Title: Operational Experience with IPv6
1Operational Experience with IPv6
- Bob FinkLBNL/ESnet
-
- NANOG19
- Albuquerque, NM
- 11-13 June 2000
2This session
- Brief talks on our IPv6 experiences
- Bob Fink 6bone, Esnet, 6tap
- Greg Miller MCI WorldCom, vBNS
- Bill Maton CRC, Canarie
- Rob Rockell Sprint
- Sean Mentzer Qwest
- Panel QA session
- Bob Fink moderator
3The 6bone
- First IPv6 packets in mid-1996 between a few
sites in Europe, Japan and US - started with tunnels (v6 encapsulated in v4),
now moving to more native links - primary goal to test implementations, their
interoperability, that the standards work and
provide feedback to the IETF - and to get early operational experience
46bone today (as of 5Jun00)
- Now in 46 countries AR, AU, AT, BE, BR, BG, CM,
CA, CN, CZ, DK, EE, FI, FR, DE, GR, HK, HU, IN,
IE, IT, JP, KZ, KR, LT, MY, MX, NL, NZ, NO, PL,
PT, RO, RU, SG, SK, SI, ZA, ES, SE, CH, TW, UA,
GB, US, UY - 571 networks/sites 135 US, 66 DE, 38 JP, 28 FR,
28 GB(UK), 20 SE, 10 CN, 9 RU, 4 MX etc. - 68 pTLAs most recent addition UUNET 27Apr00
5Primary lessons
- Besides implementations and standards issues that
come up and are corrected... - overall IPv6 does act and work like IPv4
- Though we started with static routes, IDRPv6 and
RIPng, everyone wanted BGP4 and we moved to it
fast - It has not been hard to setup, manage, maintain
and operate IPv6 nets (again, it is like IPv4
funny thing -) Ill leave it to other panel
members for more
66bone (IPv6) Registry
- From the start of the 6bone project a registry
was used to keep track of at least the top tier
networks, their peerings and prefix allocations - RIPE-style db developed by David Kessens then of
ISI, now of Nokia - He added ipv6-site and inet6num objects
- Has proven invaluable for net reports,
measurement, management and peering
7Network Tools for tracking problems
- Merit IPv6 Routing report (daily)
- Size of 6Bone Routing Table
- BGP4 Traffic Summary
- Unknown AS Numbers
- Unknown Prefixes
- Poorly Aggregated Announcements
- Prefixes from Different Origin AS
- Most Active Prefixes
8Contd.
- CSELT (Italy) BGP4 Operational report using
Aspath-tree tool - Graphic display of BGP4 routing entries
- Odd routes reports
- invalid unaggregated prefixes
- Routing Stability Report
- Routing History graphs
9Example CSELT routing history graph
10Contd.
- SLAC PingER services uses a modified IPv6 ping
service to probe and report on path reliability - data is databased and historically accessibel
hour, day, month - TCP Throughput
- Zero packet loss frequency
- PING unpredictability unreachability
- Packet loss
- PING response history
11Automated mapping services
- Among other things the 6bone registry is used
for, it can help generates pictures of the 6bone
pTLA backbone network peering relationships - Lancaster Univ. have done much of this work
12 136bone future
- Will stay in place until no longer needed
- excellent place for an ISP to get early
experience before going into production - probably not the best way to support transition
to IPv6, but if needed for this it will stay in
place - expect that the 6to4 Transition Mechanism and
native IPv6 support by your IPv4 ISP is best to
support transition
14Aggregatable Unicast Addressing
3
13
24
16
64
8
R s r v
TLA
NLA
SLA
Interface ID
001
48-bit Public Topology Routing Prefix
80-bit end-site specific usage ISP cannot change
this
TLA Top-Level Aggregation ID - are assigned
to ISPs and Exchanges that act in a default-free
way with a routing table entry for every active
TLA ID (helps constrain the routing
complexity) Rsrv Reserved for either TLA or NLA
expansion NLA Next Level Aggregation ID - are
assigned by TLAs to create a multi-level
hierarchy underneath it as the ISP chooses
(i.e., multiple NLA levels allow more ISPs and
then the end site) SLA Site Level Aggregation
IDs are used to create local addressing hierarchy
(e.g., a flat subnet space allowing 65K
subnets) Interface ID unique ID on subnet
(typically formed automatically)
15Current TLA assignments
- TESTING (3FFE/16) assigned for IETF ngtrans
use by RFC 2471 for use by 6bone project -
currently used for pTLAs - Sub-TLAs (2001/16) assigned to RIRs for
allocation of Sub-TLAs - 6to4 prefix (2002/16) assigned for use by the
Connection of IPv6 Domains via IPv4 Clouds
without Explicit Tunnels spec (soon to be at PS)
to hold the IPv4 tunnel endpoint address in the
32-bit Rsrv NLA fields
166bone (IPv6 testbed) pTLAs
3
13
24 or 20
16
64
8 or 12
TLA 0x1FFE
pseudo TLA
NLA
SLA
Interface ID
001
- The 6bone uses a variation of this concept called
pseudo-TLAs (pTLAs)3FFE0000/24 to
3FFE7F00/24 old 8-bit pTLA space3FFE8000/2
8 to 3FFEFFF0/28 new 12-bit pTLA space
17Sub-TLAs
3
13
13
16
64
13
6
TLA 0xOOO1
Sub TLA
R s r v
NLA
SLA
Interface ID
001
- To assist in the slow start of TLA assignment, a
Sub-TLA was defined which allows the
international address registries to slow start
TLA usage by just assigning a single TLA for
Sub-TLAs - an ISP must demonstrate high usage of its Sub-TLA
space before qualifying for a TLA or another
Sub-TLA - in practice, the RIRs are slow starting the /29
space by only assigning /35s to start again, a
high usage required before getting more of the
/29
18Sub-TLA usage today
- RIRs started to assign in July 1999,34 assigned
to date - APNIC (13 Sub-TLAs assigned)
- ARIN (4 Sub-TLAs assigned)
- RIPE-NCC (17 Sub-TLAs assigned)
196to4
- Specifies the 16-bit TLA prefix 2002/16as a
6to4 flag indicating that the 32-bit sized NLA
below it carries an IPv4 Tunnel Endpoint Address
of the sites egress router
3
13
16
64
32
TLA 0x002
IPv4 TEA
SLA
Interface ID
001
20ESnet
- ESnet serves the network needs of the US Energy
Research national labs, which is now IPv4 (just
turned off DecNET -) - early participant in 6bone using tunnels, then
moved to native IPv6 in 1999 - operates a Cisco IPv6 EFT router mesh over the
ESnet ATM cloud - as of July 1999 ESnet operational staff handles
IPv6 peering and routing in parallel with, and
the same as IPv4
21Application usage
- For now usage limited to early application
conversion (to the IPv6 API) and demonstration
that high-profile scientific apps run over IPv6
the same as IPv4 - also a significant network measurement activity
in place - software (versus hardware) IPv6 packet forwarding
ok for most early purposes, but it does get in
way of the high-speed scientific apps were
waiting too!
22The 6TAP
- To facilitate peering of native IPv6 providers,
ESnet and Canarie/Viagenie formed the 6tap IPv6
routing service in August 1999 at the
StarTAP/ChicagoNAP - Working with Sun and Merit to get early an early
IPv6 Route Server up - Working with early IPv6 ISPs to establish BGP4
routing policies and practices - A 6to4 Relay service and a Site-Tunnel-Server
service will also be provided soon
23What next
- Waiting for production hardware-based routers so
we can operate v4 v6 in same routers - and of course for production IPv6 host code to
become widely available - meanwhile, it is quite cheap and easy to put up
some IPv6 routing to gain knowledge and early
operational experience
24Thanks for listening
- Pointer to everything IPv6
- www.6bone.net
- Questions on anything IPv6 (dont worry, Ill
forward you to the right place) - fink_at_es.net