Evaluating Resource Isolation for Virtual PLCs: A Full-Factorial Timing Study on an Industrial Edge Platform


Altun I., Aydoğuş H., Küçük K., Gök O.

ELECTRONICS (Basel), cilt.15, sa.22, ss.1-40, 2026 (Scopus)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 15 Sayı: 22
  • Basım Tarihi: 2026
  • Doi Numarası: 10.3390/electronics15225424
  • Dergi Adı: ELECTRONICS (Basel)
  • Derginin Tarandığı İndeksler: Scopus, Technology Collection (ProQuest), Aerospace Database, Compendex, INSPEC
  • Sayfa Sayıları: ss.1-40
  • Kocaeli Üniversitesi Adresli: Evet

Özet

Predictable control-cycle timing is challenging when virtual programmable logic controllers (vPLCs) share industrial edge resources with computational and communication workloads. The individual and combined benefits of CPU core and network-interface-card (NIC) isolation remain insufficiently characterized beyond maximum-jitter measurements. This study compares four isolation configurations (none, core, NIC, core + NIC) under eight combinations of real-time communication, CPU stress, and non-real-time traffic. A commercial vPLC with a nominal 1 ms cycle was evaluated in 160 one-hour runs (32 scenarios, 576 million cycles). Metrics include per-run maximum positive period jitter (maximum jitter), upper percentiles, and exceedance probabilities. Core isolation was associated with a 75.4% reduction in the grand mean of per-run maximum jitter relative to the unisolated setup. NIC isolation alone lowered this maximum-jitter metric but was associated with more long cycles under real-time load, mainly with CPU stress. Adding the NIC claim and its associated interrupt tuning to core isolation reduced mean maximum jitter by a further 6.5%. Each configuration package bundles a resource claim with kernel and interrupt tuning, so effects cannot be assigned to single mechanisms, and the fixed measurement order limits causal attribution. Lower maxima do not imply better tail behavior; the results are workload-specific associations, not hard real-time guarantees.