110 likes | 237 Views
LHCONE “Point-to-Point Connection Service” Service Definition. Jerry Sobieski. The Service Definition. Specifies technically what the user experiences due to the service Specifies ancillary processes and procedures that will be available and/or followed to support the service
E N D
LHCONE “Point-to-Point Connection Service”Service Definition Jerry Sobieski
The Service Definition • Specifies technically what the user experiences due to the service • Specifies ancillary processes and procedures that will be available and/or followed to support the service • Intended to provide the LHCONE community with service guarantees *and* ability to interoperate with non-LHC infrastructure
Technical description • The LHCONE point to point Connection Service - “P2PCS” - will: • Deliver performance guaranteed point-to-point connections on a global basis for the LHCONE community • Provides a single contiguous service region with a global service reach. • The global service region is the result of numerous autonomous network service domains and open exchange points inter-connecting to provide a single consistent end-to-end P2PCS capability. • The service employs open and standardized protocols to deliver these capabilities.
Connection Instance • P2PCS Connections are Point to Point only (initially) • P2PCS payload is IEEE 802.1Q frames. • The service can deliver frames between ethernet endpoints associated with different VLANs • Payload frames are delivered end to end unmodified, with two explicit exceptions: • VLAN tags may be re-written (which means the FCS field will need be re-computed as well.) • Inter-arrival time of frames associated with a particular Connection is not preserved end to end (i.e. no jitter control is provided.) • P2PCS Connection Endpoints are individually identified by the full topological specifications that uniquely identify the flow to be carried end to end • Ex: <networkid>…<switchid><bladeid><portid><VLANid> • P2PCS connections are not VLANs – they are pipes that carry payload datagrams formatted in a particular fashion. • This allows us to deliver ethernet frames without requiring the intermediate network transport infrastructure to be ethernet equipment.
Reservation Parameters • Origination STP* • Destination STP* • Capacity* – Megabits per second • Max Payload datagram size – MTU • default MTU 9100 bytes • Shaping parameters • Max Burst size • Burst window • Schedule* – start & end time/date • Authorization credentials* • (*) indicates a required user specified parameter • Other parameters may be allowed to default
Scheduling • P2PCS Connections can be scheduled in advance • They may also be immediate (“now”) • They all have a start and end time/date • The service supports the following functions: • Reserve/Cancel • Provision/Release • Query • All such functions are will be individually authorized against credentials provided by the requesting agent • Security and privacy is a primary concern for global services – particularly where resources may be acquired from or by outside agents
Scheduling • Scheduling is sacrosanct. • Once confirmed, a reservation will not be removed or overridden except to insure the health of the system itself. • Failure by the service provider to deliver a confirmed connection – for any reason - constitutes a “critical outage”. • In general… • These connections are not “protected” • It is the provider’s responsibility to correct an outage as soon as possible 24x7x365 • It is user’s option to mitigate an outage (provide an alternative) • We do not allow bursting above reserved constraints (at least initially) • ABR policing requires special buffer management and datagram marking to guaranteethe performance
Eerror Threshold • Connections are imperfect… • Frame loss rate not to exceed 1*10^-8 • If these are jumbo frames, then this is ~ 1*10^-12 BER • If FLR is less than this, the circuit is considered to be operating within acceptable limits • If FLR exceeds this rate, then the connection is considered to be failing • A “critical outage”
Process • Outage procedures • Change Mangement • User support • Accounting • Policy coherency across domains
Global Architecture Other collaborating inter-operable R&E GOLEs Other inter-operable R&E Networks and their available inter-domain connectivity LHCONE specific GOLEs LHCONE Networks (end sites – labs, campuses,…) And their associated LHCONE interconnectivty Result: A high capacity, highly interconnected, dynamically provisioned, multi-domain “express” core GOLE network with minimal transit policy constraints
Global Architecture A second connection is provisioned across the interoperable R&E GOLE core A P2PCS connection instance is provisioned across the LHCONE core A subsequent connection utilizes a path across the core and a collaborating R&E network