When an HNC is connected to a 1 GbE customer network, this can be fully utilized thanks to the 16GFC connection. Up to 2x4 1GbE ports are supported.
For a connection to a 10 GbE customer network, the HNC can be upgraded by one or two 10 Gigabit PCIe Controllers (each with 2 ports). A 10 GbE can be utilized by employing link aggregation. This enables multiple Virtual Channel Adapters (VCAs) to be bundled. One device pair [read/write] in BS2000 corresponds to each VCA. The following applies here:
- By means of link aggregation via one 16GFC (1FC4VCA), this one FC channel operates at up to 60% of the theoretical capacity. Therefore, the 10GbE can only be used up to 75%.
A 10 GbE can be fully utilized by means of link aggregation via two 16GFCs (2FC8VCA).
The results (TA/s, MB/s) for the LAN connection of an SU730-70 (3,600 RPF via an HNC M4 with software version V6.5) are presented below. Measurements were made under BS2000 OSD V21.0A with openNet Server V4.0. The results may also apply to an SU710 with the same performance.
The following variants were measured:
Connection of the HNC to a 1GbE customer network and a 10GbE customer network
Use of standard frames and jumbo frames.
The following results are achieved here:
Transaction-oriented load
Maximum transaction rate of over 271.000 TA/s (here the CPU utilization of the SU730-70 was around 85%)
Mean response time with one connection around 0.37 milliseconds
Throughput-oriented load over the 1GbE
The network is fully utilized: over 110 MB/s is achieved.
The CPU utilization of the SU730-70 is up to approx. 12%.
The setting of jumbo frames results in enhanced performance: the CPU requirement is lower with a similar throughput.
Throughput-oriented load over the 10GbE
The maximum throughput which can be achieved here depends on whether and how link aggregation was configured.
Without link aggregation:
When sending 280 MB/s
When receiving 380 MB/s.
Here the CPU utilization of the SU730-70 is up to 13%. The configuration limits the performance (i.e. the device pair in BS2000). This throughput is already archieved with one connection.
With link aggregation via one 16GFC and 4 VCAs per FC (1FC4VCA):
When sending more than 900 MB/s
When receiving more than 900 MB/s
Here the CPU utilization of the SU730-70 is around 30%.
With link aggregation via two 16GFCs (2FC8VCA):
When sending 1,100 MB/s
When receiving 1,100 MB/s
Here the CPU utilization of the SU730-70 is around 40%. The 10GbE limits the performance.
These values are not attained for communication with less powerful partner systems, nor are they possible when other components such as disks or a less than ideal network slow data transfer down.
With the 10GbE the setting of jumbo frames results in only a very low performance enhancement in the CPU requirement compared to standard frames.