RANCANG BANGUN DAN EVALUASI PIPELINE PENGUMPULAN DATA LALU LINTAS JALAN MENGGUNAKAN GOOGLE MAPS DISTANCE MATRIX API PADA AWS LAMBDA DAN AZURE FUNCTIONS

Dimas Danang Andhika Santoso, Dr. Sahirul Alam, S.T.,M.Eng.

2026 | Tugas Akhir | D4 TEKNOLOGI JARINGAN

Kemacetan pada simpang perkotaan dapat terjadi karena pengendalian sinyal lalu lintas yang masih fixed-time, sehingga durasi hijau tidak selalu sesuai kondisi arus. Penelitian ini membandingkan pipeline serverless AWS dan Azure untuk mengambil data waktu tempuh Google Maps Distance Matrix API setiap 30 menit di Simpang Patangpuluhan dan membentuk dataset time-series harian (CSV). Kedua platform konsisten menjalankan eksekusi terjadwal dan menghasilkan dataset harian. Azure menunjukkan latency end-to-end lebih rendah (mean 822,15 ms; p95 1181,53 ms) dibanding AWS (mean 2514,99 ms; p95 2852,09 ms), dengan indikasi cold start proxy yang sama-sama jarang (AWS 4/48; Azure 3/48). Pada penjadwalan, AWS mengalami drift stabil berupa offset ? 12 menit 51,457 detik, sedangkan Azure umumnya tepat waktu namun memiliki outlier keterlambatan besar hingga ? 29 menit 11,332 detik. Temuan ini menunjukkan trade-off antara efisiensi eksekusi dan konsistensi waktu sampling.

Traffic congestion at urban intersections can occur because traffic signal control is still fixed-time, so green duration does not always match actual flow conditions. This study compares serverless pipelines on AWS and Azure to automatically collect travel-time data from the Google Maps Distance Matrix API every 30 minutes at the Patangpuluhan intersection and generate daily time-series datasets (CSV). Both platforms successfully executed scheduled runs and consistently produced daily datasets. Azure achieved lower end-to-end latency (mean 822.15 ms; p95 1181.53 ms) than AWS (mean 2514.99 ms; p95 2852.09 ms), while cold-start proxy indications were rare on both platforms (AWS 4/48; Azure 3/48). In scheduling accuracy, AWS exhibited a stable drift as an offset of approximately 12 minutes 51.457 seconds, whereas Azure was generally on time but showed large delay outliers up to approximately 29 minutes 11.332 seconds. These results highlight a trade-off between execution efficiency and time-sampling consistency.

Kata Kunci : serverless, Google Maps Distance Matrix API, AWS Lambda, Azure Functions, time-series