Examining Video Surveillance Storage ClustersBy John Honovich, Published on Aug 06, 2008
In the next few years, we will see a significant transition away from storing video inside DVR/NVR servers. This 'on-board' storage will be replaced by clustered storage that communicates with servers over an IP network. The economics are becoming very attractive for such a move, especially in larger scale deployments.
This review examines the background, advantages and constraints of storage clusters as a replacement for traditional DVR/NVR storage.
Storage clusters are appliances that are separate from your DVR/NVR and communicate with them across your IP network. Storage clusters are modular and more storage can be added over time, starting from as low as a few TBs to more than 1000 TBs.
Here is a summary of the key advantages and constraints:
- Price differential between internal storage and storage clusters have dropped dramatically
- Storage Clusters can reduce storage needs by ~ 30% over internal storage
- Storage Clusters are actually cheaper for large camera counts and storage durations
- Storage Clusters are cheaper and better for megapixel cameras
- Storage Clusters offer RAID 'standard'
- Storage Clusters are not cost-effective for smaller camera counts
- Storage Clusters cannot centralize storage across most distributed facilities
Where historically the price differential per unit of storage was huge, today, the differential is small or, in many cases, not at all. This is critical in spurring broader adoption.
Almost all observers recognized that storage clusters were superior to on-board storage but the historical pricing for storage clusters was 300% to 600% more for the clusters. As such, it was incredibly difficult to justify the significant increase in expense and very few video surveillance systems used this solution.
In the past few years, the increasing maturity of these solutions and the utilization of standards based IP networks has shrunk the price differential. The price of the supporting infrastructure to build storage clusters has dropped, enabling the price of storage clusters to fall faster than the price of internal DVR/NVR storage.
Per TB, the price of storage clusters is very competitive with internal DVR/NVR storage. Storage clusters cost about $2,000 per TB. For most mainstream DVRs, the MSRP to add 1 TB of storage is $2500 to $3800. Storage clusters actually can be cheaper than internal storage. I was fairly shocked about this difference but I confirmed with multiple price lists from multiple dealers.
Note: The minimum size storage cluster available today is 4TB which is a big factor in small camera count deployment. This affects total cost for this scenario and will be examined in the constraints section.
Advantage 2: Reduce Storage Needs
When you deploy multiple DVRs, even in the same building and with the same configurations, you often obtain different storage durations. Let's say you target 90 days of storage, some will get 75, others 105, a few 55 and one or two 125 days. This is because storage utilization is a factor of amount of motion or traffic in a camera's view. (PTZs are famous for chewing up storage because of this).
Because storage duration falls in a range, you generally need to deploy more storage than the average system needs or reducing recording settings on systems that do not record as long as you need. In the former, the direct impact is higher installation costs. In the later, the direct impact is higher service costs and the indirect impact can be issues with evidence usability.
With storage clusters, DVRs/NVRs record to a central pool of storage. Let's say you have 10 DVRs and the average DVR consumes 800 GBs of storage to record 90 days. However, because some of the DVRs will need more storage to reach 90 days, you use 1000GB storage in each DVR instead, resulting in a 25% premium. In a storage cluster, because storage is pooled, units that need more storage and balanced off with units that need less. Therefore, you would simply use 8TB of storage and eliminate the 25% padding and premium.
Even in a modest scenario like this you can save a few thousand dollars simply by this pooling effect.
Advantage 3: Cheaper for Large Camera Counts / Extended Storage
Once you get into large camera counts or extended storage, it pushes the limits of what internal storage can provide. As such, the economics of storage clusters are preferable.
Using interal storage will require using the largest hard drives possible (so you can fit in chassis). Larger hard drives are much more expensive (per unit of storage) than smaller hard drives. With a storage cluster, you could use the most cost-effective hard drives sizes and reduce costs.
Using directed attached storage requires an external appliance. Such an appliance will cost a few thousand, even without the drives. At that point, and for roughly the same price, you might as well use a storage cluster that provides far superior scalability.
Advantage 4: Better for Megapixel Cameras
The storage demands of megapixel cameras are severe. You can easily use multiple TBs per megapixel camera because of the increased resolution and the need for MJPEG compression.
Most DVR/NVR appliances are not designed to handle this demand for storage coming from multiple megapixel cameras. You can use direct attached storage but again, for roughly the same price, the added benefits of a storage cluster generally make it the appropriate choice.
Advantage 5: Offers RAID 'standard'
While DVRs have offered RAID for years, the additional cost has been quite high. As such, most users elect not to use RAID. A check of leading appliances indicates going from 1TB non-RAID to RAID increases cost by $1500 to $3000 MSRP. Some systems like 3VR are now offering RAID standard but it's certainly more of an exception than the norm. All in all, though, these premiums are hard to justify.
By contrast, storage clusters offer RAID 'standard'. The incremental cost of using RAID is very low. Plus, these systems are all designed to provide many flavors of RAID right out of the box.
RAID offers two primary benefits for video surveillance systems: (1) save the box from dying and (2) save video from being lost.
Saving the box from dying is the more important of the two. Video surveillance systems that die require emergency service calls and increase the risk that a security incident occurs where no recording nor live monitoring is available.
Saving video from being lost is usually a nice to have for video surveillance. It is generally not critical because the probability of a incident being lost is usually low. This is based on my experience in government, military and Fortune 2000 companies. Video storage is generally more like insurance than it is like your children. If you lose your insurance for a few days, it's a risk but you are usually fine. Video surveillance storage is pretty much the same.
Certainly, a lot of reasons exist for moving to storage clusters. Nevertheless, the value for smaller, distributed sites (a major segment of video surveillance - banks, smaller retailers, QSRs) is not as strong.
Constraint #1: Not Effective for Smaller Camera Counts
Storage clusters have a startup cost that is notably higher than the internal storage DVRs/NVRs use. Just like any PC, internal storage is available by default and the only incremental cost is usually adding in the drives themselves. With a storage cluster, you usually have a separate appliance with electronics and computing infrastructure. As such, before you deploy any drives with a storage cluster, you first have to pay for this additional appliance.
With a storage cluster, the minimum available size seems to be 4TB at about $8,000 MSRP. If you use significantly less than 4TB, the cost for a storage cluster will be significantly higher than simply adding in drives to your DVR/NVR. Moreover, the benefits of pooling decrease because you are likely using less DVR/NVR units and do not need to worry about padding to achieve storage durations.
This is not trivial because large corporations have millions of facilities around the world that fit these characteristics. Until and if storage clusters can become more competitive at lower storage entry levels, the value for customers will be quite questionable.
Constraint #2: Not Capable of Centralization Across Distributed Facilities
Even though storage clusters have significant economic benefit for large amount of storage use, this is generally not feasible for aggregating storage from facilities across the country. To the credit of the vendors in this space, they have not made this claim. However, you do hear this from time to time by some in the industry.
Take a fast food restaurant with 1000 locations. In each location they probably have about 250 GB of storage (more if they are rolling out new systems). In total, that's 250 TB of storage (at least), which is quite significant. Hypothetically, they could save a few hundred thousand dollars if they could eliminate hard drives in the restaurants and just have one central storage cluster.
The problem is that you need significant amounts of bandwidth to accomplish this that simply is not available to most sites. For 8 or 16 cameras, you might need 5 - 20 Mb/s in upstream bandwidth. This is a huge amount for most stores where DSL/cable modem is the norm. In other words, the cost of bandwidth is far higher than the cost of storage. As such, this is not very realistic.
Driven by price competitiveness and a number of significant advantage, storage clusters will quickly become a major force in sites with modest to large numbers of cameras. Nevertheless, the sizable segment of the market with small camera counts per site will not see significant advantages in this. All integrators and security managers should carefully track and learn more about security clusters so they can take utilize their significant advantages.