Do some high-concurrency new pressure tests per second and find that when the L4 of the F5 data sheet reaches one-fifth of the new per second, the CPU utilization of the device will reach 100%, and then new connections will be lost. The device model is i10800, VS in Standard mode, use an iRules with tcp option 254 inserted ,use a snatpool and profile is tcp as L4 forward traffic, not use http or https offload function, 300K new connections per second will cause cpu usage 100%
The old device F5-10250V, 200K new connections per second in a high concurrency real scene will also make the cpu 100%, and its data sheet is L4 connections per second: 1.0M
F5 i10800 data sheet:
L7 requests per second: 3.5M L4 connections per second: 1.5M L4 HTTP requests per second: 22M Maximum L4 concurrent connections: 100M Throughput: 160 Gbps/80 Gbps L4/L7
I don't know how the L4 connections per second of the device data sheet is calculated. In practice, if VS is in Standard mode, it is estimated that when it reaches 25% of the new connections per second, the CPU utilization will be 100%.
VS in standard mode, 300K new connections per second(use irules) cause the CPU utilization 100%
VS in PerformanceL4 mode, 800K+ new connections per second(it not use irules) cause the CPU utilization 100%