You've observed a daily spike in network usage within your Google Cloud project and need to pinpoint the virtual machine (VM) instances and type of traffic contributing to this surge while minimizing cost and management overhead. What should be your approach?