Re: [sig-policy] prop-098 Optimizing IPv6 allocation strategies(simplifi
Sent from my iPad
On Aug 24, 2011, at 23:56, "Terence Zhang YH" <zhangyinghao at cnnic dot cn> wrote:
> (NO link to co-chair's opinion)
>
> Hi Owen,
>
> We have intensive interest in prop-098, generally speaking, we support the idea of
> more generous IPv6 allocations policy, and nibble boundary allocation etc.
>
> We have a few questions about the proposal:
>
> 1. About Provider Allocation Unit:
> It is defined as the smallest reassignment unit used by the provider.
>
> Q1: Is the size of the PAU defined by the provider? Do they need to justify the size?
>
Yes. By choosing the smallest assignment unit they give to a customer, they set the size of their PAU. So, if they give a /48 to all customers, then, their PAU is /48. If they give /60s to some customers, then their PAU is /60, etc.
Any PAU up to /48 may be chosen by the provider without additional justification. (Or at least that is my intent in writing the policy).
> 2. About End site:
> It is defined as a single structure or service delivery address
>
> Q2: When it comes to mobile IP services, Is the 'end site' applied to
> a PDA or Cell phone devices, a personal area network (PAN), an Access Point,
> or a POP?
>
I believe end-site in that case would apply to the device where the carrier's responsibility ends, for example, the cell phone, PDA, or USB-modem. All devices served by a single mobile hot-spot, for example, would be considered a single end-site.
I confess I didn't thoroughly think through mobile when writing the policy, but, I believe that the above represents the most logical application of the policy as written to mobile and that would certainly be the intent if I were to write clarifying language to address the question.
> 3. Section 4.2: ......5 years of projected customer
> utilization based on assigning each customer end-site one
> provider allocation unit without exceeding a 75% utilization.
>
> Q3: Does this assume each customer end-site will be allocated one same size PAU?
>
It does not assume that the ISP necessarily will do so, but, it does measure anticipated utilization based on the assumption that the ISP will do so. An ISP will have a single PAU size which will be used to measure all of their utilization. It will be their smallest assignment unit. So, an ISP which gives /60s to some customers and /48s to others will have to count those /48s as multiple /60s. They will have to justify the number of /60s they give to sites that receive /48s. If they give /48s to all end-sites, then, they will not need to justify anything until an end-site receives a second /48.
Does that answer your question?
> 4. Section 5.1 subsequent allocation criteria:
>
> - 75% or more utilization of their total address space, OR
> - One or more facilities which have reached a 90% utilization......
> there are no available blocks of sufficient size in the providers current
> allocation(s) to expand those facilities.
>
> Q4: How to justify 'there are no available blocks of sufficient size', specifically
> when the LIR has sparsely assigned 160 * /40 from a /32, now one of the /40 site becomes full,
> and no more contiguous /40 left, will this be considered meet the criteria.
>
I believe that the allocation of a second non-contiguous /40 to that serving site would be expected in that case. If you have utilized 75% of your /40s (192 /40s assigned) and you fill 90% of one of them, then I believe you would qualify for an additional block.
I agree this is an area where clarification could be useful in the policy. Perhaps we can gather community feedback on the desired outcome in Busan and incorporate that into a clarification for last call? I confess I am not 100% well versed in the APNIC policy development process.
> 5. Section 2.4. The HD ratio ......Using nibble-boundaries and rounding up actually yields similar
> results with simpler math.
>
> Q5: How to understand 'nibble-boundaries round up' will have the similiar results with HD?
> With a consistent 75% usage and unpredictable 1-8 times rounding up, my humble feeling is
> the utilization is unpredictable, but the HD-ratio requirement is predictable.
>
> Prefix HD require Utlization:
> ----------------------------
> /32 36.9%
> /28 31.2%
> /24 26.4%
> /20 22.4%
>
> In my premature estimation, the 75% usage plus round up maybe more relax than HD requirement
> in small network, but it maybe more restricted than HD in large network.
>
Yes, it can be (slightly) more restrictive in a large network. It is (significantly) more liberal in some small network corner cases.
> Q6: Does this proposal remove HD-ratio criteria completely?
>
Yes.
Owen