Data flow affinity for heterogenous virtual machines

Abstract

A method and technique for data flow affinity for heterogeneous virtual machines includes disabling large receive offload (LRO) processing of data by a physical adapter. First and second logical partitions are connected to a shared virtual adapter where the first logical partition has a maximum receive unit (MRU) different than a MRU of the second logical partition. Responsive to receiving data by the shared virtual adapter destined for the first or second logical partitions, transmission control protocol (TCP) segmentation offload processing of the data is performed by the shared virtual adapter utilizing the MRU of the first logical partition as the transmission unit for data destined for the first logical partition and utilizing the MRU of the second logical partition as the transmission unit for data destined for the second logical partition.

What is claimed is: 
     
         1 . A method, comprising:
 disabling large receive offload (LRO) processing of data by a physical adapter;   connecting first and second logical partitions to a shared virtual adapter, the first logical partition having a maximum receive unit (MRU) different than a MRU of the second logical partition; and   responsive to receiving data by the shared virtual adapter destined for the first or second logical partitions, performing transmission control protocol (TCP) segmentation offload processing of the data by the shared virtual adapter utilizing the MRU of the first logical partition as the transmission unit for data destined for the first logical partition and utilizing the MRU of the second logical partition as the transmission unit for data destined for the second logical partition.   
     
     
         2 . The method of  claim 1 , further comprising, responsive to connecting of the first and second logical partitions to the shared virtual adapter, identifying the MRU for the respective first and second logical partitions. 
     
     
         3 . The method of  claim 1 , further comprising aggregating, by the shared virtual adapter, incoming packets of the data up to the respective MRU. 
     
     
         4 . The method of  claim 1 , further comprising storing, by the shared virtual adapter, the MRU corresponding to the first and second logical partitions. 
     
     
         5 . A system, comprising:
 a physical input/output (I/O) adapter having large receive offload (LRO) processing of data by the physical adapter disabled;   a hypervisor operable to manage a virtualized environment comprising first and second logical partitions, the first logical partition having a maximum receive unit (MRU) different than a MRU of the second logical partition; and   a virtual input/output server (VIOS) having a shared virtual adapter connected to the first and second logical partitions, the shared virtual adapter configured to, responsive to receiving data via the physical I/O adapter destined for the first or second logical partitions, perform transmission control protocol (TCP) segmentation offload processing of the data utilizing the MRU of the first logical partition as the transmission unit for data destined for the first logical partition and utilizing the MRU of the second logical partition as the transmission unit for data destined for the second logical partition.   
     
     
         6 . The system of  claim 5 , wherein the shared virtual adapter is operable to, responsive to a connection of the respective first and second logical partitions to the shared virtual adapter, identify the MRU for the respective first and second logical partitions. 
     
     
         7 . The system of  claim 6 , wherein the shared virtual adapter is operable to aggregate incoming packets of the data up to the respective MRU. 
     
     
         8 . The system of  claim 5 , wherein the shared virtual adapter is operable to store the MRU corresponding to the first and second logical partition. 
     
     
         9 . A computer program product for data flow affinity for heterogeneous virtual machines, the computer program product comprising:
 a non-transitory computer readable medium having computer readable program code embodied therewith, the computer readable program code comprising computer readable program code configured to:
 disable large receive offload (LRO) processing of data by a physical adapter; 
 connect first and second logical partitions to a shared virtual adapter, the first logical partition having a maximum receive unit (MRU) different than a MRU of the second logical partition; 
 responsive to receiving data by the shared virtual adapter destined for the first or second logical partitions, perform transmission control protocol (TCP) segmentation offload processing of the data utilizing the MRU of the first logical partition as the transmission unit for data destined for the first logical partition and utilizing the MRU of the second logical partition as the transmission unit for data destined for the second logical partition. 
   
     
     
         10 . The computer program product of  claim 9 , wherein the computer readable program code is configured to, responsive to a connection of the first and second logical partitions to the shared virtual adapter, identify the MRU for the respective first and second logical partitions. 
     
     
         11 . The computer program product of  claim 9 , wherein the computer readable program code is configured to aggregate incoming packets of the data up to the respective MRU. 
     
     
         12 . The computer program product of  claim 9 , wherein the computer readable program code is configured to store the MRU corresponding to the respective first and second logical partitions.