Throughput & I/O Considerations (Storage Performance)

Overview

Storage system performance is defined not just by capacity, but by how quickly data can be written and read.

In CCTV and other continuous recording systems, throughput and I/O behaviour are critical to ensuring:

• Stable recording
• No dropped frames
• Predictable retention

A system may have sufficient capacity, but still fail if performance requirements are not met.

Key Concepts

Throughput (Bandwidth)

Throughput refers to the amount of data that can be transferred per second.

Typically measured in:

• MB/s (megabytes per second)
• Gb/s (network throughput)

IOPS (Input/Output Operations Per Second)

IOPS measures how many read/write operations a storage system can perform per second.

• High IOPS = better for random workloads
• Low IOPS = acceptable for sequential workloads

Latency

Latency is the delay between a request and a response.

• Measured in milliseconds (ms)
• Lower latency = more responsive system

CCTV Workload Characteristics

CCTV systems have a very specific workload profile:

• Continuous writes
• Sequential data streams
• Minimal read activity during recording
• High sustained utilisation

This is fundamentally different from:

• Databases (random I/O)
• Office systems (bursty workloads)
• Web servers (mixed read/write)

Throughput Calculation (Simplified)

Total system write throughput is approximately:

Total Throughput = Number of Cameras × Bitrate per Camera

Example:

• 32 cameras
• 4 Mbps per camera

→ Total ≈ 128 Mbps (~16 MB/s)

Important Considerations

• Bitrate may fluctuate (VBR systems)
• Overhead must be included (filesystem, metadata, etc.)
• Peaks may exceed average values

Disk Throughput

Typical HDD performance:

• Sequential write: ~150–250 MB/s (single disk)
• Depends on drive type and workload

In RAID systems:

• Throughput may increase (striping)
• Parity calculations may introduce overhead

RAID Performance Impact

RAID 0

• High throughput
• No redundancy

RAID 1

• Write speed similar to a single disk
• Read speed may improve

RAID 5 / RAID 6

• Write penalty due to parity calculations
• Reduced effective write throughput
• More noticeable under heavy load

RAID 10

• High write and read performance
• No parity overhead
• Higher disk requirement

NAS / SAN Performance Considerations

When storage is network-based:

Network Bandwidth

• 1GbE ≈ 125 MB/s (theoretical max)
• 10GbE ≈ 1,250 MB/s

Real-world performance is lower due to overhead.

Bottlenecks

Performance may be limited by:

• Network interface
• Switch capacity
• Storage controller
• Disk performance

Latency Impact

Higher latency can result in:

• Write delays
• Buffering issues
• Potential frame loss in high-load systems

Sequential vs Random I/O

Sequential I/O (CCTV)

• Large, continuous data streams
• Efficient for HDDs
• Lower IOPS requirement

Random I/O

• Small, scattered reads/writes
• Requires high IOPS
• More suited to SSDs

Write Saturation

If system throughput exceeds disk capability:

• Recording instability may occur
• Frames may be dropped
• System buffers may overflow

This is often not immediately visible but impacts data integrity.

Practical CCTV Design Guidance

• Ensure total system bitrate is well within disk throughput limits
• Allow headroom (typically 20–30%)
• Avoid running systems at maximum sustained capacity
• Consider RAID overhead when calculating performance
• Validate network capacity when using NAS/SAN

Common Mistakes

1. Designing Only for Capacity

Ignoring throughput leads to unstable systems.

2. Ignoring RAID Write Penalties

Parity-based RAID reduces effective write performance.

3. Underestimating Network Limits

1Gb networks can quickly become a bottleneck.

4. Assuming Manufacturer Specs Reflect Real Performance

Real-world workloads differ significantly from lab conditions.

Related Topics

• RAID (Redundant Array of Independent Disks)
• Disk Types (HDD vs SSD vs Surveillance Drives)
• NAS vs DAS vs SAN
• Storage in CCTV Systems
• CCTV Storage Calculator

Summary

Throughput and I/O behaviour are critical to storage system performance.

In CCTV systems:

• Workloads are continuous and write-heavy
• Throughput matters more than IOPS
• Network and RAID overhead must be considered

A correctly sized system must account for both:

• Capacity (how much data)
• Performance (how fast data is written)