Examples relate to scaling persistent connections for cloud computing. In some examples, a data packet is used to determine connection information of the first connection. At this stage, server portion of the first connection is closed by using the connection information to send a close command to the cloud server. In response to a keepalive signal from the client computing device, the connection information is used to send a keepalive response to the client computing device to maintain a client portion of the first connection. In response to a service request from the client computing device, a service notification including the service request is sent to the client computing device, where the client computing device initiates a second connection with the cloud server to process the service request.