Packet Juggling - Floating Data Storage

Filthmaster writes "I just saw an interesting paper that has been posted to bugtraq, full-disclosure and vulnwatch. It deals with the principles of stealthily using network infrastructure as either short-term or long-term storage. Not sure if I'm ready to implement it, but it makes interesting food for thought." There's also a mirror up.

Comments

If this were quick enough

[Amiga Lover]

If it were quick enough and timed well enough, could network bounced packets be enough for some really quick swapped out swap space? gig ethernet gets around 1ms to my next machine, and thats 10 times quicker than my hard drive (10ms access time) so if I could store swap in that space, would that work?

I was thinking of tunneling ssh over sms before this, but that sounds just silly now.

it's alive!

[geoff lane]

First you give it comms, then unlimited CPU and now distributed memory.

Can Skynet be far behind...

Re:Bandwidth?

[Zocalo]

Ah, but in this context that could be a good thing since a slower Internet = more latency = longer TTL on your data. Take the ping for example; if it takes twice as long for your ping to echo back with your data, then you only need to retransmit back to storage at half the rate.

I wasn't trying DoS the Internet your honour... I was trying to improve data retention times! ;)

Provisioned storage within the Internet

[FastDownload]

While the authors try to use existing protocols to simulate temporary storage in the Internet, we are working on a scalable, shared storage resource that is open to the community.

We currently have over 20 TB of storage around the world available in the public Logistical Networking Testbed and other groups have another 10-20 TB provisioned in private use testbeds.

In additon to storage, we are also working on providing simple computational services at the storage nodes (work on the data in place while it is stored rather than moving it to computation centers).

For more info, visit the LoCI Lab at http://loci.cs.utk.edu.

Reminds us of the old days...

[ControlFreal]

... in which complete computer memories worked like this: those were called mercury delay line memories, in which pressure waves in mercury lines basically held information.

The UNIVAC I had such an 18-channel memory. More information can be found here, here, and here.

These channels could hold a whopping kilobit!

You don't need the Internet to do this

[melonman]

Just burn a CD a day and post it to a non-existant address on the other side of the world. That way you can probably keep a terabyte of data int he air without taking any space in your office, and, unlike TCP/IP, you may be able to reuse the wrappers.

Main problem with this is...

[zen parse]

Not everyone can benefit, because of side effects of it's parasitic nature.

The amount of storage this system gives in the text would be total available for ALL users of the system. More users, less avaiable storage.

Parasites can do better when there little competition from other parasites, but if the system get's infested, the host it lives of may die. Or someone may develop a cure.

Either way, after a certain threshold, the more popular any system using this gets, the less useful it would be.

Just some random thoughts I had when I was talking about a similar idea with someone.

The BOFH was there already

[red_dragon]

It was 1997 when Simon the BOFH wrote about such a contraption, which won him the IT Idiot Award for Least Intelligent Supervisor.

... This year I've decided to sell the boss on using the network as a storage medium. I casually drop a couple of remarks until the boss decides to channel his massive intelligence away from tying his shoelaces and onto the matter at hand.

"It's simplicity itself!" I cry "We've got these Gigabit Ethernet switches all around the place that we just aren't using! Instead of letting them go to waste we could be sending data continuously around them until it's needed which would actually cut down on the amount of physical disk storage we would need! And just think of the time we would save with read and write latency when the data's already on the net!"
...

This is not new

[eris_crow]

When I was in college back in the Good Old Days (tm) of the Internet, I had a friend who sometimes stored files "in transit", so to speak, by emailing them to himself with explicit routing in the To address. He would send the message on a long circuit of several machines around the country and he had a script to automatically reforward them once they got back if he didn't save them within a certain period of time. Back in the day you could do this by setting the To address to something like "@hostone.com,@hosttwo.com,me@myhost.com" (see RFC 821 sec 3.6) and since the network and the machines on it were much slower in those days, if you added enough hosts then you could introduce a significant delay and have lots of files stored in transit (actually, on the various mail servers) even though your own disk quota was nearly used.

Explicit routing is long gone, but it is an interesting early manifestation of the same principle: the network is my hard drive.

Satellites !

[wtarreau]

I remember that when I was at the University, I explained to someone that with satellites at 37000 km from us, information took a quarter of a second to go there and back to earth. So if you use a 500 Mbps link, you can store 15 MBytes of data in the distance between, on an absolutely zero cost medium, during 0.25 second. And if you were confident enough in the reliability, you could even put a bouncer on earth, which goal would only be to resend the stream to the satellite and keep it looping. You would have 15 Mbytes of free storage with an average access time of 125ms (250ms max). Although absolutely useless, that would be as fun as TCP/IP over pigeon routing :-)

Willy
I will never put a sig.

the heck?

[Frac]

How does using a scarce resource (bandwidth) to create an abundant resource (disk space) make any economic sense?

Headline: How to turn gold into copper! News at 11.