How we check fit, prices and rates
Updated
This page describes how nasdrives.ca decides what fits a NAS, where its prices and electricity rates come from, and how the calculators compute their results. When something we published turns out to be wrong, it goes in the corrections log.
Which parts fit a NAS model
For each NAS model we record the facts that decide fit from the maker’s own specification pages, data sheets and compatibility lists, with the page and the date we read it: bay form factors, SATA or SAS, any per-drive size rating, M.2 slot count, length and use, memory generation, module form, slot count, per-module and total maximum, and ECC. Each product listing is read for the same properties. A part then passes four checks in order:
- Physical fit: form factor, interface, memory generation and form, slots and capacity limits.
- Model and software policy: for example, which Synology models accept third-party drives for new storage pools.
- The maker’s compatibility list, where the maker requires it for that kind of part (Synology does for M.2 NVMe).
- Intended use: a slot the maker describes as cache-only is never recommended for a storage pool.
Only parts that pass every check appear in a model’s main shortlist. A part that fits physically but is not on a list the maker requires is shown separately as “not on the list”. A part that depends on a detail the maker does not publish is shown separately as “fit not confirmed”. A part that does not fit is left out, and the page says how many listings were left out and why. These checks run as automated tests before every release.
How prices are checked
We track listings on Amazon.ca through Amazon’s product data service and refresh them every six hours; each product row shows when that listing was last checked, who sells it, its condition and what the price covers (a bare drive, one memory module, a NAS without drives). Pages show a separate “Prices checked” time, which is not the same as the “Updated” date: that one changes only when the page’s content changes.
When our Amazon.ca feed has no offer for a product, the page says “No current Amazon.ca offer” with the date checked. That describes our feed, not the Canadian market: other Canadian retailers may stock it, and we only track Amazon. Where Amazon.com has the product from Amazon or the maker at a normal price, we show that price separately in US dollars, without exchange, duties or shipping, and never add it to a Canadian total. Listings sold only through third-party resellers above list price show no price.
Value rankings compare like with like: new SATA drives by price per terabyte, with renewed, used and SAS drives listed separately. A build shows a total only when every required part has a verified price; otherwise it shows a known subtotal and names the parts without one. A price is called a drop only after at least 30 days of recorded prices for the same product, condition and seller; until then a deal page shows current offers. We earn a commission on some Amazon purchases; it does not decide what fits or how products rank.
Where electricity rates come from
Each province’s default rate in the running-cost calculators is read from the utility’s or regulator’s own published rates. Every default shows its source, the date it took effect and the date we read it, and has a review date matched to the utility’s schedule (Ontario every November 1, Quebec and British Columbia every April 1). A rate past its review date, or one we could not confirm with the utility, is flagged on the page instead of being hidden. The rate field is always editable.
Defaults are the electricity (energy) charge only. Delivery, regulatory charges, fixed daily or monthly charges and taxes are not included, and fixed charges are not spread over the NAS. A NAS runs around the clock, so for Ontario’s time-of-use plan we weight each price by the hours a constant load spends in it: (5,748 h × 9.8¢ + 1,506 h × 15.7¢ + 1,506 h × 20.3¢) ÷ 8,760 h = 12.62¢/kWh (2026: 251 weekdays, 114 weekend days and OEB holidays).
How the calculators work
- Usable capacity: RAID 1 keeps the smallest drive; RAID 5 keeps all drives but one at the size of the smallest; RAID 6 all but two; RAID 10 half; SHR keeps the total minus the largest drive. Results are shown in TB (the drive maker’s unit) and TiB (what the NAS reports): 1 TB = 0.9095 TiB.
- Running cost: average watts × 8,760 hours ÷ 1,000 = kWh a year, times the rate. The calculator adds a cooling allowance (5 percent by default) and UPS conversion loss (8 percent by default), both adjustable. Model wattages are estimates from the processor class and are labelled as such.
- Surveillance storage: bitrate in Mbps × 3,600 × recording hours ÷ 8 ÷ 1,024 = GB a day, times cameras and retention days, plus the overhead you set.
How failure rates are computed
Backblaze publishes raw daily snapshots of its fleet, one row per drive per day, and no summary dataset. We stream every daily file for a period and count, per drive model, the number of rows (drive days), the number of rows flagged as a failure, and the fleet size on the last day. The annualized failure rate is then computed by us:
AFR % = failures / drive days x 365 x 100We do not reproduce the published rate, so the arithmetic can be checked against the same public files. Models below 250 drives or 50,000 drive days are marked as having too small a sample to rank: at that size one extra failure moves the rate by whole percentage points. Every rate is shown with its drive days and a 95 percent range, computed with a normal approximation on the failure count. Brand totals are weighted by drive days rather than averaged across models, because averaging lets a 300-drive model count as much as a 60,000-drive one.
Component prices and cited figures
The component pages (hard drives, SSDs and memory) price from Amazon.com in US dollars. Our own price history begins the first day we recorded a reading for that item and is never backfilled, so a chart showing our data cannot imply we were tracking a price before we were. Third-party price series are stored separately, each row carrying the URL it came from, and are drawn as their own attributed layer: our readings and someone else’s are never combined into one line. A cited figure we re-read ourselves carries the date we read it; one taken as published says so.
Reporting an error
If a fit, price, rate or calculation looks wrong, please write to us through the contact page. Material corrections are listed with their date in the corrections log.