Limited to fixed speed single core performanceħ Design: CHR testing lab setup Test Network for CHR AS 5459 AS 6695 CHR ebgp peer ebgp peer /29 VLAN 763 VLAN1400 Baltic Networks Vengeance HV /29 VLAN 762 iperf /29 VLAN 7 iperf /29 VLAN 777 CHR BGP AS BGP AS (15459,16696)Ĩ 4 1 UID ilo W 94% 4 1 UID ilo 1 4 RSP720-3CXL-GE PEEL LABEL FOR SUPPLEMENTAL GROUND INSTALLATION RSP720 WITH INTEGRATED SWITCH FABRIC/PFC3CXL-10GE RSP720-3CXL-GE ACTIVE ACTIVE RSP720 WITH INTEGRATED SWITCH FABRIC/PFC3CXL-10GE PWR MGMT PWR MGMT EJECT EJECT RESET RESET DISK 0 DISK UID ilo 1 4 EJECT EJECT DISK 1 DISK 1 CONSOLE CONSOLE LINK LINK Cisco 7606-S PORT 1 PORT 1 LINK LINK PORT 2 LINK PORT 2 LINK THERM S 500W 94% 500W 94% Design: CHR testing lab setup Physical Test Network for CHR AS 5459 AS 6695 ebgp peer ebgp peer HP 1U x86 Server Baltic Networks Vengeance HV 10 Gbps DAC 10 Gbps DAC Lab 7606 STATUS SYSTEM RESET STATUS SYSTEM RESET 10 Gbps DAC 10 Gbps DAC HP 1U x86 Server HP 1U x86 Server iperf3 iperf3 BGP AS BGP AS (15459,16696)ĩ Design: CHR testing lab setup Hypervisor details VM provisioning 8 vcpus Total Performance was better when using more sockets vs. Tilera Handles throughput at different frame sizes slightly better than x86 Tilera Not currently limited by ASIC. x86 More CPU and power is required to move data at the same speed as a CCR x86 Not limited by ASIC, x86 can process multiple tables more quickly than CCR if the HV host has a high clock speed CPU. Throughput At 1500 bytes, 512 bytes and 64 bytes Routing table size Impact of multiple tables x86 Better for heavy computational work. CCR for full BGP tables? Platform CPU BGP routers with full tables have a high computational requirement.
0 kommentar(er)
