03.07.2008
Roque Gagliano, LACNIC.
In the last “FLIP-6” event an increasingly important topic was introduced: how to measure IPv6 deployment? This question is particularly relevant if we would like to compare the relative deployment in our region against other regions or the impact of a promotion campaign all over the region or in a particular country.
There are three families of metrics that we can think of: those from the data plane, those from the control plane and those from administrative data.
The first family (from the data plane) deals with actual IPv6 packets and help us understand the amount of IPv6 traffic in the network. However, due to the extension and localization of the network and its services, we can only get localized and non-objective (as it depends on where and what we are measuring) data.
When we talk about data from the control plane, normally we refer to the information available in the inter-domain BGP table. From the BGP table we can measure how many IPv6 prefixes are announced and so, probably, in use. We can also study how many ASNs are present in the IPv6 table and compare them with the ones available in the IPv4 table. It is predictable that the amount of IPv6 prefixes and ASNs in the IPv6 BGP table will increase in the years to come, so it is important to track those changes. From the BGP table it is not possible to establish the amount of internal IPv6 deployment. However, we can also see that if certain ASN provides transit to other ASNs, it has certain amount of IPv6 deployment. Also, we can check for AAAA or PTR DNS records for certain particularly interesting sites.
Finally, the metrics from administrative data refers to information about allocations and assignments done by RIRs. This information is relevant as it is a prerequisite for any IPv6 deployment, however it does not tell us if those addresses are in use.
There are several initiatives that are taking measurements from all the three areas that we mentioned and that can be found in the references [GERT] [HE] [FE] [APNIC].
At LACNIC, our interest is to measure the deployment of IPv6 in the region and to understand the effect of the future “IPv6 Tour” that will take place in the coming 12 mounts.
For this reasin, we have implemented a section named “Statistics” in the IPv6 Transition Portal: http://portalipv6.lacnic.net. In this section we will add statistics about IPv6 and the depletion of IPv4.
The IPv6 statistics are divided in 4 groups:
- Global Statistics, which refer to data from all RIRs.
- Regional Statistics, which refer to data specific for the LACNIC region. In this case we consider more metrics as they help us to better understand the region as a all.
- Per Country Statistics, which are the same data as in the previous group, but specific for certain country.
- LACNIC´s Services Statistics, which refer to access information to LACNIC services.
We would like to mention certain measurements in particular.
First, we have a plot that show us the total number of allocations and assignments done by LACNIC and how many of these are routed (see Figure 1).
Figure 1: Total number of allocations and assignments in the LACNIC region (blue) and total number of routes form those allocations and assignments (green).
As we expected, the number of allocations and assignments increases over time and in particular we can see that the growth rate has increase during this year.
One important question is if these allocations and assignments are of minimum allocation size (/32) or they include important allocations to big organizations. The following plot (Figure 2) shows the total number of addresses allocated and assign measured in units of /32s.
Figure 2: Total amount of addresses allocated and assigned, measured in /32 units (blue). Total amount of addresses routed measured in /32 units (green).
We can see that this plot has several steps, they reflect allocations of more that one /32 in the region. We expect that the number of steps and its high will increase over time.
There is a difference in the two previous plots of the amount of prefixes allocated or assigned by LACNIC and the number of routes seen. With the current IPv6 policy, obtaining IPv6 resources is an administrative task. However, announcing prefixes at the inter-domain routing table requires equipment support and configuration. One important question would be what happens with the mean time that it takes from getting those addresses and announcing them as a unique aggregate. The following plot (Figure 3) shows this data but only for those prefixes that were announced at the time of the measurement.
Figure 3: Mean time from allocation or assignment to announcement for IPv6 prefixes in the LACNIC region.
We can appreciate how this measurement has increase over time. At the beginning, most of the allocations and assignments were done to organizations that were already connected to the 6Bone, as their infrastructure was already IPv6 capable, the announcement was done fairly quickly. As new providers with no previous experience were getting IPv6 addresses, the time to announce them increased. We can expect this measurement to increase even more in the short term, as there are several organizations with IPv6 allocations and assignments for a long time. As the number of providers giving IPv6 connectivity increases and there is more availability of IPv6 training, the measurement should decrease importantly.
The previous data gave us important information of how much has been done, but it will be very useful to see that information in perspective of how much needs to be done.
What we can do from the inter-domain routing table is to count all ASNs allocated from LACNIC that are only present in the IPv4 table, those that are only present in the IPv6 table, those that are in both and those that are in none of them. Figure 4 shows this information.
Figure 4: Autonomous System Numbers that are only present at the IPv4 inter-domain routing table (blue) , the IPv6 inter-domain routing table (green), both (red) or none (pink).
The goal could be for each ASN at the IPv4 routing table to also be present at the IPv6 routing table, which means that we still have some 900 organizations in the LACNIC region that requires beginning announcing IPv6 prefixes. Please note that no organization is announcing only IPv6 prefixes. This can mean that the organizations running IPv6 are using the same infrastructure as in IPv4.
The same plots that we have shown here can be found at the “Per Country” section of the statistics webpage of the Portal, giving the information for each country.
The last section of statistics refers to LACNIC services statistics. As we mentioned, the measurements done from the information in the data plane is no-objective as it depends on which is the particular public for the services we are measuring. If we take measurements from the LACNIC services, we are indeed looking at the use of IPv6 in our own community. Particularly, if we measure the amount of successful login to our registration services in IPv6, we will be measuring IPv6 deployment in our members’ networks. We consider the information as a grand-total and also per unique address in a weekly base.
We are also taking access information for our web servers and our joint-whois service. In this last case, we are also checking which is the content inside the whois request, measuring if the request is for IPv4, IPv6, ASNs or other.
References:
