Hardware and Accessories Buyer's GuideBy: John Honovich, Published on Oct 03, 2011
What type of hardware your surveillance server will require differs between manufacturers. The manufacturer's documentation or tech support should always be consulted before purchasing hardware. Each system handles video in a different way, and this variance may be slight or extreme. Processor needs could be as low as a single dual core processor or as high as dual quad cores, depending on a variety of factors, such as framerate, resolution, server-side motion detection, and more.
One deceptive practice to watch out for is hardware specified in terms of number of cameras. This can be wildly inaccurate, because number of cameras means nothing without knowing how much bandwidth each of these cameras will use. Running 32 VGA camera using 512Kb apiece is vastly different from running 32 HD cameras, each with a 3Mb stream. This is another reason why manufacturers should be consulted before specifying hardware.
Obviously, since we are discussing network video, a network interface of some kind is necessary. The speed required will depend on how many cameras will be managed by the system. For small systems, a 10/100 fast Ethernet interface is generally all that is required, and is typically built in to the machine's motherboard. For larger systems, a gigabit Ethernet interface will be needed. 10/100/1000 Ethernet is installed by default in most, if not all, servers.
Virtualization provides a handful of benefits, such as better server utilization, running multiple server instances on a single physical machine, and easier imaging for back-up. However, despite these benefits, virtualization does provide some challenges. For the most part, we recommend that servers used for surveillance not be virtualized.
The first drawback is that each VM adds a certain amount of processing overhead to the machine, which reduces the number of cameras it can handle. It is not uncommon fro this overhead to be in the range of 20-25%, so a server which could previously handle 48 cameras may be able to only handle 36, for example, if running in a VM.
Second, VMs add an additional level of complexity to an already complex system. Many users and technicians can find their way around Windows, even server versions, just fine. Adding a VM to this, however, adds another point of failure which may require troubleshooting if things go wrong. Not every tech, and we would guess not many, will have in-depth knowledge of how to fix these issues.
There are two ways storage for the surveillance system may be handled: on-board and external. On-board refers to just that: hard disk drives are installed in the server machine and used for storing video. This is the simplest and most common way to store video in the vast majority of surveillance systems, since most systems are relatively low camera count (4-24), and require short retention times (5-15 days). When camera counts are higher or video must be stored for an extended period of time, external storage may be needed. External storage can generally be broken down into three categories:
Direct-attached (DAS): Storage of this type is literally directly attached to the server. This connection is most commonly made via USB or eSATA.
Network-attached (NAS): Network attached storage consists of a unit attached to the network, typically containing a web interface, through which users create folders where files are stored. Small NAS units are relatively inexpensive, and may be used for other purposes, such as file storage, as well.
Storage Area Network (SAN): A SAN is a device or devices attached to a closed network which is used only for storage traffice. While it is network-attached, storage connected via a SAN differs from NAS in one way: where the NAS device contains an operating system and web interface, the SAN does not. It appears simply as a disk or disks to the server, where a NAS appears as a server. SANs are the most commonly used device for storing large amounts of video.
Brands: Dell, HP, IBM, Lenovo, etc.
Good Stuff: Readily available hardware with no proprietary additions, results in the most flexibility. May be VM'd if the application truly requires it. Well-established companies, less risk of failure compared to young appliance providers. "No one ever got fired for buying IBM" (or Dell or HP) is true in this case.
Bad Stuff: Requires full installation of software from scratch, as well as Windows updates, which may be pre-installed on some of the appliance manufacturers.
Bottom Line: Unless there is a very specific reason to use an application-specific server, using COTS hardware is always the safest bet.
Brands: Supermicro, ASUS, etc.
Good Stuff: Same as above, except typically lower cost than major name options. Generally stocked and available for shipping from online distributors.
Bad Stuff: Some of the benefits of company health of the big brands are lost when using lesser-known brands, as some "white box" suppliers may appear and disappear quickly. There is a perception that these brands are lower quality, whether deserved or not.
Bottom Line: While these lesser brands do offer some price advantage, quality may be compromised. We recommend careful thought before purchasing these brands instead of major manufacturers.
Third Party Surveillance Appliances
Brands: Intransa, Pivot3
Good Stuff: Speeds installation by pre-installation of the latest Windows version with prerequisites for VMS software, Windows Updates, etc. May also have VMS pre-installed. Multiple VMs may be pre-installed, so a single server appliance may run both video surveillance and access control, for example
Bad Stuff: Adding any VM uses some processor, meaning a server with a VM will not handle as many cameras as a server without. Often expensive compared to COTS servers. Company health and future may be a concern, as unlike commodity manufacturers, appliance providers are young, may have questionable futures.
Bottom Line: For systems using both surveillance and another application requiring a server (such as access control, license plate recognition, etc.), using an appliance with multiple VMs typically saves cost versus using multiple servers. For surveillance-only deployments, appliances have questionable technical value.
Power over Ethernet should be used in all cases except for niche scenarios. PoE is far simpler to use and deploy than using low voltage power because it allows everything to be run over one cable and eliminates the need for terminating power. On average, it reduces cost by $10-$20 per camera compared to low voltage power. Finally, PoE is nearly ubiqutiously available on IP cameras. Only the cheapest cube cameras do not support PoE.
PoE Enabled Switch
The simplest way to use PoE is to select a PoE-enabled switch. Most switches do NOT support PoE so you will need to find one that does. This is not hard but just do not take it for granted. The most important issue when selecting a PoE switch is to get one that supports maximum power (15 watts) to all ports. Many PoE switches only support half power assuming that only a few ports will be used with PoE. This may make sense for general deployments but can be a real problem with IP video surveillance. The difference between half power and max power is generally less than $100 per 8 port switch so it's not a big increase in cost. However, cutting corners on a PoE switch that does not support full power can cause painful and repeated service calls that look like the cameras are bad when really power is insufficient. Bottom line, get a PoE enabled switch that supports max power (15 watts) for each port (e.g., an 8 port switch should support 120 watts PoE total).
More advanced users and more advanced applications can consider PoE injectors or PoE midspans to power PoE cameras independently of switches. As this is somewhat more complicated and costly, do this only if you know what you are doing and have a specific reason why.
When selecting switches, three scenarios typically entail three different scenarios:
- Small stand-alone system: If you are deploying 50 or fewer cameras per site and this is going to be on a new network, just buy SMB switches with gigabit uplink ports and you should be fine (remembering max power PoE for the switches). If you are using H.264 cameras, you will almost never come close to a bandwidth bottleneck.
- Running IP cameras on existing networks: Determine what the existing standard is in your network and use that.
- Large and/or integrated system: If you are deploying 100s of cameras per site or if the cameras will be running over an existing corporate/organization network, this requires real planning and expertise. This is beyond the scope of an introductory buyer's guide.
For beginner and intermiediate users, go with the lens recommended by the manufacturer. You will not save much by picking your own lens and you might actually make things worse. Generally, camera manufacturers test lens and ship with validated lenses.
If you need to capture an image that is far away from the camera, you will need a lens with a longer focus, typically this is a 5-50 mm lens. This almost never comes stock with cameras. It's best to ask the camera manufacturer for a recommendation.