[ Date Index ] [ Thread Index ] [ <= Previous by date / thread ] [ Next by date / thread => ]
On Sun, 22 Jan 2012, tom wrote:
On 22/01/12 17:02, Martijn Grooten wrote:On Sun, Jan 22, 2012 at 4:13 PM, Neil Winchurst wrote:There has been a lot of emails on the list recently about IPv6. I am a little confused about all this, and I suspect that I am not alone. I get the impression that IPv6 is coming in to replace IPv4 but is taking a long time to happen. At the moment for most people there is no point in worrying about it, IPv4 will carry on working for the foreseeable future . Also many ISP's are dragging their heels about this so are not yet ready for it. And, there are not yet many suitable routers on the market. These are the 'facts' as I see them at the moment, but I am ready to be corrected. So can anyone put me right if necessary please? I think others on the list will be interested too.You put it all quite well. IPv4, the 'current' IP protocol, supports 32-bit IP-addresses like 80.68.88.22, where every 'part' has an 8-bit value which can thus range between 0 and 255. This gives just over 4 billion possible IP addresses (for a number of reasons not all of which can be used). With a world population of around 7 billion, many of whom use (or will use) more than one device, that is nowhere near enough.
That's not quite true - I've got one IP address and I can, using IP4 have millions of devices behind it and not suffer any conflicts. I can host millions of web sites on that same IP address. There are nowhere near 4 billion devices connected directly to the internet - its just thereʼs a lot of squatting going on - people with allocations that are not used and are technically tricky to use anywhere else.
How would you handle 100 virtual PBX servers behind one IP address? (Or 100 boxes requiring FTP and shell access) Tell all your clients to use different ports? That's not going to go down well, I can assure you!
So while what you say is mostly true, there are limitations. e.g. virtual servers - I host many physical servers and many many more virtual servers. I need to give the virtual servers their own IP address. I also need to give websites running on the same physical host their own IP address if they are to have an SSL certificate. The majority of IP addresses I have in the data centre go to SSH based sites and virtual servers.
The other limitation is the administrative overhead - somewhere the front-end (e.g. apache or a dedicated appliance) needs to track all that data and keep track of all the incoming requests, match against target web URL, direct the commands to the right bit of server/code and back again - it's a sort of NAT at the layer 3 level, but it still needs resources to manage - so bigger servers/routers with more memory and fast data paths and processors...
Big NAT engines also require resources - money to buy the boxes that will handle 1000's of clients behind them at multi 10/40 Gb speeds.
So there are limitations and it's not always going to be easy.
You could (theoretically) re-allocate IP ranges and the problem would go away - almost certainly for long enough for us not to worry about it again in our lifetimes - but that is politically tricky and wont sell a whole bunch of hardware we dont really need.
There really isn't that much to re-allocate - and as you pointed out, "politically" it just won't happen.
I think we will see issues in out lifetime and I'll be surprised if we don't start to see them soomer rather than later. Certianly by the end of this decade - the Internet won't stop working, but it's going to change.
So there's no point burying your head about it - it costs nothing but some of your time to read up on IPv6 and how it works and what those funny numbers and clons mean, and anyone can get an account with a tunnel broker and tunnel IPv6 from their Linux PC - so what are you waiting for?
Gordon
-- The Mailing List for the Devon & Cornwall LUG http://mailman.dclug.org.uk/listinfo/list FAQ: http://www.dcglug.org.uk/listfaq