Melvin's digital garden

Zipper ZMap

speaker: Lukas event: Papers We Love SG ** 10 GPS on 3GHz core 200 cycles per probe L2 cache miss cost 100 cycles ** parallelized address generation avoid flooding a network iterate over a random permutation

  • random start address a_0
  • a_i+1 = g * a_i mod p split the address space into shards each shard computes
  • a_i+1 = g^n * a_i mod p ** optimized address constraints model blacklist/whitelist as a segment tree ** zero-copy packet send bypass the kernel using PF_RING ZC multiple packet creation thread and single send thread

Links to this note