N-ANL5 versus N-ANL3A
This table gives an overview of the differences between the N-ANL5 feature set in the Sand Beach release and the N-ANL3A feature set in the Green Bay 2 release.
| Feature | N-ANL3A | N-ANL5 | Comment |
|---|---|---|---|
| Packet Descriptors | |||
| Standard packet descriptor | Yes | Yes | A number of fields are not used in Sand Beach (see DN-0968). |
| PCAP packet descriptor | Yes | Yes | PCAP descriptors are not 8-byte aligned in Sand Beach, and the FCS is removed from the frames, which is reflected in the stored length. This means that the PCAP descriptors are now fully PCAP-compliant. |
| Extended packet descriptor7 | Yes | No | - |
| Extended packet descriptor8 | Yes | No | - |
| Extended packet descriptor9 | Yes | Yes | A number of fields are not used in Sand Beach, and the descriptor now has a fixed length (see DN-0968). |
| Dynamic packet descriptor 1 | No | Yes | - |
| Dynamic packet descriptor 2 | No | Yes | - |
| RX Time-Stamping | |||
| End-of-frame time stamps | Yes | Yes | - |
| Start-of-frame time stamps | Yes | No | - |
| Native time stamp format | Yes | No | - |
| Native UNIX time stamp format | Yes | Yes | - |
| Native NDIS time stamp format | Yes | No | - |
| PCAP-ns time stamp format | Yes | Yes | - |
| PCAP-μs time stamp format | Yes | Yes | - |
| Time Synchronization | |||
| OS time synchronization | Yes | Yes | - |
| Synchronization of OS time to accelerator time | Yes | Yes | - |
| NT-TS time synchronization | Yes | Yes | Sand Beach only supports the NT-TS signal on the internal time synchronization connectors. |
| PPS time synchronization | Yes | Yes | - |
| PPS output | Yes | Yes | - |
| PPS-triggered clock sampling | Yes | Yes | - |
| PTP time synchronization | Yes | Yes | There are a number of limitations in Sand Beach (see below). |
| IEEE 1588-2008 PTP slave | Yes | Yes | - |
| IEEE 1588-2008 PTP master | Yes | No | - |
| PTP Default profile | Yes | Yes | - |
| PTP Telecom profile | Yes | Yes | - |
| PTP Power profile | Yes | No | - |
| SyncE slave frequency synchronization | Yes | No | - |
| SyncE master frequency synchronization | Yes | No | - |
| Time of day and phase reference failover | Yes | Yes | - |
| Frequency reference failover | Yes | No | - |
| Multi-Port Packet Merge | |||
| Software merge across accelerators | Yes | Yes | - |
| Hardware merge on a single accelerator | Yes | Yes | - |
| QPI bypass in an NT200C01 solution | No | Yes | - |
| Frame Classification | |||
| Layer 2: ISL encapsulation, VLAN tags, MPLS encapsulation | Yes | Yes | There is a slight difference for ISL. See DN-0968 for the implementation in Sand Beach. |
| Layer 2: EtherII, Novell_RAW, SNAP, LLC, others | Yes | Yes | There are slight differences, and the Other class is new in Sand Beach. See DN-0968 for the implementation in Sand Beach. |
| Layer 3: IPv4, IPv6, IPX, others | Yes | Yes | IPX is not supported in Sand Beach, and there are slight differences for IPv4 and IPv6. See DN-0968 for the implementation in Sand Beach. |
| Layer 4: TCP, UDP, ICMP, GRE_v0, SCTP, others | Yes | Yes | In Sand Beach GRE_v0 is included in the tunnel type classification instead of layer 4, and there is a slight difference for TCP. See DN-0968 for the implementation in Sand Beach. |
| Layer 4 port: GTPv0-U, GTPv1v2-C, GTPv1-U, others | Yes | No | - |
| Tunnel type: GTPv0-U, GTPv1v2-C, GTPv1-U_signaling, GTPv1-U_GPDU, IPinIP, others | Yes | No | There is a slight difference for GTPv0-U but IPinIP and TunnelType_Other are the same in Green Bay 2 and Sand Beach. See DN-0968 for the implementation in Sand Beach. |
| Tunnel type: GTPv0-U, GTPv1-C, GTPv2-C, GTPv1-U, GRE_v0, GRE_v1, IPinIP, EtherIP, others | No | Yes | |
| Inner layer 2: VLAN tags, MPLS encapsulation | No | Yes | - |
| Inner layer 3: IPv4, IPv6, others | Yes | Yes | There are slight differences. See DN-0968 for the implementation in Sand Beach. |
| Inner layer 4: TCP, UDP, ICMP, GRE_v0, SCTP, others | Yes | Yes | There are slight differences, and GRE_v0 is not included in Sand Beach. See DN-0968 for the implementation in Sand Beach. |
| Ethernet checksum verification | Yes | Yes | - |
| IP, UDP and TCP checksum verification | Yes | No | - |
| IP match classification | Yes | Yes | This functionality has been extended in Sand Beach. The number of match entries has been increased, and address groups and wildcard matches have been included. See DN-0968 and DN-0449 for the implementation in Sand Beach. |
| Filtering and Coloring | |||
| 64 general-purpose filters | Yes | Yes | The filter architecture is new in Sand Beach. See DN-0968 for the implementation in Sand Beach. |
| Physical port number filter | Yes | Yes | - |
| Pattern compare filter | Yes | No | - |
| Size compare filter | Yes | No | - |
| Protocol filter | Yes | Yes | More protocols are supported in Sand Beach. See DN-0968 for the implementation in Sand Beach. |
| Error conditions filter | Yes | Yes | Not all error conditions are supported in Sand Beach. See DN-0968 for the implementation in Sand Beach. |
| IP match filter | Yes | Yes | This functionality has been extended in Sand Beach. The number of match entries has been increased, and address groups and wildcard matches have been included. See DN-0968 and DN-0449 for the implementation in Sand Beach. |
| Coloring | Yes | Yes | - |
| Color mask | No | Yes | - |
| Deduplication and Correlation Key | |||
| Deduplication | Yes | No | - |
| Correlation key | Yes | No | - |
| Slicing and decapsulation | |||
| Hard-slicing | Yes | Yes | In Green Bay 2 frames are hard-sliced to 16383 bytes in the PHY/MAC and 10000 bytes during the frame processing, so the NT_NET_GET_PKT_WIRE_LENGTH macro returns a number between 10000 and 16383 for hard-sliced frames. In Sand Beach frames are hard-sliced to 10000 bytes in the PHY/MAC, so the NT_NET_GET_PKT_WIRE_LENGTH macro always returns 10000 for hard-sliced frames. |
| Fixed slicing | Yes | Yes | - |
| Dynamic slicing | Yes | Yes | An end-of-frame dynamic offset has been added in Sand Beach enabling bytes to be sliced off from the end of the frame by applying a negative offset. This can be used, for instance, for frame checksum removal. See DN-0968 and DN-0449 for the implementation in Sand Beach. |
| MPLS decapsulation | Yes | No | - |
| Hashing | |||
| Hash key types | Yes | Yes | - |
| Dynamic hash key selection | Yes | Yes | There are differences with respect to IPMatch. See DN-0449 for the implementation in Sand Beach. |
| GTP tunneling support | Yes | Yes | - |
| IPinIP tunneling support | Yes | Yes | - |
| GRE tunneling support | Yes | Yes | - |
| NVGRE tunneling support | No | Yes | - |
| Tuple swap controlled by IP match filter | Yes | Yes | The implementation differs. Sand Beach uses a workaround. See DN-0449 for the implementation in Sand Beach. |
| Tuple swap controlled by receive port number | Yes | Yes | - |
| Hash key masks | Yes | Yes | There are slight differences. See DN-0449 for the implementation in Sand Beach. |
| Hashing with configurable field selection | No | Yes | - |
| Multi-CPU Distribution | |||
| Per-port CPU distribution | Yes | Yes | - |
| Hash-based CPU distribution | Yes | Yes | - |
| Filter-based CPU distribution | Yes | Yes | - |
| Combined hash- and filter-based CPU distribution | Yes | Yes | - |
| IP Fragment Handling | |||
| IP fragment handling | Yes | No | - |
| Local Retransmission | |||
| Local retransmission between ports on one accelerator | Yes | No | - |
| Line loopback | Yes | Yes | In Sand Beach line loopback and filtering/capturing can be used independently of each other for NT100E3-1-PTP. |
| Host-Based Transmission | |||
| Host-based transmission | Yes | No | - |
| TX time-stamping | Yes | No | - |
| Limited low-rate CPU-consuming transmission | No | Yes | - |
| Buffer System | |||
| Packet-based interface | Yes | Yes | The maximum number of host buffers has been increased in Sand Beach from 32 to 128, and the maximum host buffer size has been increased to 1 TByte, and the total host buffer memory has been increased from 128 GBytes 1 TByte. |
| Segment-based interface | Yes | Yes | The maximum number of host buffers has been increased in Sand Beach from 32 to 64, and the maximum host buffer size has been increased from 128 GBytes to 1 TByte, and the total host buffer memory has been increased from 128 GBytes to 1 TByte. |
| Onboard packet buffering | Yes | Yes | - |
| IOMMU support | Yes | Yes | - |
| Statistics | |||
| Frame classifications | Yes | Yes | In Sand Beach giant undersize frames are no automatically considered as bad frames, and a frame can be large and still be a baby giant. This makes the counters RMON-compatible. See DN-0968 for the implementation in Sand Beach. |
| RMON1 counters per port | Yes | Yes | In Sand Beach jabbers counts all large erroneous frames including hard-sliced frames. See DN-0968 for the implementation in Sand Beach. |
| Extended RMON1 counters per port | Yes | Yes | In Sand Beach pktsHardSlice is always 0, since all hard-sliced frames are considered to be erroneous. See DN-0968 for the implementation in Sand Beach. |
| Checksum counters per port | Yes | Yes | In Sand Beach IP the checksum counters are always 0 since IP, UDP and TCP checksum verification is not supported. See DN-0968 for the implementation in Sand Beach. |
| Packet decoding counters per port | Yes | Yes | - |
| Extended drop counters per port | Yes | Yes | In Sand Beach pktsDedup and octetsDedup are always 0, since deduplication is not supported, and pktsMacBandwidth is also always 0. See DN-0968 for the implementation in Sand Beach. |
| IP fragment counters per port | Yes | No | - |
| Statistics per color | Yes | Yes | - |
| Statistics per stream ID | Yes | Yes | - |
| Synchronized statistics across accelerators | Yes | Yes | - |
| Performance | |||
| PCIe performance for NT100E3-1-PTP and NT80E3-2-PTP (Sand Beach only) | 56 Gbit/s | 112 Gbit/s | - |
| PCIe performance for NT40E3-4-PTP and NT20E3-2-PTP | 56 Gbit/s | 56 Gbit/s | - |