-lived, small, or unpredictable workloads because it can scale down. Owning can cost less for a steady, heavy load over many years, once staff time is added to the comparison.
Reliability in numbers
Providers state availability as the percentage of time a service is up. A year has 8,760 hours, so 99% availability allows 87.6 hours of downtime a year, 99.9% allows 8.76 hours, and 99.99% allows about 53 minutes. Each additional nine cuts the allowed downtime by a factor of 10.
A cloud service also needs the school's own Internet link. If the link is up 99.5% of the time and the provider 99.9%, both must work at once: 0.995 × 0.999 ≈ 0.994, or about 99.4%. The weakest link in the chain limits the whole system.
Cloud computing in more detail
The U.S. National Institute of Standards and Technology (NIST) describes cloud computing as on-demand network access to a shared pool of configurable computing resources (networks, servers, storage, applications, and services) that can be provisioned rapidly with minimal management effort.
Service models
| Model | Provider manages | Customer manages | Examples |
|---|---|---|---|
| IaaS (infrastructure as a service) | Physical hardware, networking, virtualization | Operating system, runtime, applications, data | Virtual machines such as AWS EC2 and Azure VMs |
| PaaS (platform as a service) | Everything through the runtime and operating system | The application code and data | Google App Engine, Heroku |
| SaaS (software as a service) | The entire application | Only its use, and the user's own data and settings | Google Workspace, Microsoft 365, learning management systems |
Deployment models
- Public cloud: shared infrastructure from a provider, available to many customers.
- Private cloud: cloud-style infrastructure dedicated to one organization.
- Hybrid cloud: a mix, for example public cloud for the website and private systems for sensitive records.
Scalability and elasticity: cloud resources can be added or removed quickly as demand changes. A school's registration site can handle a surge on the first day of sign-ups, then scale back down.
Shared responsibility for security
Moving to the cloud moves some security work to the provider, but never all of it.
| Model | Provider secures | Customer still secures |
|---|---|---|
| IaaS | Data centers, hardware, network, virtualization | Operating system updates, applications, firewall settings, accounts, data |
| PaaS | Everything through the runtime | Application code, accounts, access to data |
| SaaS | The whole application | User accounts, passwords and multifactor authentication, sharing settings, and what data is uploaded |
A secure data center does not help if a teacher reuses a leaked password or shares a gradebook with "anyone with the link."
Means of storing binary data
Every storage medium needs a physical property with two distinguishable states.
| Medium | How a bit is stored | Characteristics | Typical use |
|---|---|---|---|
| Hard disk drive (HDD) | Direction of magnetization on spinning platters | Large, cheap per GB; slower; moving parts | Bulk storage, backups |
| Magnetic tape | Magnetization along a long tape | Very cheap per GB; sequential access only | Long-term archives |
| Solid-state drive (SSD), USB flash, SD card | Electric charge trapped in flash memory cells | Fast, silent, shock-resistant; cells wear out after many writes | Laptops, phones, portable storage |
| Optical disc (CD, DVD, Blu-ray) | Pits and lands on a reflective surface, read by a laser | Portable and inexpensive; low capacity; slow | Distributing media; some archives |
| RAM (DRAM) | Charge in tiny capacitors that must be refreshed constantly | Very fast; volatile | Working memory for running programs |
| Cache (SRAM) | State of transistor circuits (flip-flops) | Fastest memory outside registers; volatile | CPU caches |
Everything that is nonvolatile keeps data without power: magnetic, flash, and optical media. RAM and cache are volatile (Section 15.1).
A company's developers upload their application code to a cloud service that automatically provides and manages the servers, operating system, and runtime environment. The developers manage only the code and its data. Which cloud service model is this?
A school moves its shared documents from a server in its building to a cloud office suite. Which statement is an accurate trade-off?
Which storage medium records bits as pits and lands on a reflective surface that a laser reads?
In which situation is local computing and storage the best choice?