Barry Norton, Fellow, Milestone Systems introduces the data storage playbook for modern video security systems.
There’s a paradox at the heart of modern video security, and most organizations are living with it every day without realizing it.
Cameras are generating more data than ever before. Resolution keeps climbing. Frame rates push higher. AI analytics layer on metadata by the gigabyte.
And yet the strategies most organizations use to store, manage and retrieve all of that data haven’t kept pace.
The result is systems that are more capable than ever, running on storage architectures designed for a simpler era. Drives fill up faster than expected.
Older data gets overwritten before anyone realizes it was needed. And when investigators or auditors come looking for a specific video, what should be a quick search turns into an hours-long project.
The good news is that modern video management platforms give security professionals real tools to close that gap.
Whether an organization is running everything on-premises, managing a mix of local and cloud resources in a hybrid environment, or looking to offload storage to a fully cloud-based model, the fundamentals of smart storage design apply across the board.
Getting the foundation right
Before jumping into cloud architectures and AI-driven storage policies, it’s worth spending a moment on some basics that still trip up even experienced system designers. They’re not glamorous, but they matter.
The first is the disk allocation unit size. Standard IT hard drives are typically formatted with a 4 KB allocation unit, which is fine for general office computing.
But high-bitrate video writes data very differently than a spreadsheet or a document, and that mismatch creates write inefficiencies that slow performance, increase drive wear and can lead to dropped frames at exactly the wrong moment.
I would recommend formatting simple recording drives at 64 KB; a straightforward configuration step that significantly improves write efficiency and extends drive life.
For deployments using a RAID controller, the controller’s stripe size should be set to match; otherwise, the benefit of the 64 KB formatting is negated.
The second is capacity planning. A common mistake is designing storage to use every available gigabyte, but that creates real problems for both system performance and data retention.
Spinning disks slow down as they fill up due to the physics of how data is written, and SSDs suffer accelerated wear and reduced endurance at high fill levels.
A practical design target is to plan for no more than 80% utilization. The system also has built-in panic limits that trigger forced archiving and deletion as a drive approaches capacity, but those are safety valves, not design targets.
Building the 80% buffer in from day one keeps performance stable and ensures video data isn’t lost before anyone realizes it was needed.
These are foundational details that apply whether the system is a single on-premises server at a small facility or a distributed enterprise deployment. Get these settings correct first, and everything built on top works better.
Thinking in tiers: the multi-stage archiving strategy
Once the basics are solid, the next step is thinking about storage not as a single bucket, but as a series of tiers, each matched to how the data is actually being used.
For on-premises deployments, this typically means keeping live or recently recorded video on high-performance SSDs for fast access during active investigations or daily monitoring.
That same video doesn’t need to stay at full resolution and frame rate forever, though. Through a process called storage archiving, older video can be automatically archived down to lower frame rates or keyframe-only formats and moved to less expensive, high-capacity SATA drives.
A common model keeps full-quality video for seven days, then grooms it down to five frames per second for longer retention. The difference in storage costs is substantial.
In hybrid environments, that second or third tier doesn’t have to live on local hardware at all. Open VMS systems can integrate directly with Amazon S3 and Amazon Glacier, allowing organizations to use cloud object storage as an archive tier.
Video that’s needed for compliance or long-term retention, but unlikely to be accessed frequently, is a natural fit for cloud archiving. It costs a fraction of what local deep storage would, and it scales without requiring new hardware.
There’s an important caveat for hybrid deployments, though. Cloud egress costs, meaning the fees incurred when pulling data back from the cloud, can add up quickly if video is being pushed offsite that teams regularly need to retrieve.
The practical rule of thumb is to use cloud tiers for compliance-driven, rarely accessed data, and keep operational video and anything likely to be reviewed in the normal course of daily work, closer to home on on-premises or edge storage.
That balance keeps costs predictable and performance where it needs to be.
For organizations running fully cloud-based or VSaaS deployments, the multi-tier principle still applies, though the mechanics shift. Cloud-native storage management handles much of the archiving logic automatically, and subscription-based VSaaS models bundle storage costs into recurring fees.
The trade-off is that the organization has less direct control over storage configuration, which makes it even more important to choose a platform with flexible retention policies built in.
The hybrid cloud as a storage architecture
According to an analysis published by ASIS International, organizations are gravitating toward hybrid architectures because modern cameras no longer just capture video; they generate metadata, AI-driven analytics and inter-device communications that each carry different storage and access requirements.
A hybrid approach lets organizations keep high-volume video on-premises where performance and control matter most, while leveraging cloud storage for redundancy, scalability and the smaller data types like analytics metadata that are well-suited to it.
It’s worth taking a moment to look at the hybrid model specifically, because it’s where a lot of organizations are landing right now, and it requires a different way of thinking about storage design.
Hybrid doesn’t mean partially committed to the cloud. It means intentionally using both on-premises infrastructure and cloud resources as complementary parts of a single, unified storage architecture.
For multi-site organizations, this distinction matters a lot. A regional retailer, a university with multiple campuses or an enterprise with branch offices across different cities faces a different storage challenge than a single facility.
One practical approach for these deployments is to use cloud-managed video surveillance as a service for remote or smaller sites where the cost of deploying and maintaining local recording hardware doesn’t make sense.
Those remote cameras stream to cloud-based storage and can be managed remotely, while the central facility continues running an on-premises VMS for full enterprise management.
The result is a single view across all locations without requiring a local server at every site.
The deployment options range from simple single-site cloud-only configurations for mid-sized to large installations, through single-site hybrid cloud and on-premises setups, all the way to multi-site geo-distributed architectures and full global deployments spanning international business locations.
Each model has a natural fit depending on the size and complexity of the organization, and the open-platform approach means the storage architecture can evolve as those needs change.
Storing intelligence, not just images
The final layer of the modern storage conversation is also the most forward-looking. As AI analytics become a standard part of video security systems, what gets stored is changing.
It’s no longer just video. It’s metadata: descriptions of what happened, when, where, and involving whom.
The shift matters for storage design because metadata is small, searchable and enormously valuable. A well-designed system stores metadata separately from the video itself, which allows investigators and security teams to search across enormous video archives in seconds without having to transcode or re-watch a single frame.
Looking for all data involving a specific vehicle type between specific hours over the past 90 days? In a metadata-indexed system, that search takes seconds rather than hours.
An open platform approach to digital evidence management takes this a step further, acting as a unified search and management layer across all storage tiers, whether that’s local on-premises drives, AWS archives or edge storage on cameras themselves.
It also enables policy-driven storage automation. An organization can set rules that say, for example, any video flagged by AI analytics as containing a firearm should be automatically moved to encrypted deep archive storage.
The system acts on that policy without human intervention, making sure the highest-priority evidence is always protected at the highest level of security.
This kind of intelligent, policy-driven storage is where the field is heading. Organizations that build for it now, with metadata indexing and AI-ready workflows in place, will be in a far better position than those that have to retrofit those capabilities later.
The open platform advantage
Running through all of these strategies is a single thread: flexibility. The ability to choose the right storage tier, the right vendor, the right hardware and the right cloud platform at every stage of the data lifecycle is only possible on an open platform.
Proprietary systems lock organizations into specific storage vendors, specific hardware configurations and specific upgrade paths.
When storage needs evolve, as they always do, proprietary systems turn what should be a straightforward technology decision into a costly, complicated migration.
An open platform approach keeps those decisions where they belong: with the organization. On-premises, hybrid, or fully cloud-based, the storage architecture should serve the security mission, not the other way around.
Getting the fundamentals right, thinking in tiers, using the hybrid model strategically and embracing metadata-driven intelligence are the steps that get you there.
This article was originally published in the May edition of Security Journal Americas. To read your FREE digital edition, click here.