Short answer. Useful company video surveillance in Algeria in 2027 holds five things. Cameras by role: each is chosen to identify, recognise or observe, with a pixel density calculated for its scene. A network: one cable per camera for data and power, on a network separate from the rest. Recording: retention days calculated before the disk is bought, then checked. Remote access through an encrypted tunnel, with no port open to the internet and one account per person. Operation: an alert when a camera goes silent, and a written way to extract a sequence. Two calculations decide everything: 10.8 GB per day for every Mbit/s recorded, and 250 pixels per metre to identify a face.
This guide is for the manager of a warehouse, shop, factory or head office who already has cameras and did not get the footage the day it was needed, for anyone preparing an installation who wants to know what to ask for, and for the founder based in France who wants to see their Algerian site from a phone without exposing it. The notions quoted come from public references: IP camera, network video recorder, the ONVIF forum, H.265 compression, power over the Ethernet cable, and IEC 62676-4 for pixel density. The numbers in the figures are a constructed example, a warehouse with sixteen cameras. This guide gives neither equipment prices nor legal rules: installing and using cameras is regulated, and authorisations, information given to the people filmed and the permitted retention period are to be checked with your legal adviser and the competent authorities before anything is fitted.
The day the footage is needed: how far back the recorder goes
The figure places the theft on a fifty-day axis and shows, for four settings of the same sixteen cameras, how far back the recording goes.
The calculation. A recorder writes in a loop: when the disk is full, it erases the oldest day. A stream of one megabit per second recorded without interruption takes 10.8 gigabytes per day (86,400 seconds, divided by eight, then by a thousand). Sixteen cameras at 4 Mbit/s therefore write 691 GB per day, and an 8 TB disk, 8,000 GB, a little less once formatted, is full in 11.6 days. The theft in the example, found at stock-take nineteen days later, is no longer on the disk. Nobody erased it: the calculation had simply never been done.
Three ways to go further back. H.265 compression, which public references credit with 25 to 50 % less data than H.264 at the same quality, brings the bitrate in the example down to 2 Mbit/s and retention up to 23.1 days: the theft is there, the thirty-day target is not. Recording on motion for the eight cameras in quiet areas, active 30 % of the time in the example, gives 35.6 days on the same disk; cameras on the passages everyone must take stay continuous, because a missed detection is a lost image. A second disk gives 46.3 days. The rule. The target is written before buying, in days; the real bitrate of each camera is read on the recorder, and it varies with the scene and with motion; and once a month, someone looks for the oldest image and notes its date. A stock discrepancy is often found late, hence the value of cycle counting: the more frequent the check, the fewer retention days are needed.
«Nobody erased the footage of the theft. The disk looped, because the calculation had never been done.»
Company video surveillance, in five layers
The accented layer is recording: it decides whether the image still exists the day you look for it.
Cameras. An IP camera is a digital camera that sends its images over the computer network; it compresses them itself, and the network video recorder receives them and writes them to disk, unlike older recorders, which process the camera signal themselves. Each camera is given a role, described in the next section. Network. A single Ethernet cable carries data and power: the 802.3af standard supplies up to 15.4 watts per port, of which 12.95 are guaranteed at the device, and 802.3at up to 30 watts, of which 25.5 at the device. What must be checked is the switch's total power budget, not just its number of ports: sixteen cameras at 7 watts at night, infrared on, in the example, need 112 watts. The cameras live on a separate network, as in the guide to the company network, so that a compromised camera sees neither accounting nor the workstations.
Recording. Disks made to write day and night, a recorder locked in the rack, because a thief who takes the recorder takes the evidence, an exact clock, without which the timestamp is worth nothing, and a copy of important sequences somewhere other than the recorder. Access. One account per person, factory passwords changed on day one, and no recorder port open to the internet. In 2016 the Mirai malware took control of IP cameras and routers by trying some sixty factory logins, then used them to attack other services. Remote viewing goes through an encrypted tunnel, as in the guide to linking sites. Operation. Monitoring reports the camera that no longer answers, the disk that is wearing out, the recording that has stopped; it becomes part of the routine described in the guide to managed IT. To avoid depending on a single brand, ask for equipment that conforms to ONVIF profiles, a forum founded in 2008 by Axis, Bosch and Sony that standardises exchanges between cameras and recorders.
What a camera really sees: the same face at 5, 10, 20 and 50 metres
The figure draws the same face with the number of pixels a 4-megapixel camera gives it at four distances.
Pixel density. A 4-megapixel camera produces an image 2,560 pixels wide. With a 90-degree horizontal angle, the scene is twice as wide as the distance: 10 metres wide at 5 metres, 100 metres at 50. The 2,560 pixels are spread over that width: 256 pixels per metre at 5 metres, 128 at 10, 64 at 20, 26 at 50. IEC 62676-4, in its 2014 edition, ties to these densities what an operator can do: 25 pixels per metre to detect a presence, 63 to observe, 125 to recognise someone seen before, 250 to identify a person. The 2025 edition replaces this scale with seven levels, from 20 to 1,500 pixels per metre, and keeps the two thresholds of 125 and 250.
What follows from it. To identify, the scene must be no wider than the image width divided by 250: 7.7 metres for a 2-megapixel camera (1,920 pixels), 10.2 metres for 4 megapixels, 15.4 metres for 8 megapixels (3,840 pixels). A wide-angle camera covering the whole yard observes; it identifies nobody at the gate. Hence the rule of one camera per role: a tight camera on each passage everyone must take, gate, warehouse door, till, loading dock, at face height rather than on the ceiling, and wide cameras to understand the scene. These thresholds hold in daylight, on a sharp image; night, the backlight of an open door, rain and compression degrade them, which is why the test is done on site, at night, before acceptance is signed. The badges and clock-ins described in the guide to time clocks and access control say who came in; the camera on the passage confirms it.
«The same camera identifies a face at five metres and only signals a presence at fifty.»
Six real situations, and what the system must do
Each row is a sentence heard after an incident, and the answer is a setting or a rule, not one more camera.
On the day of the theft, the recording was gone: retention is calculated, written and checked, as in the first figure. The image is there, but nobody is recognisable: the camera was fitted to see everything; one is needed to identify, on the passage everyone must take. At night, all you see is headlights: a camera's infrared lighting has a range, headlights and open doors dazzle; light is added, a camera made for backlight is chosen, and the result is looked at at ten in the evening.
I cannot see my cameras from my phone: remote viewing depends on the site's upload rate, not the phone's. Cameras generally offer a main stream, the one that is recorded, and a lighter sub-stream for display; four cameras watched in sub-stream at 0.5 Mbit/s need 2 Mbit/s of upload, four main streams at 4 Mbit/s need 16, which many links do not provide. The manufacturer's relay service, which avoids any configuration, routes the connection through its servers: it is a decision to take knowingly, with a password of its own and two-step verification. Nobody knows the password: administration access is kept in the company's vault and everyone has their own account, as in the guide to roles. A camera had been down for a month: nobody watches a screen where nothing happens; an alert has to say it.
Video surveillance in 2027, in weeks
Four weeks in the example, the first with nothing installed.
Week 1, survey and targets: the site plan, the role of each camera, the pixel density of each scene, the retention days wanted, who watches what, and a night visit; the outcome is a camera plan and a disk calculation, and only then is equipment chosen. Week 2, network and fitting: cables, powered switch, separate network, cameras fixed and set one by one, recorder locked in the rack.
Week 3, recording and access: bitrates set camera by camera, retention checked by calculation, named accounts, factory passwords changed, a tunnel for remote access, a test from a phone outside the site. Week 4, monitoring and file: alerts set, the extraction of a sequence tried end to end, two people trained, then the file handed to the company: camera plan, addresses, accounts, retention calculation. An existing installation is taken over in the same order: many cameras already fitted remain usable; it is the settings, the disk and the access that change.
The day the footage is needed, in six steps
The second step is accented: until the sequence is locked, the recorder keeps erasing it.
- What the manager sees: cameras online, the date of the oldest image, the month's alerts, and the picture of their site from their phone with no open port.
- What the company no longer does: find out after a theft that the disk had looped, hunt for the recorder's password, film a screen with a phone.
- What is measured: real retention days, cameras online, pixel density on the passages everyone must take, time taken to find a sequence.
- What Symloop does: the survey and the camera plan by role, the retention calculation, the separate powered network, the recording, remote access through a tunnel, monitoring with alerts, and the file handed to the company; for an existing installation, a stock-take and an upgrade.
Note the date, the time and the cameras involved, before touching anything. Lock the sequence in the recorder, or export it straight away: loop erasing does not stop because there was an incident. Export with the timestamp, in the original format, with the player supplied if the format requires it; a screen filmed with a phone loses the date and the quality.
Copy twice, one copy off site. Log who exported, when and for whom: that is what makes it possible to say later where the image came from. Hand over the sequence following the procedure the company has agreed with its legal adviser. These six steps are written on one page posted near the recorder and rehearsed once a year, like a backup restore, as in the guide to backups. Video analytics, line crossing after closing time, counting, plate reading, can signal an event instead of waiting for someone to look for it; it is tuned on site, produces false alerts until it is, and is only worth something if someone receives the alert.
- How many days of recording should be kept?
- As many as you need to discover an incident, within what the applicable rules allow, to be checked with your legal adviser. A theft at the till is seen the same evening; a stock discrepancy is seen at stock-take, sometimes weeks later. Write the target in days, then calculate the disk: 10.8 GB per day for every Mbit/s, multiplied by the number of cameras.
- What is the difference between a DVR and an NVR?
- In a DVR, the recorder itself encodes and processes the video from the cameras wired to it. An NVR, a network video recorder, receives over the network images already encoded by IP cameras. IP cameras generally offer more definition and a single cable for data and power; an older installation can often be kept in part and extended.
- How many cameras does a warehouse or a shop need?
- The number follows from the roles, not the other way round: one tight camera per passage everyone must take, entrance, exit, till, dock, and wide cameras to follow what happens in between. The one-week survey gives the count. Adding wide cameras does not make anyone identifiable.
- Can I watch my cameras from abroad?
- Yes, through an encrypted tunnel to the site's network, or through the manufacturer's relay service if you accept the principle. Smoothness depends on the site's upload rate: you watch the lighter sub-stream. What must not be done is to open a recorder port to the internet.
- Are 4K cameras better?
- They give more pixels, so a wider identifiable scene, 15.4 metres against 10.2 for 4 megapixels, but they fill the disk faster and often need more light at night. A 4-megapixel camera well placed on a narrow passage identifies better than a 4K camera covering a whole car park.
- Should recording be continuous or on motion detection?
- Continuous on the passages everyone must take and on risk areas, on motion in quiet areas. Detection saves disk space, but a missed detection is an image that does not exist; a few seconds are kept before and after each trigger.
- Can the cameras already installed be kept?
- Often yes. If they conform to ONVIF profiles, another recorder can read them; otherwise they stay tied to their brand. The stock-take says which to keep, which to move and which to replace; most often it is the settings, the disk and the remote access that need redoing.
- How much does a video surveillance installation cost?
- The cost depends on the number of cameras and their roles, cable lengths, retention days, and therefore disks, and monitoring. This guide gives no equipment prices. Ask for a quotation that writes, for each camera, its role and its pixel density, and for the recorder, the number of days the calculation guarantees.
- Who is allowed to watch the footage, and is an authorisation needed?
- Installing cameras and using the footage is regulated, and this guide gives no legal rules: authorisations, information given to the people filmed, retention period and handing over footage are to be checked with your legal adviser and the competent authorities. On the technical side, the system must be able to apply whatever is decided: one account per person, rights per camera, and a log of who watched and exported.
- What happens if the recorder is stolen or fails?
- Without another copy, the images are lost. Hence the recorder locked in the rack, a UPS for power cuts, an alert when it stops answering, and for the most important cameras a second copy, on a memory card in the camera when it allows it, or on storage off site.