Slashdot Mirror
Kim Dotcom's Mega Claims 1 Million Users Within 24 Hours
Kim Dotcom's new "Mega" cloud service appears to be a hit. According to Dotcom over 1 million have signed up for their free 50 gigabytes of storage. Although that is about 1% of the Dropbox user base, it's not a bad start. From the article: "Mega quickly jumped up to around 100,000 users within an hour or so of the site's official launch. A few hours after that, Mega had ballooned up to approximately a quarter of a million users. Demand was great enough to knock Mega offline for a number of users attempting to either connect up or sign up for new accounts, and Mega's availability remains spotty as of this articles' writing."
Considering the reputation that megaupload had, I don't think he'll have any problems getting users. I think, like so many other websites, he will have trouble monitizing the service without becoming obnoxious.
...
I'm sure adblock will deal with the obnoxious ads
- Nec Impar Pluribus, or so I'm told.
That's why you spend the next few weeks downloading porn, followed by the next few months uploading it all to Mega and freeing the space in your hard drive, and then... you'll have to download it *again* from Mega just to be able to watch it.
The patchy availability will be resolved soon I hope, but there's a major flaw I ran into, which is that when you sign up it doesn't ask you to confirm your password by typing it twice. This means you can make typos without realising it. Because the password is also an encryption key, you can't reset it. You can't delete the account either, nor can you register two accounts to one email address. I made a typo in my password. Net result: I permanently can't access my account, nor can I register a new one with my preferred email address.
is it encrypted transmission and storage? otherwise its just another dropbox clone. also, 1st post!
Yes, Yes, No it's not, and no you weren't.
Be seeing you...
I'm sure adblock will deal with the obnoxious ads ...
But isn't that their monitizing plan? To you mega you will need to run their ad blocker which replaces normal advertisments with ads from mega.
http://michaelsmith.id.au
This weird criminal somehow has 50 GB * 1,000,000 = 47.6 petabytes of enterprise storage? Without getting one dollar? How is this paid for? Not to mention all the data traffic back and forth which will be even more expensive?
1. Not every user is using 50gb.
2. He has lots of money.
3. He is investing in a new enterprise and knows that he has to spend money first in order to make money in the future.
I assumed all that was fairly obvious. What's your theory, by the way?
I think, like so many other websites, he will have trouble monitizing the service without becoming obnoxious.
I assume he may be going for paid premium accounts
When I use a free (valuable) service, I always consider (and sometimes purchase) the premium account. Seems fair.
Try an E-mail multiplier, such as SpamGourmet.
You can set up any number of separate E-mail addresses which get forwarded to your main E-mail, and if you set mega as the "exclusive sender" there's no limit count on that address.
Depending on the backend SAN he has, you can use thin-provisioning since there will not be a demand from all users for the entirety of their storage immediately.
He also doesn't need thin provisioning, to have a file system smaller than the capacity promises.
It's very possible also, the storage organization is not just your files uploaded stored on a filesystem verbatim.
Your files, might, for example be divided into 4K or 8K chunks.
And each 'chunk' instead of each file, might be persisted to a file on a filesystem/volume somewhere, with a "chunk ID" based on a crypto hash of that chunk, placed in a directory based on its ID; different pieces might be on different volumes as required.
There may be a database table indicating what files you have on your account, and what "chunk IDs" in what order belong to the files you uploaded.
Chunks might be identified by checksums, so if you have two users with a very similar file, or a very similar 1K, 2K, 4K, or 8K block. The different users' files will share the exact same "chunks".
Instead of having to store the same exact chunk multiple times.... then you increase an "in use" counter on the chunk, and have records from both separate user accounts pointing to the same "chunk"
The result is... you could upload 50GB of data, but use zero, or very close to zero extra space on the server --- as long as all the files you uploaded, are identical or 99% the same as some file uploaded by a different user of the service.
Links in the summary... NONE of them to the actual service. Brilliant!
Here is the actual site: https://mega.co.nz/
I assumed all that was fairly obvious. What's your theory, by the way?
I'm thinking all those elves that are unemployed now that Christmas is over are working hard for Mr Kim.
Interesting. Mega seeks to achieve profitability by sharing revenue with participating artists - creating a channel with as little rent-taking as possible. As opposed to the super-rent-seekers: today's media and telecom conglomerates.
Kim says Megaupload was killed by the Obama administration, as a gimme to the media cartels - in return for financing and as a replacement for failing with SOPA. I'd add that Megaupload was SPECIFICALLY targeted over Eastern European hosters for enforceability, and over others because of Dotcom's incipient "MegaKey" agreement with big-name urban artists.
So, from where will the source of this revenue come? Ads are obvious - but really another nut to crack. I don't think this is what the new Mega has in mind for a foundation pillar.
Rather, I suspect that the artist agreements are expected to drive enough subscriber interest, for real takes, vs. simple freeloaders. The volume of signup in the past 24 hours is a great validation for Dotcom, if prospective participants need prompting.
"Flyin' in just a sweet place,
Never been known to fail..."
The artists want out of these RIAA handcuffs as badly as do their fans. They see there is a different, more direct model that doesn't fatten the talentless go-betweens sitting in air-conditioned offices, producing no value at either end of the production pipeline.
Sorry, Mr. Ego Hat, David Geffen.
"Flyin' in just a sweet place,
Never been known to fail..."
This would work if the files weren't encrypted.
I'm pretty sure everyone loves to hate the RIAA/MPAA so Kim Dotcom had little trouble rounding up support when they moved to shut down MegaUpload.
Unfortunately, he's now picking a fight with bigger opponent and possible a mass of small website owners who rely on their Adsense revenues to help pay the bills.
Kicking the RIAA/MPAA for their sins is one thing, taking money out of the mouths of independent content creators (by hijacking their ad-revenues to fund his Mega-services) is something altogether different.
I admire KD for what he's doing with the MegaKey service but I really wonder if he's got an oar out of the water in picking a fight with Google and the many websites who rely on that company's ad-revenue sharing.
BTW: I'm one of those sites and I'll be mighty pissed if Kim starts replacing the ads on *my* webpages that should be generating money to pay for *my* efforts -- because I have *nothing* to do with MegaKey so why should *I* be paying for it?
Mega encrypts files using the hash of the content as encryption key, this allows them to have dedup without knowing how to decrypt the files.
That is called deduplication and most modern SAN systems have this feature. You can have both thin-provisioning and deduplication for increased savings. In Mr. Dotcoms business model I doubt he will get many exact duplicate files, but that really doesn't matter because you can still deduplicate similar binary strings within differing binary files or as you said duplicate blocks. In any case dedupe and thin-prov are not mutually exclusive, you can do both.
Normally dedupe is more efficient for backups or when used on the disk target for a virtual environment since you only need one copy of notepad.exe if you are hosting 200+ windows servers. The same applies to unchanging files in *nix systems. The thing is you *have* to have some way to "present" the 50GB of promised space. While you may use dedupe or any other method to reduce your storage footprint the end user wants to see that storage. You either have to present that space raw, which comitts it from the SAN or as a thin-provisioned LUN with only the bare minimum of space actually reserved. How you store those files after the fact is up to you as the hosting company, but if you promise 50GB of space the user will want to see that space available.
Chunks might be identified by checksums, so if you have two users with a very similar file, or a very similar 1K, 2K, 4K, or 8K block. The different users' files will share the exact same "chunks".
Instead of having to store the same exact chunk multiple times.... then you increase an "in use" counter on the chunk, and have records from both separate user accounts pointing to the same "chunk"
That is a good description of data compression - see LZW, for example. Unfortunately, encrypted data can't be compressed. Any sufficiently strong encryption algorithm produces a bitstream which is virtually indistinguishable from random. For any given size dataset, random data does not repeat in sufficiently long sequences to achieve useful compression relative to the dataset size.
Stop-Prism.org: Opt Out of Surveillance
Your encrypted data, you mean? I don't mind them selling my encrypted data, honestly. Would take more time to unencrypt it than it's really worth and they'd just lose money. Renting out those botnets DOES cost something and it'll take them a while to break AES128.
Nobody cares what the CAPTCHA for your post was.
If it is anything like megaupload, the differences on the first tier and free account will be the download links. You get a faster download and the people you give the links to does not have to wait for a specific server or be limited in speed because of public servers being overloaded.
This is probably only valuable if you are hosting files for work or something and need a quick way to disseminate them outside the building. Most smaller companies who are not into web services do not have a lot of extra upstream bandwidth.
but I don't have 50GB of porn to fill it...
You need to hand in your geek card. Immediately.
Deduplication can be done perfectly well on a file level even for encrypted files. If this is anything like PGP, you could do it by having multiple encrypted messages attached to the encrypted file (or linked to the file), one for each person with access to the file. Each message contains the same [unencrypted] data — a decryption key for the file — but can only be decrypted by the person who the message is for.
In a well-secured system, all these [small] decryption messages would be stored in a random order in the same place, so you attempt to decrypt all these messages with your personal key in order to retrieve the decryption key for the file. That way, even if you have access to the decryption messages, you can't tell if any particular person will be able to decrypt that file without knowing their decryption key. A more reasonable compromise (faster, a bit less secure / deniable) would be for each person to have their own decryption key store, which is queried whenever they want a particular file.
As long as the file encryption is done on the client side (in the same way by all clients), the server can do deduplication without having any idea what the file contains.
Presumably Mega has the encrypted hashes (or not, it doesn't really matter), which are useless for accessing the contents of the file without access to the keys that are used to decrypt and extract the file hashes. Mega does not need to store these user keys on the server, so they probably don't.
However, Mega should be able to identify files stored on their server if they have access to the original files. Consider if a Mega has access to a file called 'The_HoBBiT_Crazy_Delta_XxXDVDRip_firsTPost.avi' (e.g. they were provided the file by some company who wanted to check for compliance with licensing). They use the Mega client system to encrypt that file, producing an encrypted file and a file decryption key. They then check to see if any files on the server have an exact match to that encrypted file (e.g. first by hash, then by comparing length, then by an exact comparison of the encrypted bytes). It would then be possible to delete the file, check to see who has uploaded this particular file, and produce a list of people who also have access to that file.
[mostly copied from my response(s) elsewhere]
Ask me about repetitive DNA
That is called deduplication and most modern SAN systems have this feature. You can have both thin-provisioning and deduplication for increased savings.
It's true that SANs have deduplication functions, but there are a few problems with SAN deduplication -- the main one, is this typically operates on a 'volume', LUN, or disk level. A SAN can't deduplicate data stored across storage systems, so there are scalability limits.
That is this doesn't scale very well to large numbers of volumes, with intentional duplicates for performance... and when you need to make backups of the filesystem, there is a tendency of the backup or replication target to Re-Duplicate the deduplicated data.
An application that actually implements this functionality can achieve far better characteristics than a SAN is capable of, because the application would be aware of consumers of the data, and logical pairings.
And at lower cost, by using non-specialized storage hardware
There are perhaps 2 SAN vendors out there with a scalable dedup option, and they are both very expensive per GB of storage options that would not be a very tenable thing to base a free service off of.
That is, the dedupping SAN vendors charge 10 to 20x as much per disk drive of a given capacity, than you can buy off the street, and claim efficiency, lower cost, lower power consumption because dedup means you get twice as much storage.
(E.g. SANs are expensive, and the storage vendors want to reap profit advantages by selling dedupped storage, for more, as well... which can be as much increase in cost as dedup supposedly saves)
Mega's availability remains spotty as of this articles' writing."
So it's only partly Cloudy.
You don't even NEED a lot of money to get 50 PB of storage.
Granted, you need some, but it's a lot less than people think.
18 months ago, BackBlaze showed how to build a 135 TB server for $7,384, and the price would be just about the same today.
That's $56,696/TB for a total of $2,834,800
For what Kim has in mind for Mega, 3 million in storage hardware isn't exactly surprising. In fact I'd be surprised if they haven't budgeted for a lot more than that.
The connection to New Zeland is via undersea cable over a monopoly called the Southern Cross.
http://en.wikipedia.org/wiki/Southern_Cross_Cable
Is one fiber cable going to be able to handle the traffic?
The truth shall set you free!
2. He has lots of money.
3. He is investing in a new enterprise and knows that he has to spend money first in order to make money in the future.
I assumed all that was fairly obvious. What's your theory, by the way?
http://arstechnica.com/tech-policy/2013/01/building-mega-ars-pre-launch-interview-with-kim-dotcom/
The Mega business plan will be a distributed model, with hundreds of companies large and small, around the world, hosting files. A hosting company can be huge or it can own just two or three servers Dotcom says--just as long as it's located outside the US.
"Each file will be kept with at least two different hosters, [in] at least two different locations," said Dotcom. "That's a great added benefit for us because you can work with the smallest, most unreliable [hosting] companies. It doesn't matter because they can't do anything with that data."
More than 1000 hosts answered a request for expressions of interest on the Mega home page. Dotcom says several hundred will be active partners within months. Successful hosts will get paid E500 per month per server; each server needs to supply 24 hard drives with 72 terabytes of storage and one gigabit of bandwidth, among other requirements.
That's all down the road, however. For now, Mega is launching with just one, professional, hosting operator--a subsidiary of Cogent, based in Dotcom's home country of Germany.
According to other articles, he has a (maxed out) 10Gbit pipe from this Cogent subsidiary
And FYI - Cogent was the US host for megaupload.com, so they believe in his business plan enough to host for him again.
If he can get Mega back into the big leagues again, it's going to put some serious strain the undersea fiber that feeds the USA.
That's the most expensive wired bandwidth around and he's planning to host nothing in the USA.
[Fuck Beta]
o0t!
Yes, together with the magic Unicorns that guard your data and the gnomes that print the money to pay for it all.
No, it's not a sting operation run by the US government that has made a deal with someone looking to a couple years of prison who desperately doesn't want to go there. They would never do that. He would never do that. None of them has done it before...
Assorted stuff I do sometimes: Lemuria.org
Not only that, to address a lookup table with 2^65536 elements you'd need an 8K index, or in other words exactly as much storage to store the addresses as the original data.
Dedupe should NOT work if every file is encrypted.
IF they do the crypto right, even if the exact same file is being encrypted it will NOT result in the same encrypted file. If it does it means they are doing the crypto wrong!
And with strong crypto it becomes exceedingly unlikely that you'd have duplicate blocks - assuming a block size that's 512 bytes or larger. If you find significant numbers of duplicate blocks it means something is wrong somewhere. What are the odds that 512 random bytes will the same as another 512 random bytes? If you flip a coin 4096 times, what are the odds that you'll get the same result sequence again if you flip a coin another 4096 times? That's basically what decent crypto is aiming for - it looks random. If it turns out to be far from that random, you've found a flaw in the crypto or the system.
You might have dupe blocks in the metadata of the file system that's handling those encrypted files but that might not be good enough, or you might not want to dedupe those anyway...
Of course for 100% backups they'll have duplicate blocks but you're not supposed to dedupe backups...
You don't even NEED a lot of money to get 50 PB of storage.
Granted, you need some, but it's a lot less than people think.
18 months ago, BackBlaze showed how to build a 135 TB server for $7,384, and the price would be just about the same today.
That's $56,696/TB for a total of $2,834,800
For what Kim has in mind for Mega, 3 million in storage hardware isn't exactly surprising. In fact I'd be surprised if they haven't budgeted for a lot more than that.
(You did ask for correction if you were wrong...)
$7,384 for 135 TB is $54.70/TB or ~$56,000/PB.
This should only serve to strengthen your point of course.
He is referring to the malpractice insurance the doctors need that inflates the cost of repairing your kid's broken leg.