Skip to content

The Complete 2023 Guide to Computer Data Storage

Data is the lifeblood of technology – but making sense of modern storage is no easy feat. As our digital universe continues exponential growth, selecting the right storage technology is crucial for performance, cost and competitive advantage. This comprehensive guide delivers an authoritative overview of storage from volatile memory to emerging media, with insights to inform strategic decisions.

The Importance of Storage

Before surveying storage tech, we must underscore why it matters. Each year the world generates over 100 zettabytes of new data as digital transformation penetrates deeper across industries [1]. Analyzing and capitalizing on data requires storing it – from customer transaction histories to streaming media libraries. The storage medium impacts:

Speed – Latency between data storage and compute limits overall application performance. Slow storage cripples databases and data analysis.

Capacity – The sheer volume of enterprise and consumer data makes capacity a perpetual concern.

Durability – Storage safeguards data against hardware failure, disasters, human error or software bugs. Durability combines backup, redundancy, and business continuity.

Agility – Storage infrastructure should flexibly support new data sources, formats and use cases. Forward-looking agility prevents expensive rip and replace upgrades.

Balancing these factors while minimizing TCO makes storage decisions highly consequential.

Primary Storage

Primary storage consists of media directly accessible to compute resources for current processing needs. It emphasizes high speed access rather than large volumes.

RAM

RAM represents the first tier in the storage hierarchy – smallest capacity but fastest to access. It temporarily buffers active workloads in executed memory:

  • Volatile – Loses data when powered off
  • Expensive – Roughly $5-10 per GB
  • Fast – 60+ nanosecond access times
  • Limited – Up to 3TB per server

RAM use cases demand the fastest possible disk access for real-time response. High frequency trading bots analyze market data streams in RAM to enable microsecond trade execution. RAM also provides critical cache memory to minimize delays retrieving data from slower media.

Storage Class Memory

Storage class memory (SCM) occupies the space between RAM and SSDs, providing nonvolatile storage with access latency around 10-100x faster than flash [2]. Two competing technologies exist:

Intel Optane – An improved variant of 3D XPoint memory that Intel jointly developed and launched in 2015 before abandoning the partnership. Optane DIMMs slot alongside DRAM while Optane SSDs replace NVMe SSDs [3].

Standards Based SCM – The Storage Network Industry Association (SNIA) promotes an open standard interface designed to support multiple SCM technologies like phase change memory.

The verdict is still out whether Intel‘s proprietary Optane or SNIA‘s standards based approach will achieve broader adoption. While costly, SCM breaks past RAM capacity limits to support bigger fast memory pools. Use cases include buffering ever larger datasets or providing primary storage for latency sensitive applications.

Secondary Storage

While primary storage serves current workloads, secondary storage delivers larger persistent capacity. It comprises systems retaining data even after powering down – essential for accessing stored data sets, applications and files over time rather than just during a session.

Hard Disk Drives

First patented in 1955 by IBM, hard disk drives (HDDs) dominate budget conscious storage needs for 60+ years due to very low $./GB costs. HDDs write data on quickly spinning magnetic platters accessed by a moving actuator arm. However, the physical moving parts incur mechanical latency penalties. Additionally, the rapidly spinning disks consume significant power relative to solid state options.

Spec Desktop HDD Datacenter HDD
Capacity Up to 20TB Up to 22TB
Cost/GB $0.017 – $0.03 $0.025 – $0.12
Host Interface SATA, USB, SAS SAS, Fibre Channel
Latency 2-10 milliseconds
Warranty 2-5 years 5 years

Despite declining relevance for primary workloads as SSDs advance, HDDs will continue playing an enormous role where cost trumps speed – high capacity cloud object stores, media archives, and of course consumer laptops and desktops.

Solid State Drives

Replacing spinning disk platters with integrated circuits, solid state drives (SSDs) deliver vastly faster access. NAND flash memory cells store charge as binary data without moving parts. Costs steadily improve but remain higher than HDDs – making SSD ideal for performance sensitive workloads.

Spec SATA SSD NVMe SSD
Interface SATA PCIe
Cost/GB $0.08 – $0.15 $0.10 – $0.30
Capacity Up to 4TB Up to 128TB
Latency .05 ms .02 ms
Warranty 5 years 5 years

NVMe – Newer Non Volatile Memory Express attaches via PCIe rather than SATA for vastly improved throughput. NVMe leverages multiple I/O queues and streams enabling up to 4x faster maximum performance than SATA.

Flash Types – Single layer cell (SLC), multi layer cell (MLC), triple layer cell (TLC) and quad level cell (QLC) flash provide various capacity/endurance/cost tradeoffs.

Parameter SLC MLC TLC QLC
Endurance (P/E Cycles) 100k 10k 3k <1k
Relative Cost 3-4x 2x Baseline 2/3x

With HDD closing the capacity gap but not the speed gap, the SSD juggernaut continues full steam. SSD makes possible real-time insight from massive datasets via faster queries. Its true value emerges when using SSD to reshape application and analysis possibilities – not just accelerate legacy systems.

Emerging Storage Class Memory

Storage class memory like Intel‘s Optane offers promising bridging between storage and memory. Covered above in the primary storage section, SCM provides memory-like speeds at higher capacities than RAM alone permits. This lets more data reside closer to compute for faster processing while the most active working set stays in DRAM.

Initial SCM usage focuses on buffering and caching larger indexes, databases, metadata stores and similar write-intensive workloads prone to high fragmentation [4]. The future may bring SCM displacing some primary storage use cases altogether.

Tape Drives

Given Contemporary data volumes, tape storage plays a crucial role offering unmatched capacities at lowest $.GB despite misconceptions of disappearing obsolete technology. Modern LTO-9 tapes store 18TB uncompressed at 700MB/s transfer rates. Meanwhile 30TB compressed enterprise cartridge capacities arrive soon [5].

Unlike HDDs and SSDs, tape data resides sequentially along the recording medium rather than random access block storage. This requires linearly streaming data – unacceptable latency for primary workloads but ideal for archival and backup use cases.

Additional tape benefits include:

  • 3 decade shelf life – Far outpacing any disk media
  • Portability – Easily eject cartridges for offsite storage
  • Energy efficiency – Nearline tapes consume just 7 watts idle [6]

While past capacity more than meets requirements for rarely accessed data, future tape generations push density further with goals for 148TB native cartridges. No other technology currently provides such scalable and inexpensive archival ability.

Optical Discs

Compact Discs revolutionized music distribution in the 1980s. But optical disc relevance continues today offering inexpensive removable random access media. Use cases take advantage of key strengths:

  • Long term reliability – Estimated data retention up to 100 years
  • Virtually unlimited writes for rewritable models
  • Dirt cheap at $.02/GB for write once discs [7]
Spec CD DVD Blu-ray
Debut 1982 1995 2006
Capacity 700MB 4.7GB 25GB
Transfer Rate 150 Kbps 11 Mbps 432 Mbps

While optical lacks performant capacities for primary storage, excellent longevity at low cost makes optical storage highly attractive for archives. M-Disc and related write-once models specifically enhance data durability for long-term cold storage needs.

Magnetic Tape Cartridges

Magnetic tape plays an equally vital role offering unmatched scale economics due to very low media costs and operational expenses. Modern LTO-9 tapes push native cartridge capacities to 18 TB with transfer rates of 700 MB/s [5]. Future LTO-10 roadmaps call for 24-30 TB capacities leveraging barium ferrite magnetic particles [8].

Unlike HDDs and SSDs, tape data resides sequentially along the recording medium rather than random access block storage. This requires linearly streaming data – unacceptable latency for primary workloads but ideal for archival and backup use cases.

Additional tape benefits include:

  • 3 decade shelf life – Far outpacing any disk media durability
  • Portability – Easily eject cartridges for offsite storage
  • Energy efficiency – Nearline tapes consume just 7 watts idle [6]

For rarely accessed data requiring retention, tape currently offers the most scalable and inexpensive archival storage solutions. Future tape technology roadmaps push density and capacities further.

Cloud Storage

Rather than rely on local infrastructure, cloud storage saves data to remote servers provided by vendors like AWS, Google and Microsoft. It offers several advantages:

Limitless capacity – Add storage as needed without procuring hardware
Availability – Redundantly stored across data centers
Collaboration – Globally accessible by distributed teams
Automated backup – Protects against site failures

Of course latency, egress fees, and reliance on internet connectivity pose challenges. Data gravity can also emerge as accumulating volumes make transferring large datasets cost prohibitive.

Storage Networking

Storage networking and shared storage enable centralized data and simplified management:

DAS – Direct attached storage links dedicated storage directly to a single server
NAS – Network attached storage connects storage to a LAN for file sharing
SAN – Storage area networks provide block level storage access over a dedicated high speed network

Storage networks unlocked critical application cluster capabilities like live VM migration across storage domains.

The Data Storage Hierarchy

Balancing cost, performance and capacity across workloads reveals a classic inverted pyramid data storage hierarchy with key principles:

  1. Store actively processed data requiring lightning fast access on the smallest, fastest media – RAM and cache
  2. Retain frequently accessed data requiring high IOPS on low latency SSDs
  3. Migrate rarely referenced data, backups and archives to the highest density, lowest cost HDD or tape

Adhering to this stratified approach minimizes wasted resources mismatched to need.

Data Storage Hierarchy Pyramid

Selecting The Right Storage

With so many technologies and tradeoffs to evaluate, here is practical guidance for common use case requirements:

Personal Computing

For home and office needs favoring affordability, stick with SATA SSD system drives combined with larger HDD storage:

  • Laptop – 500GB SSD + external HDD backup
  • Desktop – 1TB NVMe SSD + 2 TB HDD

Mission Critical Enterprise Applications

Invest in all flash storage with data redundancy mechanisms and backup strategies:

  • Virtual server infrastructure – Leverage shared NVMe flash arrays
  • SQL databases – Opt for high IOPS low latency SAN
  • NoSQL databases – Direct attached SSD clustered storage

High Performance Compute

HPC workloads demand bleeding edge storage performance to feed data hungary applications, AI/ML training, and simulations. Consider:

  • Parallel file storage – Scale IOPS and bandwidth with high RPM HDDs or SSD clusters
  • Burst buffering – Tiered storage using NVMe to absorb I/O spikes without bottlnecks

Media Libraries & Archives

When prioritizing cost efficient capacity for video repositories, aggregates analytics datasets, medical imagery and records, prefer high density cool storage options:

  • Cloud object storage – Virtually unlimited capacity
  • MAID – Massive array of idle disks spin down when inactive
  • Deep flash – QLC flash prioritizes density over performance
  • Shingled magnetic recording – Overlapping tracks boost HDD density

Backup Targets

Emphasize maximizing capacities for infrequently accessed backups while maintaining data integrity:

  • Magnetic tape – Unmatched scalability and 30 year shelf life
  • Optical disc – Inexpensive media with 100 year lifespan
  • Cloud archives – Geographic distribution and encryption

Adjust strategies based on frequency of access, capacity demand, latency needs and budget considerations.

The Future of Data Storage

If the previous decade brought SSDs mainstream, what does the next era hold? Storage innovation continues advancing in labs and R&D clean rooms around the world.

Holographic Storage – Uses laser beams to encode data throughout photosensitive crystal media for over 50x Blu-Ray density.

DNA Storage– Leverages DNA molecule quadrillions of base pair combinations providing extraordinary density – 1 gram stores 1 exabyte!

Quantum Storage – Exploits quantum effects like superposition for exponentially greater capacity and parallelism compared to classical bits.

Atomic Storage – IBM demonstrated encoding data on holmium atoms holding potential for high density immutable records.

Widespread viability remains years away but demonstrates continued push for groundbreaking storage technologies.

Conclusion

We‘ve covered a vast range of primary and secondary storage alternatives spanning volatile system memory to cheaply scalable tape archives – each optimized around key vectors like speed, density durability and cost.

Carefully evaluate your workload patterns and metrics for data access frequency, capacity demand, growth projections, latency needs and budget constraints.

Matching the appropriate storage media optimally aligns infrastructure investments to strategic goals rather than reactively adding more of what you already have. Consult the data storage hierarchy pyramid to right size tiers – allocating higher cost media only where justified by real performance ROI rather than overprovisioning.

With digital transformation permeating deeper globally every year, selecting the best storage technology for current and future objectives delivers strategic advantage. Master modern storage options – don‘t let infrastructure limitations throttle innovation.


Citations

[1]. Forbes Technology Council Post, 2022

[2]. StorageNewsletter.com, 2021

[3]. IBM Systems Magazine, 2021

[4]. Supercomputing Frontiers and Innovations, 2017

[5]. LTO Program Media Roadmap, 2022

[6]. Clipper Notes Report, 2021

[7]. Optical Storage Technology Association, 2022

[8]. LTO Program Technology Provider Companies, 2021