Because of server over-supply, cloud data centers are underused. Cloud providers are offering consumers the option of run king workloads like BID analysis of under-used resources as cheap yet revokable transition servers to increase their use of the data center (e.g., EC2 spot instances, GCE preemptible instances). Although at very low pricing, large data analysis can drastically impact work performance on unreliable cloud transient servers due to instance revocations. This study offers a cost-effective transient server delivery mechanism, iSpot, to address this issue by focusing on Spark as a model of the large-format data analysis system (DAG)-style Directed Acyclic Graph (DAG). First of all, a precise long-term short-term memory (LSTM) pricing prediction approach detects the stable cloud transient servers during the workflow execution. By employing automated work step profiling, Spark's DAG data acquisitions may create the iSpot Supply Strategy to ensure task performance on steady transient servers and develop an analytical model and provide Spark with a lightweight crucial data control mechanism. Extensive EC2- and GCE-instance prototype studies reveal that while saving workplace costs up to 83:8% as compared to state-of-the-art server supply policies, iSpot is able to ensure the performance of large-data analytics running on cloud transient servers. The overhead overtime is still acceptable.