Slashdot Mirror
Lockheed Snags $31 Million To Reinvent the Internet, Microsoft To Help
DARPA has awarded a $31 million contract to megacorp Lockheed Martin which will, with some assistance from Microsoft, attempt to reinvent the Internet and make it more military-friendly. "The main thrust of the effort will be to develop a new Military Network Protocol, which will differ from old hat such as TCP/IP in that it will offer 'improved security, dynamic bandwidth allocation, and policy-based prioritization levels at the individual and unit level.' Lockheed will be partnered with Anagran, Juniper Networks, LGS Innovations, Stanford University and — of course — Microsoft in developing the MNP. Apart from that, Lockheed's own Information Systems & Global Services-Defense tentacle will work on amazing new hardware."
This makes a lot of sense, the military has unique requirements of all sorts, from security to e.g. their inability to hook up an aircraft carrier to fiber (except while at dock) to their need to carry both operational and personal traffic (the latter to keep their people in touch with home) over necessarily constrained links.
I like the bit about "self configuration capabilities to ... reduce the need for trained network personnel and lower overall life cycle costs for network management". While the current state of the art keeps us well employed, things could be easier. Heck, the more the systems I maintain for my parent self-configure, the happier I am.
Yes. Also, how did they decide the effort should cost exactly $31 million of taxpayer money?
... and I can tell you that this sounds like a disaster in the making. LM is so top-heavy with bureaucracy and process-bloat that the company might as well be a mini-Pentagon itself (not so mini, either, now that I think about it). Nothing happens quickly at Lock-Mart, and the things that do happen cost a bloody blue fortune.
If nothing else, they'd better hire in some outside IT guys. If this work gets anywhere near the corporate IT bozos, the military can look forward to a future of XP Pro with daily forced updates, and new hardware every five years or so (which again, is not terribly far away from the way the armed forces IT already works)...
Did you not know that many hospitals already run windows equipment?
sure, in comparison to the piles of money previously given to large contractors to flail around pretending to solve the unique mission critical requirements of the military, its nothing!
USB device? The DoD hasn't been allowed to use those things in almost a year!
I have a bad feeling about this...
no they wouldn't as you say, 'stuff it up'.
They'd patent the sh1t out of it so it is 'stuffed up' for the rest of us.
Remember that the US Military are exempt from patents awarded for work funded by them.
Then all Microsoft need to do is make 'The Internet V2' standard in Windows 8 and watch pretty well every company fall over backwards to implement it.
They would control who the licensed 'Internet V2' to thus kille FOSS, ORacle and probably Apple in one stroke of the pen from the US Patent Office.
Embrace - Done
Extend - Take IPv6, add a few bells & Whistles, patent it
Extinguish - Message from Steve B to Bill G, 'Looks Good'
I'd rather be riding my '63 Triumph T120.
'improved security,'
Like IPSec? Don't fix the network layer, that's pointless. Fix the application layer - run it through TLS or similar if you must.
'dynamic bandwidth allocation,'
Like RSVP on an MPLS circuit? Or like DiffServ?
'policy-based prioritization levels at the individual and unit level.'"
Like CoS?
Seriously, all this has been thought of before - and we ended up with CLNA, IS-IS and networks so complicated it never took off - instead, IP took off because it was easy to use and easy to route.
If we're going to change IPv4 for anything, it should be IPv6 - it's easy to understand, easy to read, easy to process and best of all - ready to use *now*. Many ISPs already have it, and there's a crapload of Usenet traffic/BitTorrent that already goes via v6.
And that is why a company I used to work for making Medical Office Management software replaced all of thier 5000+ installed desktops with a version of Linux I created for them, and dramaticaly cut their support costs. 3 Customer service types, one System Architect (Me), and two developers were easily able to support 5000+ desktops, and around 200 servers, remotely.
Try that with Windows...and you will need many, many more people!
ttyl
Farrell
CAN-CON 2019 - Ottawa's only book oriented Science Fiction Convention! October 18-20, Sheraton Hotel, Ottawa, Canada h
How does this affect pr0n?
If I were implementing it in HTB, I'd do it like this:
prio,rate(%),burst(S,M,L): desc
Everything is guaranteed the percentage (relative to peers) given; IE, the queue with SMTP will get 1% (5% * 20%) of bandwidth as a guarnateed minimum (enough to keep connections alive when other things are bursting hard, and eventually deliver email even if higher priorities never relent).
Extra bandwidth is given exclusively to higher priority bands (ie, lower prio numbers): If there are whole bunch of videoconferences going on between officers in bases about non-immediate military needs (prio 1.0), and suddenly 20 drone pilots need realtime video feeds to interactively fly a coordinated airstrike, the pilots get all the bandwidth they need, leaving the videoconferences only 6% (smart codecs will degrade gracefully; fixed bandwidth ones will just have to call back after the airstrike). Similarly, if they need to VoIP about building a bigger mess, your counterstrike game will lag. FTP gets best effort in between your porn page loads (which burst quickly with the medium-size burst; FTP gets a small burst so it's always ready to yield).
The level of detail you get into for the queues depends on how much bandwidth you have, and how much contention there is for it. If there's high contention, more detail helps more. There are also smarter queueing disciplines than HTB, but it's the simplest to describe like this.
Statically reserved bandwidth guarantees per-connection is better for many realtime needs. With RSVP, each drone pilot can reserve a guaranteed 5% slot for their flow, to prevent problems where there was lots of extra bandwidth, and then a lower priority suddenly needs its minimum guarantee, thus screwing up traffic that was flowing before. IE, it's better to tell the pilot from the start that there's not enough bandwidth that can be guaranteed to them, than to have them start flying and then get jitters when a bunch of troops hit push-to-talk, right as their drone was on final approach.
So in short, porn is pretty low on the list, but not the bottom of the stack. :)
Most tactical systems use UDP so you could argue TCP has already been replaced ;)
The trouble with policy based management at unit / sub-unit level is not with traffic within the unit's AOR but with traffic which crosses multiple unit's networks. Not only that but you have two conflicting isssues: ;) )
Traffic prioritisation based on the traffic type (E.g voip - low jitter requirement) and priotiy based on user needs (E.g Flash, Priority, Immediate, Routine etc... Or to get really stupid flash override override
How to square those two issues is a large part of any problem they will need to address.
--- Users are like bacteria -> Each one causing a thousand tiny crises until the host finally gives up and dies.
It's actually quite a good system, and I really don't think they're going to want to replace all that any time in the next few years.
I would worry a bit about transitioning to anything more complex than what exists already.
Currently, the training for enlisted soldiers who will be the operator/maintainers of the JNN & SSSv3 is 39 weeks long (up from 25). Even with this length of training, there is a lot to be desired. The General Dynamics trainers at the signal school at Ft. Gordon are retired senior NCO's (>E6), but not one has actually used the JNN in the Army. Their experience is all with the old circuit switched comms gear. Knowledge of basic computer networking is seriously lacking for many. So, the end result is that soldiers spend more time learning the maximum length of a CX-11230 cable, memorizing the location of each jack on the signal entry panels, and mopping the floors of the school than actually using the equipment. When soldiers do actually use the gear, it's 100% scripted. The soldiers read the commands off a "cut sheet" and enter them verbatim into the command prompt.
With this level of training, anything more complex than TCP/IP is going to be a no-go unless it's implemented in a very transparent way to the operators.