GLIX/Diagnostics/Looking glass and specifications
Diagnostics and standards
Looking glass and technical specifications
Dual-stack IPv4 and IPv6, standard BGP communities, RPKI validation, and the layer 2 rules that keep a shared switching fabric stable.
Looking glass
The public looking glass is not online yet. This page intentionally carries no query form: a query box on a static site would have to post somewhere, and there is nothing to post to. When the route servers are live, link the hosted looking glass here and the sample output below can be removed.
Once running, the looking glass will accept show route,
show route detail, ping, and BGP summary queries against both
route reflectors. Sample output is shown below purely to document the expected format.
glix-rs1> show route all 203.0.113.0/24
203.0.113.0/24 unicast [EXAMPLE_PEER 00:00:00.000] * (100)
Type: BGP univ
BGP.origin: IGP
BGP.as_path: 64500
BGP.next_hop: 198.51.100.10
BGP.local_pref: 100
BGP.community: (64500,1000)
RPKI ROA State: VALID
IRR Validation: MATCH
Addresses, AS numbers, and communities above are documentation
examples from RFC 5737 and RFC 5398, not live exchange data.
Maximum transmission unit
The core switching fabric carries generous frame buffers, and a consistent MTU across every participant is what keeps them useful.
- Standard MTU
- 1500 bytes
- Jumbo MTU
- 9000 bytes
- Default at turn-up
- 1500 bytes
Jumbo frames are available on request at no surcharge. Members negotiate a matching MTU with their peers; mismatches are the single most common cause of an exchange that appears to work until a large packet crosses it.
Allowed EtherTypes
| EtherType | Protocol | Status |
|---|---|---|
| 0x0800 | Internet Protocol version 4 | Permitted |
| 0x86DD | Internet Protocol version 6 | Permitted |
| 0x0806 | Address Resolution Protocol | Rate limited |
| Others | Any other EtherType | Dropped |
Peering ports operate in untagged 802.1Q access mode by default. Tagged sub-interfaces are provided for dual-service participants.
Layer 2 broadcast domain rules
- No spanning tree. STP, RSTP, and MSTP bridge protocol data units are dropped instantly at the ingress edge port.
- No proxy ARP. Member router interfaces must disable proxy ARP responses.
- No discovery protocols. LLDP, CDP, and neighbor discovery toward the switching fabric are prohibited.
- Unicast BGP only. All peering traffic must be unicast IPv4 or IPv6.
- One MAC per port. Exactly one MAC address is permitted on each allocated port.
- Storm policing. Broadcast, unknown unicast, and multicast traffic is policed per interface.
Peering policy and neutrality
Open peering with the route servers
GLIX maintains a strictly open policy across its public route server infrastructure. Every connected ASN that peers with the route servers receives route advertisements from all participating networks that have chosen to announce prefixes to the multilateral switching fabric.
- No minimum traffic ratio or throughput quota.
- No non-disclosure agreement required to peer.
Bilateral and private peering
While the route servers expedite day-one connectivity, direct bilateral sessions between any two consenting participants are expressly permitted over the shared peering VLAN without surcharge. Members may also negotiate private network interconnects or dedicated metro VLANs.
- Direct member sessions run concurrently with route server sessions.
- Private interconnect VLANs are available on request.
Need something the documentation does not cover?
Link aggregation, jumbo frame negotiation, high-speed handoffs, and private VLANs are all discussed directly with the engineers running the fabric.