Around 1% of the people in the world suffer from epilepsy, which is the second most neurological disorder in the human after stroke. The spontaneous recurrence of seizures is the main clinical manifestation of the epilepsy. Real time detecting the seizure in the Electroencephalogram (EEG) signal is a clinical way in the diagnosis and treatment of epilepsy. The unpredicted nature of the epileptic seizures, necessitates continuous monitoring and recording of the brain activities using high-throughput neurophysiological data acquisition systems over extended periods of time. The sheer volume and the velocity of the data generated from continuous monitoring the brain activities make real-time seizure detection a big data analytic problem. In this paper, we present a Spark-based machine-learning approach to the seizure detection problem using linear dimensionality reduction and classification. Using this approach, we achieved an average accuracy, sensitivity, specificity across all patients $(\mathrm{N=24)$ of 99.32%, 99.41%, and 95.25%, respectively. Also, the average Iatency of the Spark-based seizure detection framework is about 0.38 ms.