I know memory is fairly cheap but e.g. there are millions of new videos on youtube everyday, each probably few hundred MBs to few GBs. It all has to take enormous amount of space. Not to mention backups.
Google just has a lot of storage space. They have dozens of data centers, each of which is an entire building dedicated to nothing but storing servers, and they’re constantly adding more servers to previous data centers and building new data centers to fit even more servers into once the ones they have are full.
IIRC, estimates tend to put Google’s current storage capacity somewhere around 10-15 exabytes. Each exabyte is a million terabytes. Each terabyte is a thousand gigabytes. That’s 10-15 billion gigabytes. And they can add storage faster than storage is used up, because they turn massive profits that they can use to pay employees to do nothing but add servers to their data centers.
Google is just a massive force in terms of storage. They probably have more storage than any other organization on the planet. And so, they can share a lot of it for free, because they’re still always turning a profit.
There are also techniques where data centers do offline storage by writing out to a high volume storage medium (I heard Blu-ray as an example, especially because it’s cheap) and storing it in racks. All automated of course. This let’s them store huge quantities of infrequently accessed data (most of it) in a more efficient way. Not everything has to be online and ready to go, as long as it’s capable of being made available on demand.
Edit: Clarifying that tape medium is typically used for the longest term storage with the caveat that read is slow, so that used for the stuff that is least likely to be accessed. For things that are accessed infrequently but still need to be available relatively frequently you can have a “caching layer” which is what I was referring to with the discs. It’s a tradeoff between speed of access and information density. Here’s an article from 2015 where Facebook/Meta is discussing their design: https://engineering.fb.com/2015/05/04/core-data/under-the-hood-facebook-s-cold-storage-system/
You can feel it on YouTube when you try to access an old video that no one has watched in a long time.
every time it lags, it’s because youtube has to send someone down to the basement to retrieve the correct blu-ray disc from a storage room
God bless those interns. Earning those college credits.
Three additional things that you have to keep in mind are that:
1 - Enterprise storage is much, much denser (as in, capacity per physical space occupied) than you would expect.
2 - These systems have capacity recovery features (primarily compression and deduplication) that save a lot more storage than you would expect.
3 - The elements in the infrastructure are periodically refreshed by migrating them to newer infrastructure (think of how you could migrate two old 500GB disks to a single modern 1TB disk to save the physical space of a disk).As an example about point 1, this is what IBM advertises in their public whitepaper for their Storage Scale systems (https://www.ibm.com/downloads/cas/R0Q1DV1X):
“IBM Storage Scale System is based on proven IBM Storage 2U24 hardware that can be expanded with additional storage enclosures up to 15.4PB of total capacity in a single node and 633YB in a single cluster. You can start with 48TB using half-populated flash nodes or create a fully-populated NVMe flash solution with 24, 2.5” drives in capacities of 3.84TB, 7.68TB, 15.36TB or 30TB. Using the largest capacity 30TB NVMe drives, up to 720TB total flash capacity, in a 2U form factor, along with associated low weight and low power consumption. Adding storage enclosures is easy as up to 8 enclosures (each 4u with 102 drives) can accommodate up to 816 drives of 10TB, 14TB or 18TB or 14.6PB of total raw HDD capacity.”
In short, you end up packing a stupid amount of storage in relatively moderate spaces. Combined with the other two points, it helps keep things somewhat under control. Kinda.
For twitter it’s not that complicated because tweets are quite short and text compresses very well. The pictures and videos people upload are of course another story, I’m not sure what Twitter uses as a backend for anything though.
Not only that but for each video on YouTube there are different versions for each resolution. So if you upload a 1080p video, it gets converted to 1080p AVC/VP9, 720p AVC/VP9, 480p… also for the audio.
If you run
youtube-dl -F <youtube url>you will see different formats.Does youtube actually store copies of each one? Or does it store 1 master copy and downsaple as required in real time. Probably stores it since storage is cheaper than cpu time
I believe they store and that’s why it processes lowest res first and works up
It’s transposed on the fly, this is a fairly simple lambda function in AWS so whatever the GCP equivalent is. You can’t up sample potato spec, the reason it looks like shit is due to bandwidth and the service determining a lower speed than is available.
Are you suggesting they don’t store different versions? This (speculative ik) suggests they do.
That response is almost 10 years old and completely outdated. I’ve designed and maintained a national media service and can confirm that on the fly transcoding is both cheaper and easier. It does make sense to store different formats of videos that are popular at the minute but in the medium to long term streams are transcoded.
Sure it’s old but the stats I posted in a lower comment show that at YouTube’s scale, it makes sense to store.
YouTube isn’t even profitable yet. Google pours billions into storage and compute, so does Amazon and Microsoft and all the others. They have so much space we probably can’t even comprehend it
TIL that Google owned YouTube. How did I not know this? I honestly thought that they were their own entity still after all these years. God I feel dumb.
acquisition happened in 2006
Absolutely huge data centers.
A full third of my towns real estate is currently covered with a sprawling Google data center. Just enormous.
NoVa?
I lived in Herndon VA for work for a while.
Was so nice gaming with 2 ping.
It’s the same story with AWS as well. They use vast amounts of storage and leverage different tiers of storage to get the service they want. It’s funny but they have insane amounts of SD cards ( cheapest storage available at the size) and use that for some storage and just replicate things everywhere for durability. Imagine how small 256GB SD cards are and they you have hardware to plug-in 200 of them practically stacked on top of each other. The storage doesn’t need to be the best, it just needs to be managed appropriately and actively to ensure that data is always replicated as devices fail. That’s just the cooler tier stuff. It gets complex as the data warms.
