What makes a company like this lose so much money? I'm paying 15$ a month for years for a tiny tiny bit of hard drive space. Where does all the money go? What's the big money sink with something like this?
I'm not doubting it, just would like to understand.
The "Personal backup" pricing plan is $99 USD / year and includes "Unlimited data backup" and "one year version history" which implies they are storing much more than just your data at a given time.
$99 would give you 5-6TB of data for consumer HDD costs. And then factor in that they are likely doing some type of additional backup/RAID it would be even less data per dollar. And that's not including maintenance, rent, and employees.
Disclosures: I’ve been a customer for almost a decade. I’m also an early (IPO) shareholder. I’ve never been employed by, or been a vendor to Backblaze.
To add some context and background —
Their “unlimited” plan has a few major capacity-saving caveats. For example, there are built-in exclusions you can’t modify to prevent the cruft and temporary files of supported OSs (and several common apps like Outlook) from taking up space. Backblaze Computer Backup won’t backup certain kinds of data like Time Machine backups on Macs. They also have an automatic policy of deleting backups of external drives that haven’t been connected to the backup host for 30 days. They have a long list of caveats that to their credit they do disclose, although it’s a bit hard to find and understand if you aren’t technical and aren’t looking.
Their backup service is built on their object storage infrastructure to prevent duplicating their own stack, and it also compresses the amount of overhead and capacity accounting they have to worry about between their two product lines. They don’t keep backups of customer data, but they’d need to suffer multiple hard drive and host failures effecting the same file at the same time to lose any single file. Their architecture is proprietary, which I expect adds a lot of human/payroll overhead they wouldn’t have if they just deployed endless ZFS pools “off the shelf”. They build their own servers which offsets the costs of going COTS from a vendor like Dell.
You're paying for the hard drive, the server it goes into, the rack is in, the building it's in, and ask the power and maintenance for you to be able to access the data.
> You're paying for the hard drive, the server it goes into, the rack is in, the building it's in, and ask the power and maintenance for you to be able to access the data.
It's not even just the one drive. Data is stored across multiple drives, in multiple servers, hosted in multiple locations.
People pay the likes of Backblaze for it's reliability and them fool themselves into believing it's a dude with a HD stashed in a shelf somewhere.
Your data is scattered across multiple drives, but that doesn’t make it take up much more space. A terabyte stored in a RAID-5 array is 1.2 terabytes. Maybe the fixed price for unlimited storage part was a mistake, but the footprint overhead is relatively small. Also, keeping a year of history isn’t that much either unless you change every bit of the files multiple times. Using my BtrFS snapshots and Time Machine backups as a reference, normal data changes very little.
> Your data is scattered across multiple drives, but that doesn’t make it take up much more space.
That's besides the point. The point is that storing data in a service such as Backblaze requires more infrastructure than just that one HD in that one server in that one location.
I was skeptical of the wave of "unlimited" (or very large) online backup/storage offerings that came out ~15 years ago, but I didn't expect it to take this long to unravel.
Look at the offering at a glance right now they are offering $99/year unlimited, versioned backup of your home storage?
That seems like peanuts compared to buying a set of drives, setting up, and managing a truly redundant & versioned backup solution takes home. I spent like $1500 on a RAID NAS a decade ago and fell behind on keeping the thing operational.
>That seems like peanuts compared to buying a set of drives, setting up, and managing a truly redundant & versioned backup solution takes home. I spent like $1500 on a RAID NAS a decade ago and fell behind on keeping the thing operational.
Running storage system is cheaper than you think. Look at hetzner for an upper bound on how much storage costs at scale. They offer 10 TB of storage for $31/month. It mentions having "RAID-based storage" and "backups", so this presumably includes redundancy. If you can get 4 customers paying $99/year, you're already breaking even on the server. Most people probably don't have 2.5 TB worth of stuff. Backblaze only does backups, so it isn't like dropbox where you can upload whatever you want onto it. Most people don't even have computers with 1TB storage, so you'll probably be able to fit far more than 4 customers per server. Of course, this doesn't cover salaries for your engineer or marketing people, but at least it's not like moviepass where they're selling $50 worth of movie tickets for $10.
Actually, kind of the opposite. A few years ago Backblaze posted that the economics of their solution relied on the average customer having less than 1.2TB of data. They explicitly do not support Linux users because they tend to average a larger amount of data and are thus not economical to support, even ignoring the extra costs to maintain a separate Linux client.
If it’s so cheap to operate why have they continually lost money, possibly even committing accounting shenanigans to minimize the loss to a smaller number?
Your $1500 includes all the overheads that a storage provider amortizes over their entire customer base. At scale, the cost comes down to the cost of the hard drives, plugging them in, keeping them on, and bandwidth. The margins are really thin, so you need to reduce costs further by doing things like powering down drives that are not in use, or leaving failed drives in racks until they can be replaced in bulk. And importantly, if you offer something unlimited, make sure that it is actually limited. In Backblaze's case, the only 'unlimited' offering is only for their custom backup software that only backs up a single PC, and more recently allows limited backup of external drives. You can only keep multiple TB in it if you have multiple TB powered up in your PC and keep your external drives attached enough so their backups don't get wiped. It is impractical for archiving and abuse. And if it was a problem, they just need to tweak the terms of service to disallow abuse as is traditional.
Their price may well be too cheap. Maybe they make a loss due to unexpected costs (thin margins == risk). Or maybe they are priced too cheap, because they know increasing prices will lose even more money by losing more profitable business than unprofitable.
Capacity is cheap especially if you dedupe. IO is expensive.
Most people don’t have that much data, and capacity grows over time with new drives.
But… you need to be really disciplined to make money in a low margin business. I remember running the numbers on replicating something like backblaze in-house. It was just barely achievable, although I wasn’t crazy enough to shuck disks or build arrays.
I bought a Drobo with the same intent of it being my online redundant array type storage while the original copies were on external HDDs that were placed back in their original retail packaging. The Drobo itself was only connected to computer when necessary but left plugged in to mains when one day it was needed and multiple drives failed to mount--more than could be recovered.
I used to really be into all of the hardware stuff like that building 64 disk arrays and maintaining multiples for edit bays. Now, I'm so sick of it. I just want a plug-n-pray system that works out of the box and does not need maintaining. But of course I know that's not how gear works. So make it someone else's problem and throw money at that problem. $99/year sounds like an excellent bargain. apparently, too excellent
Put another way - if you are a Silicon Valley SWE, spending more than 15-60 minutes per year of your own time is more “expensive” than just giving backblaze $99/yr.
I'm not doubting it, just would like to understand.