Abstract: Methods and structure for enabling re-training of a DDR memory controller in a storage device without loss of data in the DDR memory devices of the cache memory in response to detecting failure of the memory subsystem during operation of the storage device. In response to detecting a failure of the memory subsystem, the memory controller is reset without resetting the memory devices. The memory controller is then re-trained for operation with the memory device. During the re-training, self-refresh mode of the memory devices is disabled and manual refresh is performed by a processor of the storage device to thereby retain any user data in the memory device.

Abstract: A method for robust preamble location and gate training in a Double Data Rate type Three (DDR3) computing environment. A single algorithm is employed to begin sampling a Data Strobe Signal (DQS) at a maximum delay value designed to fall within the driven region of a DQS. The method then begins sampling the DQS in a sequence of delay values from right to left. Each result of the sampling indicating a high state and a low state are stored as well as the occasions where the DQS transitioned from high to low indicating a rising edge. At a consecutive number of samples returning a low state, the method determines the preamble has been reached and discontinues sampling. The method retains the most recently stored rising edge as the first rising edge and configures the result for gate training.

Abstract: Methods and structure for enabling re-training of a DDR memory controller in a storage device without loss of data in the DDR memory devices of the cache memory in response to detecting failure of the memory subsystem during operation of the storage device. In response to detecting a failure of the memory subsystem, the memory controller is reset without resetting the memory devices. The memory controller is then re-trained for operation with the memory device. During the re-training, self-refresh mode of the memory devices is disabled and manual refresh is performed by a processor of the storage device to thereby retain any user data in the memory device.

Abstract: A method for detecting a failure in a serial topology. The method may comprise sending a predetermined pattern to a plurality of devices communicatively connected to an initiator in a serial topology; receiving a return result from each of the plurality of devices in response to the predetermined pattern; recognizing a problem associated with a particular device among the plurality of devices, the problem being recognized based on the return result from the particular device; sending a plurality of test patterns to the particular device; receiving a plurality of test results from the particular device in response to the plurality of test patterns; and determining a cause of the problem based on the plurality of test results, the cause of the problem being at least one of: a cable failure and a device failure.

Abstract: A method for robust preamble location and gate training in a Double Data Rate type Three (DDR3) computing environment. A single algorithm is employed to begin sampling a Data Strobe Signal (DQS) at a maximum delay value designed to fall within the driven region of a DQS. The method then begins sampling the DQS in a sequence of delay values from right to left. Each result of the sampling indicating a high state and a low state are stored as well as the occasions where the DQS transitioned from high to low indicating a rising edge. At a consecutive number of samples returning a low state, the method determines the preamble has been reached and discontinues sampling. The method retains the most recently stored rising edge as the first rising edge and configures the result for gate training.

Abstract: A method for training an electronic memory may include receiving a first delay value and a second delay value. The first delay value and the second delay value may be associated with a first data strobe indicating when to sample data on a first memory lane of the electronic memory. The method may also include determining a difference between the first delay value and the second delay value. The method may further include receiving a third delay value associated with a second data strobe indicating when to sample data on a second memory lane of the electronic memory. The method may also include determining a fourth delay value for the second memory lane of the electronic memory utilizing the third delay value and the determined difference between the first delay value and the second delay value.

Abstract: A method for training an electronic memory may include receiving a first delay value and a second delay value. The first delay value and the second delay value may be associated with a first data strobe indicating when to sample data on a first memory lane of the electronic memory. The method may also include determining a difference between the first delay value and the second delay value. The method may further include receiving a third delay value associated with a second data strobe indicating when to sample data on a second memory lane of the electronic memory. The method may also include determining a fourth delay value for the second memory lane of the electronic memory utilizing the third delay value and the determined difference between the first delay value and the second delay value.

Abstract: A method for detecting a failure in a serial topology. The method may comprise sending a predetermined pattern to a plurality of devices communicatively connected to an initiator in a serial topology; receiving a return result from each of the plurality of devices in response to the predetermined pattern; recognizing a problem associated with a particular device among the plurality of devices, the problem being recognized based on the return result from the particular device; sending a plurality of test patterns to the particular device; receiving a plurality of test results from the particular device in response to the plurality of test patterns; and determining a cause of the problem based on the plurality of test results, the cause of the problem being at least one of: a cable failure and a device failure.

Abstract: A computer system includes a communication adapter that connects a plurality of virtualized servers to one or more support system devices. The communication adapter includes a master lock register, a processing device, a queue, and a multitude of adapter access registers. Upon initialization, a virtual server asserts ownership over the communication adapter by writing its identification into the master lock register, if the register is empty. Service requests by images are transmitted to the communication adapter with an origination identification (“ID”). This ID is placed in one of the adapter access registers and the service request is placed in the queue. When a support system device responds to the service request, the response is married to the ID and broadcast back to all connected virtualized servers.

Abstract: A computer system includes a support system that report events, faults, and failures to a master virtual server. While the support system may be accessed and used by a multitude of virtual servers, only the master virtual server can manage the support system. The support system include a master lock register, a heartbeat timer, and a digital processing device (“processor”). Upon initialization and if the master lock register is empty, a virtual server asserts ownership over the support system by writing its identification into the master lock register, becoming the master virtual server. The master virtual server transmits periodic heartbeats to the support system to communicate that it is still viable and in control. If the heartbeat timer expires without communication from the master virtual server, the processor clears the master lock register and transmits a broadcast message inviting all connected virtual servers to attempt to assert control.