Slashdot Mirror
What's Your Timeline for IPv6 Migration?
SgtChaireBourne asks: "IPv4 has, over the last 20 years, seen unexpectedly wide adoption. During this time it's proven to be both flexible and robust, but also several problems, though once small, have grown. IPv6 looks to solve some scalability problems, add needed privacy and authentication mechanisms, address quality of service, and provide better routing and addressing capabilities. What kind of timeline does your site/institution/business have for rolling out IPv6 and how?" Those interested in IPv6 migration may also be interested in this article, from a year ago.
None of the organizations I work directly with are even thinking about IPv6.
I'm still trying to figure out the mess that is IPv4! Once I get our internal networks configured as perfect as I can get them, I'll start researching IPv6. Until then, I'll continue to figure out all the problems with the older protocol.
I can't say where I work, but it's on my list of things to look at. I'm looking forward to having an entire block of IP's and not having to use NAT to the Internet. I'll just have to make sure the firewall works right in an IPv6 world.
We have no plans at all to migrate to IPv6. Don't see any need in the next five years.
ISPs are going to have to add support for this in a real-world environment before it begins to really move in businesses. Right now, a fairly complicated tunneling process has to happen before machines using IPv6 can hit the internet in general. Yes, I know you can run IPv6 and IPv4 at the same time, but doesn't that kind of defeat the purpose? Besides, until IPv6 addresses are being assigned by ISPs, the addressing schemes are going not conform to the standard that is finally settled on, meaning that individual addresses will have to change numerous times for people who adopt it early.
Mind you, the above statements are highly uninformed, based on what I've read of IPv6 and my own brief experience setting up a tunnel for it with 6bone (which, I understand, is no longer with us).
End of lesson. You may press the button.
Right now, never. Seriously - not even considering it.
Realistically speaking, I'd say 5 - 10 years, right after I get my flying car.
anyone running anything less than Win2k (that is, if they're running Windows) has to be out of their mind.
There's tons of older software implementations out there. Take the check-in kiosks for airlines: Continental's runs on NT4, and they're still rolling more of those kiosks out every day. One of those situations where if it works, why mess with it, especially when it would just cost more money to convert the existing check-in kiosks in Armpit, Iowa simply to be IPv6 compliant.
What's your damage, Heather?
We'll see but it could easily be another 20 years before the world adopts (wholely) IPv6.
"Killer App" is right, they arent going to take off until there is one.
IMO the killer app for IPv6 is going to be Mobile Phones. I've heard that the next generation phones are going to be IPv6 based using IPv6 Mobility. If this happens, there will be a good reason for people to use IPv6.
For something like mobile phones, IPv6 is really needed - there just arent enough IPs with IPv4 to assign every mobile a unique IP (and mobility typically needs multiple IPs anyway)
were you expecting to see a sig here? perhaps you'd rather see the inside of an ambulance!
..and even sells routing and other IPv6 equipment... and yet, we're not even dreaming of planning of designing a possible IPv6 migration for our own (50.000+ node) network. NAT does it for talking to the outside world, and we still have plenty unallocated public addresses.
Business cases have been made, feasibility plans created, consultations and meetings have been held, and it all points to: IPv4 works just fine, thank you. Our network-related problems have absolutely nothing to do with IPv4, so nobody is going to put his job on the line for the fancyness of a new technology that nobody really needs. OK, maybe somebody needs it, but heck, I really didn't see any such company around.
So, you see, if even the cook doesn't want to eat his own soup, you probably can stick to the tried-and-tested Big Mac (so I like Big Macs. Got a problem with that?) too.
Sigged!
The reason we are running out of IPv4 addresses isn't technical, its a poor implementation of routing by Cisco that is mostly to blame combined with a political mess.
/19 minimal allocation is a complete waste of resources.
/24, that means for 15 interfaces I need 8 meg of memory to decide where the packet goes for mostly static tables.
99+% of the net today can look at the reset of the world as a "default route". That means for most of the world, the
A very small number of companies fit in the dual homed category. While they may need better routing, most of the time its not for efficiency of the routes but redundancy. Note that is is virtually impossible for a small business to be dual homed and have things work when one of the links goes down.
The remaining is the core routers. A core router shouldn't be using routing tables they way they are done now. For most routes in a core routers, its just a switch. Stuff to 1.2.3/24 goes to interface 2 and that's it. There tend to be a few dynamic routes for some of the stuff that's close but everything else is far away and very static (relative to the routers ability to change all of it). Since no one is switching far away traffic in smaller groups than a
Ipv6 isn't going to fix any of this. It doubles the amount of bits that are needed for the hardware routing and then double that for the local address. That doesn't seem like a good idea to me.
I would like to play with IPv6 on a public network but Racksapce (where I keep a server) won't give me an IPv6 address.
Cisco is still earning about $3B/yr on about $18B in revenue.
The money you buried is now soggy... isn't it time you switched to IPv6 (which Cisco mostly only gives lip service to anyway)?
You can still do NAT like stuff with IPv6, you just don't have to. Wouldn't it be nice to be able to swap a switch or router or just a patch cable and have all your open internet connections still work instead of being dropped?
Why only a move from 32bit to 128bit addresses? I mean, I know there are a hell of a lot of assignable addresses through IPv6, but wouldn't it have made more sense (and be more futureproof) to just have an address that can be dynamic in length?
It is probably not regarded as a pressing issue to increase the range of addresses above 128bits, but then 32bits (and 640K RAM cough) seemed a lot at the time. As has been stated in previous comments, this addresses will eventually be consumed by even the most trivial of objects like light switches or microwave oven bells.
A similar point could be made for dates, where fixes for the year 2000 suddenly allowed dates up to 9999, but what about when we hit the year 10000? Sounds silly, yes, and no doubt we will have moved on to much bigger and better things by then... but what if, for example, we suddenly (within years) moved to a new style calendar system where we started counting from 18209 years ago? Yeah, the point for dates is probably stupid, but why not just let the date/address be any length it needs to be?
Just start with the lowest bit and then work towards the most significant bit that will uniquely identify an object? Perhaps this is unworkable, but it seems to make more sense than just relying on no one filling out the address space (again... will we never learn?) It also seems to follow logically from how the domain name system works where there is a hierarchy involved from some top level towards the actual machine address. I imagine I am missing some vital concept of addresses needing to be a fixed number of bits or something though, I haven't delved into it enough to understand exactly the issues involved.
... I guess
Just look at it this way, the content provider sends a single stream, not 1000, just one. That stream is sent to the first router then stops. If someone on the other side of the router reqests the stream, the router sends it through, still a single stream. But, if another person on the same segment as the original requester want's the same stream, they just "listen" on that that particular muticast address. You still only have one stream, router to router, ever. With proper multicast you should never have multiple streams per route.
So, in this situation, the content provider is using very little bandwidth at all, much better than normal unicast. This should "save" the content provider big bucks on ISP bandwidth. And so for the end point user, since they only need one stream that 1000 people can listen in on, it's still one stream. It should save them big bucks too.
Multicast...allowing multiple clients to "listen-in" on the same IP address data stream. It's just like publicly acceptable snooping!
Do you hear me now?
5 pints.
You people in the nice non-gov't world can fret about luxuries like IPv6 and IPSEC while I have to battle in 2003 to even get rid of telnet. Yes, in this day and age we are still running telnet. At this rate if the world adopted IPv6 tomorrow I would get to implement it a few years after I can teleport to work.
You're on to something here, but the problems are not just content (receiver) control and charging. There are lots of technical problems as well, primarily scalability. During the last 10 years, several new protocols for use in IP multicast have appeared, trying to make up for some of these problems, but it's not really ready for widespread use yet.
The main problem is the amount of per-group state routers need to keep. Currently, a router needs to know about all multicast groups that are being sent through it, and where they should be forwarded to. If multicasting becomes more widespread, the amount of multicast groups will soar, and thereby the multicast state will exceed what current routers can handle.
Correct me if I'm wrong in something here... The scalability problem is partly taken care of with Rendez Vous points, but I'm not entirely sure about those. Basically, they are local control points for routers on a given network, so that some state is lifted from the routers.
All of this conflicts with the rule of always pulling functionality as far up the network layers as possible.
The point of all this is that IP multicast is not really fully researched yet, and I hope IPv6 deployment is delayed until the multicast problem has been properly addressed, so we can get the full functionality in one packet.
I'd say at least 10 years.
what the hell are you babbling about? i'm a router guy and you're not making any sense to me. you would get laughed off of nanog if you ever posted gibberish like this.
Local IPv6 addresses don't offer any advantages over 10.* IPv4 addresses.
Global IPv6 addresses don't work. Most client computers around the Internet can't talk to a server on a global IPv6 address, and most server computers around the Internet can't talk to a client on a global IPv6 address. Sure, a few people could connect to my IPv6 addresses; so what? Why should I go to extra effort to make those addresses work?
All the operating systems I use have been claiming ``IPv6 support'' for years. But they still require manual action by the system administrator before they can talk to IPv6 addresses. What do I gain by spending time setting up IPv6?
(All of this boils down to a small protocol design error in IPv6. A small change to IPv6 software would make IPv6 addresses work without any administrator action. I have a web page, http://cr.yp.to/djbdns/ipv6mess.html, explaining this in much more detail.)
In today's business climate, we can't imagine migrating without a financial incentive to do so.
Well this company refuses to spend out any money to investigate ipv6. Yes there is an IP shortage. And do you know what causes it? Primarily IANA who are holding about 1/3rd of the total IPV4 address space in reserve.
dont believe me? check this.
I took a look into v6 and found some nice things in there. But I wouldn't want to use it currently. Having PA-Addresses on every networkstation in our internal network and not having to use NAT sounds like a good Idea. Until you change your provider. Big renumbering ante Portas. And the autoassignments might be nice, only I do not trust them yet. There will always be boxes configured by Hand.
... All that little things, already supporting IP. They only can do v4.
And then there is the vendor support. Not only the software missing (that was mentioned before), but also the missing Hardware: I'm talking about network printers, Barcodescanners, telephonesystems, powerswitches
So we cannot switch until all the devices we need support it. And that won't be the case within the next 5 to 10 years I guess.
Nils
I am already there. My Dutch ISP supports IPv6, my Cisco routers support IPv6 (as of IOS v12.2), my Windows XP machines support IPv6, my Linux boxes support IPv6, Windows Server 2003 supports IPv6, and we are rolling that out right now, and of course both Bind and Microsoft DNS support AAAA records, so there is no need to wait. :)
On the other hand, learning the new numbering scheme is quite a pain...