bpftime

BPFtime Uprobe Benchmark Results

Generated on 2025-12-12 02:53:10

Environment

• OS: Linux 6.11.0-1018-azure
• CPU: AMD EPYC 7763 64-Core Processor (2 cores, 4 threads)
• Memory: 15.62 GB
• Python: 3.12.3

Summary

This benchmark compares three different eBPF execution environments:

• Kernel Uprobe: Traditional kernel-based eBPF uprobes
• Userspace Uprobe: BPFtime’s userspace implementation of uprobes
• Embedded VM: BPFtime’s embedded eBPF VM

Times shown in nanoseconds (ns) - lower is better

Core Uprobe Performance Summary

Operation	Kernel Uprobe	Userspace Uprobe	Speedup
__bench_uprobe	2244.11	361.83	6.20x
__bench_uretprobe	2967.79	365.23	8.13x
__bench_uprobe_uretprobe	3091.67	368.81	8.38x

Kernel vs Userspace eBPF Detailed Comparison

Operation	Environment	Min (ns)	Max (ns)	Avg (ns)	Std Dev
__bench_array_map_delete	Kernel	9411.68	9420.82	9416.65	3.78
__bench_array_map_delete	Userspace	6288.50	6310.77	6296.96	9.85
__bench_array_map_lookup	Kernel	3206.72	3225.79	3214.39	8.22
__bench_array_map_lookup	Userspace	5964.90	5976.94	5969.33	5.41
__bench_array_map_update	Kernel	44246.96	44587.92	44370.51	154.20
__bench_array_map_update	Userspace	9696.07	9777.83	9734.88	33.51
__bench_hash_map_delete	Kernel	55168.02	57033.53	55809.72	865.71
__bench_hash_map_delete	Userspace	11897.56	11902.74	11900.85	2.33
__bench_hash_map_lookup	Kernel	41794.06	43009.93	42338.45	504.45
__bench_hash_map_lookup	Userspace	21682.45	22528.11	21986.02	384.23
__bench_hash_map_update	Kernel	147510.14	148003.09	147798.39	209.74
__bench_hash_map_update	Userspace	25986.34	26834.53	26298.73	380.60
__bench_per_cpu_array_map_delete	Kernel	9430.00	10034.35	9637.12	280.97
__bench_per_cpu_array_map_delete	Userspace	6282.90	6289.20	6286.06	2.58
__bench_per_cpu_array_map_lookup	Kernel	3255.23	3570.79	3365.69	145.17
__bench_per_cpu_array_map_lookup	Userspace	11898.58	11908.97	11902.46	4.63
__bench_per_cpu_array_map_update	Kernel	44784.38	44911.17	44827.21	59.37
__bench_per_cpu_array_map_update	Userspace	28467.63	28470.11	28468.74	1.03
__bench_per_cpu_hash_map_delete	Kernel	56023.99	56418.30	56214.64	161.24
__bench_per_cpu_hash_map_delete	Userspace	89639.52	104971.59	99672.69	7098.27
__bench_per_cpu_hash_map_lookup	Kernel	41785.77	42094.16	41905.30	135.11
__bench_per_cpu_hash_map_lookup	Userspace	71736.29	71876.25	71818.78	59.81
__bench_per_cpu_hash_map_update	Kernel	109944.47	110391.21	110175.09	182.67
__bench_per_cpu_hash_map_update	Userspace	113350.83	116054.26	114485.74	1145.47
__bench_read	Kernel	77063.34	78884.87	78077.86	757.98
__bench_read	Userspace	4569.66	4724.70	4672.05	72.41
__bench_uprobe	Kernel	2237.61	2251.05	2244.11	5.50
__bench_uprobe	Userspace	358.41	366.62	361.83	3.49
__bench_uprobe_uretprobe	Kernel	3081.72	3100.68	3091.67	7.77
__bench_uprobe_uretprobe	Userspace	363.53	379.30	368.81	7.42
__bench_uretprobe	Kernel	2964.06	2972.35	2967.79	3.44
__bench_uretprobe	Userspace	364.89	365.63	365.23	0.30
__bench_write	Kernel	78222.97	78276.55	78250.67	21.91
__bench_write	Userspace	4319.05	4412.77	4380.38	43.39

Embedded VM Performance

Operation	Min (ns)	Max (ns)	Avg (ns)	Std Dev
embed	96.52	99.82	97.63	1.55

Benchmark Metadata

• Number of runs: 3
• Timestamp: 2025-12-12 02:53:10
• Total duration: 359.04 seconds

This site is open source. Improve this page.