Data centers contain 90% crap data

&gt; One organization I knew of had 1,500 terabytes of data, with less than 2% ever having been accessed after it was first stored.<p>On a related note, probably a similar percentage of people claim on their car insurance. If only the rest realised they had &quot;crap insurance&quot; and were paying for nothing, they could save so much money!<p>This is obviously sarcasm, but I think it&#x27;s important to remember that much of the data is stored because we don&#x27;t know what we will need later. Photos of kids? Maybe that one will be The One that we end up framing? Miscellaneous business records? Maybe those will be the ones we have to dig out for a tax audit? Web pages on government sites? Maybe there will suddenly be an interest in obscure pages on public health policy if a global pandemic happens.<p>Complaining that data is mostly junk is not a particularly interesting conclusion without acknowledging this. Is there wastage? Yeah sure, but accuracy on what needs storing is directly traded off with time spent figuring that out, and often it&#x27;s cheaper to store the data.

About a decade and a half ago I worked on a large data migration project at a FAANG. Multi-exabyte scale, many clusters across many countries. Once everyone was moved the old storage platform wasn&#x27;t completely empty, because the number of migrations was large and users were (naturally) more focused on ensuring their data was in place and available on the target platform rather than ensuring every last thing was deleted on the legacy platform. We weren&#x27;t initially concerned about it because it would all get deleted when we turned down the old setup.<p>As we were gearing up to declare victory and start turning down the several dozen legacy storage clusters someone mused that given some users were subject to litigation holds -- not allowed to delete any data -- that at least some of the leftover data on the old system might be subject to litigation hold, and we&#x27;d need to figure that out before we could delete it or incur legal risk. IIRC the leftover &#x27;junk&#x27; data amounted to a few dozen petabytes spread across multiple clusters around the world, in different jurisdictions. We spent several months talking with the lawyers figuring that out. It was an interesting dance, because on the one hand we were quite confident that there was unlikely to be anything in the leftovers which was both meaningful and <i>not</i> migrated to the new platform, while on the other hand explaining that it wasn&#x27;t practical to just &quot;go and look&quot; through a few dozen PB of data. I recall we ended up somewhere in between, coming up with ways to distinguish categories of data like caches and working data from various pipelines. It added over six months to the project, but was quite an interesting problem to work through that hadn&#x27;t occurred to any of us earlier on, as we were thinking entirely in technical terms about infrastructure migration.

Some fun math: according to some estimates there is 175 zettabytes of data worldwide. Assuming 20-terabyte harddrives, this could be stored on 8.8 billion drives. Assuming 10 drives per rack unit, 42 RU per rack cabinet, 16 square feet per cabinet (including aisle space), that means you need about 330 million square feet of data center space to host this data. If it was hosted in a single square data center, it would be 3.5 miles long and wide. (I always like to picture the physical space something would occupy.) And energy wise, assuming 5 watts per drive, it would consume 44 gigawatt, so it could be powered by about two large hydro dams similar to the Three Gorges Dam (22 gigawatts capacity). I am assuming a PUE close to 1.0 for simplicity. Of course one would <i>not</i> be able to spin up all these drives at once, since a drive spinning up consumes about three times more power (15 watts). So you would definitely staggered spin up when booting the servers :-)<p>If 90% of this data is &quot;crap&quot; and could be cut down, it would still be just a drop in the bucket compared to worldwide energy use.

I agree with the general sentiment here, but don&#x27;t like the examples. 200 photos per person per year isn&#x27;t very much! That is all fine.<p>What really bloats things out is surveillance (video and online behavioral) and logging&#x2F;tracking&#x2F;tracing data. Some of this ends up cold, but a lot of it is also warm, for analytics. It bloats CPU&#x2F;RAM&#x2F;network, which is pretty resource intensive.<p>The cost is justified because the margins of big tech companies are so wildly large. I&#x27;d argue those profits are mostly because of network effects and rentier behavior, not the actual value in the data being stored. If there was more competition pressure, these systems could be orders of magnitude more efficient without any significant different in value&#x2F;quality&#x2F;outcome, or really even productivity.

Don&#x27;t forget emails.. I have everything I ever sent or received, and I have it backed up. I expect 90% of my inbox is the jpg signature logo they attach to the bottom of my clients email rather than hyperlink.

Isn&#x27;t this just a specific case of sturgeons law?<p><a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Sturgeon%27s_law" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Sturgeon%27s_law</a>

To repurpose an old saying: &quot;90% of my data is crap. I just don&#x27;t know which 90%&quot;

&quot;We’re destroying our environment to (...)&quot;<p>No we&#x27;re not. I really dislike this &quot;environmental&quot; anti-technologist angle. A single steel plant in china has tenfold &quot;environmental impact&quot; than all photos stored on a platter everywhere.<p>Would you prefer the photos are a cocktail of weird chemicals on a negative and printed on glossy photo paper?<p>Digital data is the most ephemeral we are able to make it through vast effort.

Storage being cheap enough that it&#x27;s not worth policing doesn&#x27;t seem very consistent with it being expensive enough to include much energy use (what I assume the &quot;destroying the environment&quot; hyperbole is referencing).

A gross underestimation, IMO. When I was in big data, fewer than 5% of data written was ever touched again, and only a single digit number of our large customers (out of tens of thousands) actually made real use of their &quot;big data&quot;, and created most of the load. That&#x27;s the trouble with &quot;checkbox driven development&quot; - 10 years ago you were required to have a &quot;big data strategy&quot; for anyone to take you seriously, even if your strategy boiled down to just ETL-ing a bunch of crap you&#x27;re never going to need into the cloud and never touching it again. Now I&#x27;m in AI, and the same thing is happening to AI. It&#x27;s great if you&#x27;re selling shovels, so to speak, but not so great if you plan on selling them for an extended period of time.<p>This, by the way, has implications on storage systems design. You want something that&#x27;s cheap yet dense to encode, potentially at the slight expense of decode speed. Normally people really lose sleep about decode speed first and foremost, which, while important, does not minimize the overall resource bill.

We recently discovered we store 500MiB of email tokens (expired). Then a copy of them in history table. Then this data is replicated. Then there are total of 4 backups of it. Backups that are done every 2 hours for half of each day. And we store those backup for up to half a year. I don&#x27;t even want to add that up...

The other day I had to go through 15 years old PowerPoint files to grab the originals of pictures made by a guy extremely proefficient at creating detailed artwork from PowerPoint forms that&#x27;s now retired. Was can now render them to full-hd PNGs instead of the 256x256 BMP files we were using before.<p>Storing &quot;useless&quot; data makes financial sense.

90% of libraries consist of books that are never opened. These books were all produced by destroying and processing trees, sometimes with toxic chemicals, and their information density is orders of magnitude lower than that of a hard disk or SSD. Same with photo processing, where 90% of photos taken are discarded, and the toxicity of the chemicals is even higher.<p>So the question isn&#x27;t simply whether storage is wasted; it&#x27;s how much waste there is relative to the environmental impact. Granted, books and photographs don&#x27;t need to be continuously fed energy to make the information available. However, the cost of storage is now so cheap that even with 90% waste, it&#x27;s economically viable to keep it online. So the problem, if you can call it one, is that energy is too cheap, and externalities are not accounted for in the cost.

There&#x27;s another dimension to this, that storage is <i>so cheap</i> that being wasteful with it isn&#x27;t really disincentivized. I know for example at work of a portal that accepts uploads of large files from external clients that stores both the initial upload and every subsequent transformation of the file (of which there are 4-6) permanently. It&#x27;s extremely useful for debugging, as one of the bits of metadata we shove on the zip archive is the git hash of the code that was running, so it&#x27;s trivial to pull down any failed step and diagnose what happened.<p>We are using 4-6 times as much storage as we need to, and these are often not small files (on the order of 100 MB - 5 GB, several dozen times a day) but fixing this overuse is so far down the priority list that I don&#x27;t think it survived the great Jira purge of mid-2024.

It&#x27;s pretty simple: sorting through the data to determine which 90% is crap is more expensive and uses more of our scarce resources than just storing it all.

When traveling, a funny thought I always have is watching other tourists take the same photos, from the same exact location, knowing it will be backed up to iCloud. I can&#x27;t help but imagine how much disk space is taken up by duplicates of the exact same photo.

Make the cost of operating data centers reflect the damage to the environment and see how quickly people optimize. I don’t know how damaging storage is, but that’s the only way to make a difference.

&gt; The Cloud made the crap data problem infinitely worse.<p>This article is mainly focusing about the unused data by website and enterprise databases, only toward the end of the article it barely touched upon &quot;the elephant in the room&quot; of data in cloud.<p>Now everywhere in the world data centers are being built at breakneck speed to cater for the AI data modeling, training and serving. Most of the AI based data are being kept in datalake in the form of raw data that will probably never see the light of that day i.e never being processed.<p>Bill Inmon warned us against this potential data swamps in data center due to the increasing popularity of the datalake [1].<p>Hopefully open table format like Apache Iceberg can rectify this unused raw data epidemic but time will tell [2].<p>[1] Lakehouses Prevent Data Swamps, Bill Inmon Says<p><a href="https:&#x2F;&#x2F;www.datanami.com&#x2F;2021&#x2F;06&#x2F;01&#x2F;lakehouses-prevent-data-swamps-bill-inmon-says&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.datanami.com&#x2F;2021&#x2F;06&#x2F;01&#x2F;lakehouses-prevent-data-...</a><p>[2] What Are Apache Iceberg Tables and How Are They Useful?<p><a href="https:&#x2F;&#x2F;www.snowflake.com&#x2F;guides&#x2F;what-are-apache-iceberg-tables&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.snowflake.com&#x2F;guides&#x2F;what-are-apache-iceberg-tab...</a>

&gt; We’re destroying our environment to store copies of copies of copies of stuff we have no intention of ever looking at again<p>While I agree that most of the stuff in data centers is probably crap, but that&#x27;s because most of everything people do is crap. That&#x27;s not for me to decide though, people save things because they find value in them. Most of what has value to another person won&#x27;t have value to you. Most of what people treasure in their life is thrown away after they die because nobody wants it, even their closest family members. Who gets to tell everyone the bad news, that objectively their memories are trash and they don&#x27;t have a right to keep them anymore? Gerry Fuckin&#x27; McGovern?<p>Secondly, we aren&#x27;t destroying the environment for any of this. Data centers use like 5% or less of the overall electricity use. It&#x27;s a lot, but we don&#x27;t have to put datacenters in random locations, we can (and do) put them where electricity is cheap. That generally means that the 5% of electricity used for data centers, kwh for kwh isn&#x27;t as impactful as an average kwh of end use. Large companies like Meta and Google claim to have zero net carbon by obtaining offsets. So in general we aren&#x27;t &quot;destroying the environment&quot; to store copies of photos.

I have seen takes about the environmental footprint of unused data before, and I have severe doubts this is a relevant issue.<p>I mean, sure, there is some impact. Storage media has to be produced. But there&#x27;s a reason storage is cheap, it&#x27;s not a whole lot of resources going into it. And hard drives that are idle in some data center without being accessed don&#x27;t consume a lot of electricity.<p>There are very real and concerning problems with the environmental impact of IT. But they are primarily found in other areas. Energy consumption is mostly a function of &quot;how much you compute with data&quot;, not &quot;how much data you have&quot;.<p>In other words: be concerned about so-called &quot;AI&quot;, be concerned about Bitcoin. Don&#x27;t worry about unused data too much.

There are a few arguments here, and yeah, I get it, some of the shit being stored is shit — or shitty versions of content already stored. But the argument that “these pages haven’t been reviewed in 20 years!” is <i>exactly</i> the reason for preservation! We want to be able to read, listen, and review content that is rare.

In this book &#x27;World Wide Waste&#x27; he states &#x27;Every time I download an email I contribute to global warming.&#x27; Is this true? Aren&#x27;t some data centers, Google for example carbon neutral?

It&#x27;s a problem for us when it comes to the GDPR and rights to be forgotten; companies will say they store your data carefully, will say they have shown you everything they have, will say they delete it, but &quot;the company&quot; in aggregate has no idea there&#x27;s a thousand SharePoint sites and ex-employees mailboxes and filestores, and copies of old filservers from before a migration, and test databases containing copies of real data from a half-abandoned project where new management fired the contractors and then never got around to hiring new ones.

I recently worked on a project that proposed using ancient cave systems, full of priceless stalagmites as data centers. The first step in the operation? Raze the cave and cover everything in a thick layer of cement. Use of the underground water table was recommended to cool the AI machines. It was all very dystopian. It was even suggested that the national park system be sold off to build these mega data centers underground.

I hate searching for vintage computer stuff only to find broken hyperlinks to lost pages that once contained useful information or software but is now gone forever.

Some fair points, but some &#x27;crap data&#x27; has storage requirements orders of magnitudes from each others.<p>One short video can equal a year worth of emails for someone. Similarly those many webpages that don&#x27;t get viewed often probably require only a negligible amount of resources to keep online and might help someone who&#x27;d otherwise be faced with linkrot.<p>Best to focus on the low hanging fruit.

&gt; the cost of storing data is less than the cost of figuring out what to do with the crap.<p>Or the cost of figuring out that it&#x27;s not worth saving...

Not my area of expertise but I don’t think storage is a big deal, relatively speaking. It’s all about the compute

So if its 90% dead cold data, the temptation must be there for &quot;ocassional hit&quot;data, to switch from ssd&#x2F; hdd in a server to some slow access magnetic tape with a delayed reply..

I imagine there&#x27;s a similar statistic for book shops and public libraries.

90% of everything is crap. I think there&#x27;s a law about that or something

Instagram and Facebook, at least, will compress a 4K video down to 360p if it doesn&#x27;t get any views for a couple of months. So, there are some mechanisms to degrade the crap.

Must be more, surely.

&gt; Why were they created?<p>Proof of work. Look at all this data I&#x2F;we created.<p>And the article didn&#x27;t talk about logs and other operational data yet

Burn the libraries, they&#x27;re full of books no one&#x27;s read in years &#x2F;sarcasm

Surprised it isn&#x27;t higher

Surely a few more nines if the base rate is already Sturgeon.

Without rules, metadata, systems to organize files even the most valuable data can be stored as a steaming pile of poo. In my case I was diving into a compressed tar containing compressed tar-s, which again may have giant uncompressed data files inside. The whole pile of crap had premature EOF so for sure some non trivial amount of data was lost. And on the top of it what was preserved should have a prefix: &quot;unobtanium_&quot;. Storing several tar.gz several terabytes in size on some puny domestic use NAS made extraction of the files a crying game.<p>In another site I found a mix of some old version windows disk images with data. With more crap inside.<p>In the end: storage may be cheap. Storing piles of disorganized crap is very costly if you want to find something

Photos not seen by humans again, but plenty of value for the AI overlords to examine. These things have value again.<p>Didn&#x27;t Facebook start to move most of their least-used data onto optical arrays a long time ago?

I wonder what this guy thinks about Internet Archive?

Fortunately with LLM&#x27;s we can generate data and not store it. If you think about it that means articles could always be relevant with additional details added as time goes on.

That&#x27;s a pretty optimistic estimate, tbh.

&gt; because when it comes to technology, these managers exist on a whole other level of stupid vanity and narcissistic pursuit of their own selfish agendas<p>Regardless of what you think about the article. This rings so true at many Fortune 500 companies.<p>The number of times I have seen teams work through pointless bullshit to push some meaningless objective for the company. Just so the middle manager (aka “Director of SVP of X product of Y branch”) can get a bullet point(s) in the quarterly “all hands”.<p>Oh and those 10 developers&#x2F;off shore people that were just hired? It was all to pump his&#x2F;her “head count” number to get to the promotion to next grade&#x2F;level.<p>Then when that person gets promoted, those people get scattered throughout the firm or just let go.<p>It’s truly just weaponized incompetence.

&gt; were destroying our environment<p>citation needed

Thats why human brains forget

One man&#x27;s crap is another man&#x27;s crap. My data, which may be 100% crap to you, is obviously not (all) crap to me. I often unerase USB drives, HDDs, and SD cards find pictures, movies, etc. Obviously, they don&#x27;t have much value to me so lots of it is crap but sometimes there&#x27;s gold hidden in the crap if you dig.

&gt; Having to deal with senior managers has always been the most unsavory part of my job, because when it comes to technology, these managers exist on a whole other level of stupid vanity and narcissistic pursuit of their own selfish agendas.<p>Whoever sent this dude made a mistake. People who don&#x27;t share your worldview need to be persuaded, not insulted! Some dude stomps in, thinks all the snaps in the cloud are crap, things the big bosses are stupid for not instantly deleting the pictures they saved into the cloud.... and then what? Download Lisp? Thought we got over this, pal.<p>WORSE IS BETTER.<p>P.S. do not erase our porn. WORSE IS BETTER.

It seems you&#x27;re all missing the point here: it&#x27;s not about storing useless data, it&#x27;s about destroying the environment in the process. I understand you all want to keep everything, just in case, because it&#x27;s cheap and you don&#x27;t see the externalities. But there are externalities, and they are big.

WON&#x27;T SOMEBODY, PLEASE, THINK ABOUT THE ENVIRONMENT!?

Not too related but... Did you ever tried to imaging a world where computing is back to interconnected desktops&#x2F;homeservers? Well, if you see trends like {fog,edge}-computing or recent ideas to distributed LLM computing (BrianknowsAI DCI Network for instance) it&#x27;s clear that giants well known they can&#x27;t keep up the modern mainframe model preferring to makes users paying the iron and bandwidth while they handle the software stack.<p>Now it&#x27;s clear the new deal could be implemented only in homes&#x2F;sheds with domestic p.v. and storage, smart cities keeps to fail since the ancient Fordlandia, see Neom, Songdo, Masdar, PlanIT Valley, Lavasa, Ordos, Santander city, Toronto Quayside (Google Sidewalk Labs), Amazon HQ2, Egypt new Cairo still nameless, Modi&#x27;s Indian 100-smart city program, Arkadag, Innopolis, Nusantara, Proton City, ... and can&#x27;t be powered with a smart-grid at such scale.<p>So well, new well insulated buildings, with ventilation of course, with p.v. and storage, with room for a domestic rack(s), with FTTH. Anyone with such settlements could have his&#x2F;her own &quot;datacenter&quot; at home, following the same trend for medical devices more and more cheaper and smaller. A LOM? Well a NanoKVM PCIe or an external JetKVM cost MUCH less than classic LOM and do much more. We have all the gear to makes such &quot;datacenter at home&quot; assemblies, anyone holding preferred crap and participating in distributed computing networks to pay at least a bit gear and bandwidth.<p>It&#x27;s not for all of course, some will be trapped in dense cities while some large owner dream an obviously not possible conversion from offices to apartments and datacenters like <a href="https:&#x2F;&#x2F;finance.yahoo.com&#x2F;news&#x2F;southern-californias-hottest-commercial-real-100007648.html" rel="nofollow">https:&#x2F;&#x2F;finance.yahoo.com&#x2F;news&#x2F;southern-californias-hottest-...</a> or <a href="https:&#x2F;&#x2F;www.euronews.com&#x2F;next&#x2F;2024&#x2F;02&#x2F;29&#x2F;madrid-to-convert-underused-offices-into-flats" rel="nofollow">https:&#x2F;&#x2F;www.euronews.com&#x2F;next&#x2F;2024&#x2F;02&#x2F;29&#x2F;madrid-to-convert-u...</a> and <a href="https:&#x2F;&#x2F;www.theguardian.com&#x2F;society&#x2F;2025&#x2F;jan&#x2F;05&#x2F;office-to-homes-conversions-london-blocks-hold-fresh-allure-since-shift-to-home-working" rel="nofollow">https:&#x2F;&#x2F;www.theguardian.com&#x2F;society&#x2F;2025&#x2F;jan&#x2F;05&#x2F;office-to-ho...</a> or <a href="https:&#x2F;&#x2F;czechdaily.cz&#x2F;half-of-pragues-office-buildings-are-aging-a-shift-towards-residential-spaces&#x2F;" rel="nofollow">https:&#x2F;&#x2F;czechdaily.cz&#x2F;half-of-pragues-office-buildings-are-a...</a> etc for all over the developed world. That&#x27;s while we admit <a href="https:&#x2F;&#x2F;doi.org&#x2F;10.1073&#x2F;pnas.2304099120" rel="nofollow">https:&#x2F;&#x2F;doi.org&#x2F;10.1073&#x2F;pnas.2304099120</a> we need a full-remote DISTRIBUTED shift.<p>Food, meds, general retail distributed by a single integrated logistic platform for maximum efficiency in a spread society, the IT evolution makes Distributism possible.<p>Doing so erase the big amount of concentrated energy, dense network, heat handling and water problem of datacenter and also reduce much the crap, because anyone keep it&#x27;s own personal and being not free keeping it they&#x27;ll learn to be storage continuous.

&gt; We’re destroying our environment to create and store trillions of blurred images, half-baked videos, rip-off AI ‘songs’, rip-off AI animations, videos and images, emails with mega attachments, never-to-be-watched-again presentations, never-to-be-read-again reports, files and drawings from cancelled projects, drafts of drafts of drafts, out of date, inaccurate and plain wrong information, and gigabytes and gigabytes of poorly written, meandering content.<p>Storing the files for Mr. McGovern’s website requires plastics, metals, power and physical space, yet I assume he believes that environmental effect is worthwhile. Who is he to decide for others that their choice to pay for the storage of data is not equally worthwhile to them?<p>That’s the beauty of a price system: each of us gets to decide what we will buy, and what we will not buy.<p>Now, perhaps his argument should be that the price of storing digital data does not adequately reflect the true cost. Perhaps there are unaccounted-for externalities. If so, then he should make <i>that</i> argument, perhaps arguing for a tax to align prices with costs.<p>Someone else might argue that data is a liability as well as an asset. That’s another argument he could make.<p>But haranguing folks for spending their money in ways he doesn’t like doesn’t seem likely to produce the outcome he appears to wish.