Method and apparatus for performing checkpointing on a network device

1. A method for checkpointing in a network device having an input port, an output port, an active virtual machine, and a standby virtual machine, a network application on the active virtual machine which manipulates data present on the input port and transmits the manipulated data from the output port; a checkpoint engine on the active virtual machine; and an interface agent on the active virtual machine, having callable functions to move data from the input port to the output port, the method comprising the steps of:

determining, by the checkpoint engine, that a checkpoint is required;

requesting, by the checkpoint engine, that the interface agent quiesce itself;

returning, by the interface agent to the network application, an indicator that no packets are available regardless of whether or not packets are arriving at the input port;

determining, by the interface agent, if there are additional packets being processed by the network application and if not, setting an indication of quiescence in the active virtual machine and if so, continuing to check for additional packets being processed;

determining, by the checkpoint engine, if an indication of quiescence has been set, and if so, declaring a checkpoint, and if not, continuing to check for the quiescence indication;

if the checkpoint is declared, pause the active virtual machine;

copy memory data;

copy state data of the active virtual machine;

resume the active virtual machine; indicate

the checkpoint is complete; restart the

interface agent; and

transfer a copy of memory data and the copy of state data from the active virtual machine to the standby virtual machine.

2. The method of claim 1 wherein the interface agent is a callable library.

3. The method of claim 2 wherein the callable library is a Data Plane development Kit (DPDK) library.

4. A method of bringing a standby virtual machine to an active state comprising the steps of:

determining, by a checkpoint engine of standby virtual machine, that the active virtual machine is no longer functioning properly;

promoting, by the standby virtual machine, itself to be the active virtual machine; applying most recent checkpoint received from the improperly functioning active virtual machine;

setting a failover flag by newly-promoted active virtual machine;

starting newly-active virtual machine at a state set by last checkpoint;

initializing in the newly active virtual machine, by an interface agent, a network adapter hardware and a memory buffer holding incoming packets when the interface agent detects that the failover occurred flag has been set; and

continuing with normal processing.

5. A method for checkpointing in a network device having an active virtual machine, and a standby virtual machine, a network application on the active virtual machine; a checkpoint engine on the active virtual machine; and an interface agent on the active virtual machine, the method comprising the steps of:

determining, by the checkpoint engine, that a checkpoint is required; stopping

receipt of packets by the interface agent;

determining, by the interface agent, if there are additional packets being processed by the network application and if not, indicating quiescence;

declaring a checkpoint by the checkpoint engine, if an indication of quiescence has been set;

pausing the active virtual machine by the hypervisor;

copying memory and state data of the active virtual machine, by the hypervisor;

resuming, by the hypervisor, the active virtual machine;

restarting the interface agent by the checkpoint engine; and

transferring, by the checkpoint engine, the copy of memory and the state data from the active virtual machine to the standby virtual machine.