Article on Slashdot, quick comments
As seen on the worlds greatest 'Stupidity Amplifier'; slashdot: http://slashdot.org/askslashdot/01/11/08/0237233.shtml ikekrull asks: "After looking to see how I could set up my company's LAN to be multi-homed? , I found that it would be next-to-impossible for me to do this. 'Providerless' IP addresses are no longer allocated to anybody in this part of the world (New Zealand) by APNIC? , unless you meet requirements (financial and political) that are pretty much unmeetable by anyone but a large ISP. Does this put control of the entire internet further and further into the hands of large corporate players, and and is anyone particularly interested in changing this situation?" [...] Since this was posted by someone in New Zealand is comments on APNIC directly (of which there are employees on this list) I should point out: It is my belief, based on experience in doing just this and from dealing with APNIC for several years, that being a small ISP or network user does NOT necessarily mean you are 'shut out' in any shape or form whatsoever; you can in-fact have one or both carriers take your case(s) for multi-homing to APNIC and potentially have them allocate a /24 (or whatever is appropriate) --- this address space would not be part of the carriers CIDR block and need not give the carrier leverage over your company or network as the address space can be delegate to the end-user not the carrier. I have done this before (after discussing this with people from APNIC and conforming that this was indeed a legitimate situation), is there any reason to believe that this situation has changed radically? --cw P.S. Thinking an email address like blortnospam(a)foo.com obscures your email address is bogus and just plain annoying; please don't do it people, many email address acquiring 'bots know of such patterns and will correct for them anyhow. - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Sat, Dec 01, 2001 at 07:32:15AM +1300, Chris Wedgwood wrote:
It is my belief, based on experience in doing just this and from dealing with APNIC for several years, that being a small ISP or network user does NOT necessarily mean you are 'shut out' in any shape or form whatsoever; you can in-fact have one or both carriers take your case(s) for multi-homing to APNIC and potentially have them allocate a /24 (or whatever is appropriate) --- this address space would not be part of the carriers CIDR block and need not give the carrier leverage over your company or network as the address space can be delegate to the end-user not the carrier.
APNIC almost have a policy on micro-allocations for the purpose of multi-homing: http://www.apnic.net/meetings/12/docs/proposal-multihome-assign.html http://www.apnic.net/meetings/12/results/index.html#2 This was due to become finalised in December, last I heard. No need to have carriers "take your case(s)" to APNIC. It's also commonplace for ISPs in NZ (and almost everywhere else in the world, as far as I can tell) to punch holes in each others' PI netblocks to allow PA-delegated blocks to be used for multi-homing. This is sufficiently recognised as common practice that it will be documented in the current-practices document in the ietf multi6 wg.
I have done this before (after discussing this with people from APNIC and conforming that this was indeed a legitimate situation), is there any reason to believe that this situation has changed radically?
Not that I know of. There are barriers in place to discourage too many people from multihoming. These are important, since the current routing system is not scaling well, and a rush of multihoming at the edge (using routing-based mechanisms) will make it blow up. Joe - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
At 12/1/2001 05:44 AM, Joe Abley wrote:
APNIC almost have a policy on micro-allocations for the purpose of multi-homing:
http://www.apnic.net/meetings/12/docs/proposal-multihome-assign.html http://www.apnic.net/meetings/12/results/index.html#2
This was due to become finalised in December, last I heard. No need to have carriers "take your case(s)" to APNIC.
I would not be so optimistic about timing for this policy. Its not a policy that has support across all three Regional Internet Registries, as far as I am aware.
It's also commonplace for ISPs in NZ (and almost everywhere else in the world, as far as I can tell) to punch holes in each others' PI netblocks to allow PA-delegated blocks to be used for multi-homing. This is sufficiently recognised as common practice that it will be documented in the current-practices document in the ietf multi6 wg.
Getting back to the issue of problem identification and priorities. IF the growth and dynamic stability of the inter-domain routing space using the current capabilities of BGP4 are your primary concern, then this practice is one that causes concern. If, of course, one looks at this from the perspective of ensuring a strongly competitive IP transit market then number portability is also a deep concern. Unfortunately multi6 has, to date, just not been able to provide a framework that can reconcile these two perspectives. i.e. documenting current practices is not an implied value judgement on those practices. There are barriers in place to discourage too many people from
multihoming. These are important, since the current routing system is not scaling well, and a rush of multihoming at the edge (using routing-based mechanisms) will make it blow up.
Not quite - it will take the deployed base of the Internet into territory which is largely uncharted, and in such a space blowing up is a possible, but not entirely certain outcome. The concern is that we'd rather NOT learn by experience in this case. thanks, Geoff - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
Chris Wedgwood wrote:
As seen on the worlds greatest 'Stupidity Amplifier'; slashdot:
(...which a significant number of people read on a daily basis, of which just a small handful use as their morning intelligence test to start the day or pass the time... new age crossword puzzle?). In any case, the problem isn't the punters (that background radiation exists *everywhere*, it is the folks that edit/moderate that particular insane asylum What strikes me as slightly odd is that most people who say they want to multihome for "fault tolerance/redundancy purposes" will reveal to you, under intense questioning, that what they *actually* want to do is load balancing. Whatever the case solutions exist without having to pull your hair out dealing with your upstream/NIC: http://www.radware.com/content/products/link.htm Which only leaves the following question: Does anyone have a definitive (at this particular point in time) list of all protocols/applications that embed layer2/3 address information in the payload? Other alternatives include closely coupling DHCPS/RADIUS/etc. to NAT, but I don't believe anyone has actually written code for that yet (for good and proper reasons). - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Sun, Dec 02, 2001 at 02:02:38PM -0800, cfb wrote:
What strikes me as slightly odd is that most people who say they want to multihome for "fault tolerance/redundancy purposes" will reveal to you, under intense questioning, that what they *actually* want to do is load balancing.
Is load balancing really ever the ultimate objective, or is it merely a requirement of the multihoming architecture?
Whatever the case solutions exist without having to pull your hair out dealing with your upstream/NIC:
You still get to pull your hair out in other ways.
Which only leaves the following question:
Does anyone have a definitive (at this particular point in time) list of all protocols/applications that embed layer2/3 address information in the payload?
Exactly. Joe - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Sun, 2 Dec 2001, Joe Abley wrote:
On Sun, Dec 02, 2001 at 02:02:38PM -0800, cfb wrote:
What strikes me as slightly odd is that most people who say they want to multihome for "fault tolerance/redundancy purposes" will reveal to you, under intense questioning, that what they *actually* want to do is load balancing.
Is load balancing really ever the ultimate objective, or is it merely a requirement of the multihoming architecture?
A small ISP has 2*T1s with an upstream, During peak time they use around 2.5Mb/s . The contract for one of these runs out and the boss finds another vendor who offers them a T1 for half the price of the original upstream. They are now multihomed and their primary objective is to ensure their 2.5Mb/s of peak bandwandith is blanced between the two T1's to ensure good customer service. They do not see fault tolerance as an important objective since they have good service from their current provider and their new provider has a good reputation as well. The boss is just happy cause he saided a load of cash and everything works okay with two providers. -- Simon Lyall. | Newsmaster | Work: simon.lyall(a)ihug.co.nz Senior Network/System Admin | Postmaster | Home: simon(a)darkmere.gen.nz ihug, Auckland, NZ | Asst Doorman | Web: http://www.darkmere.gen.nz - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
What strikes me as slightly odd is that most people who say they want to multihome for "fault tolerance/redundancy purposes" will reveal to you, under intense questioning, that what they *actually* want to do is load balancing.
Is load balancing really ever the ultimate objective, or is it merely a requirement of the multihoming architecture?
Our primary objective with a change to multihoming was to force our existing carrier to meet the price for International bandwidth being offered by another. Asking them to bring the price down did not work. You have to be multihomed and be sending service orders to reduce bandwidth to get their attention. Having said that, we have enjoyed the fault tolerance benefit twice since making the transition. Peter Mott Chief Enthusiast 2DAY INTERNET LIMITED It's kind of fun to do the impossible - Walt Disney -/- - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
I regularly have to disucuss multihoming with enterprises in my role. Generally they are all after HA for their precious and valuable web-servers, and would like to connect to two or often three upstreams for resiliency. Then comes the supplementary problems of how do I balance the traffic ? Going to APNIC usually proves fruitless for these guys, then it's down to the SP's who are often less than co-operative. I agree Service Providers often had different motives for multihoming, being price with a side-effect of resiliency, and even at Waikato we'd buy off more than one provider and shift traffic around depending on the "special of the week". Personally, I think both can be false economies, causing less efficient bandwidth utilisation, and less reliable links for many users, trying to iron out routing problems, resiliency decisions, global filtering and dampening of long prefixes.
From the New Zealand enterprise perspective I still think a connection to one primary ISP using their Super-nets for aggregration, following Best Practices and designing deployments to support renumbering is the best place to start. It is then often enough to have a second link to a well connected ISP for NZ domestic, or even direct APE/WIX connectivity as many NZ sites don't need huge global 24/7 availability.
Just my opinion, Arron -----Original Message----- From: owner-nznog(a)list.waikato.ac.nz [mailto:owner-nznog(a)list.waikato.ac.nz]On Behalf Of Peter Mott Sent: Monday, 3 December 2001 10:07 AM To: Simon Lyall; nznog(a)list.waikato.ac.nz Subject: RE: Article on Slashdot, quick comments
What strikes me as slightly odd is that most people who say they want to multihome for "fault tolerance/redundancy purposes" will reveal to you, under intense questioning, that what they *actually* want to do is load balancing.
Is load balancing really ever the ultimate objective, or is it merely a requirement of the multihoming architecture?
Our primary objective with a change to multihoming was to force our existing carrier to meet the price for International bandwidth being offered by another. Asking them to bring the price down did not work. You have to be multihomed and be sending service orders to reduce bandwidth to get their attention. Having said that, we have enjoyed the fault tolerance benefit twice since making the transition. Peter Mott Chief Enthusiast 2DAY INTERNET LIMITED It's kind of fun to do the impossible - Walt Disney -/- - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
Arron Scott wrote:
Personally, I think both can be false economies, causing less efficient bandwidth utilisation, and less reliable links for many users, trying to iron out routing problems, resiliency decisions, global filtering and dampening of long prefixes.
Less efficient bandwidth utilisation is a likely outcome, however it is possible to obtain some efficiencies by moving less time sensitive traffic to a cheaper, crappier, and/or more heavily loaded links. ISPs can do this in a multi-homed environment to a reasonable degree without upstream support. On the issue of reliability, there is almost always increased complexity at larger, upstream providers. The result is that given two organisations with clue, the larger, upstream provider will always be less reliable. This is one reason ISPs need to multi-home, but others (like having more than one telco, cable into the building, etc.) are also important. -Craig - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
I read in a Nortel Software release notes that they now to Aggregate (ci$co etherchannel) bonding between 8000 series passport chassis, in otherwords Load balancing(sharing) and pretty well instantaneous recovery in the event that 1 should fail. The only requirement is that Multilink Trunking (802.1q) is running between the chassis. I have played with this and it is way cool. Russ
On Sun, Dec 02, 2001 at 02:02:38PM -0800, cfb wrote:
What strikes me as slightly odd is that most people who say they want to multihome for "fault tolerance/redundancy purposes" will reveal to you, under intense questioning, that what they *actually* want to do is load balancing.
Is load balancing really ever the ultimate objective, or is it merely a requirement of the multihoming architecture?
Whatever the case solutions exist without having to pull your hair out dealing with your upstream/NIC:
You still get to pull your hair out in other ways.
Which only leaves the following question:
Does anyone have a definitive (at this particular point in time) list of all protocols/applications that embed layer2/3 address information in the payload?
Exactly.
- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
participants (9)
-
Arron Scott -
cfb -
Chris Wedgwood -
Craig Anderson -
Geoff Huston -
Joe Abley -
Peter Mott -
Russell Sharpe -
Simon Lyall