HP BL260c HP Flex-10 technology - Page 3

data traffic on the multiple networks employed in a typical data center. These heterogeneous networks, listed in Table 1, carry data server communication, management, computing, and storage. Table 1. Typical data center bandwidth requirements Network Communication - All business communication that is based on Internet Protocol (IP). This is primarily data moved over LAN. Management - This data is usually IP -based remote desktop and management services. Although some companies may combine general IP traffic with management traffic, most administrators separate these networks. Computing - Inter-Process Communication (IPC) is a method for exchanging data among two or more threads in one or more processes. HPC Cluster computing is a typical example of IPC use. IPC is employed mostly for passing instructions and redistributing large amounts of data between shared, distributed applications. IPC functions include methods for passing messages, synchronization, shared memory, and remote procedure calls. Storage - All data communication to and from storage media. This includes storage area networks (SAN). Data Traffic Email, file sharing, web services, streaming media Console OS to manage physical server and/or virtual machines, and virtual machine migration (such as VMotion) Cluster communications, Virtual machine (VM) traffic (production) Backup traffic for Fibre Channel or iSCSI Bandwidth 1Gb 2Gb 4Gb 2Gb 10GbE technology accommodates this bandwidth and allows cable aggregation to reduce cabling uplinks. But there are the additional issues of network connection availability. Using all four networks as described in Table 1 typically requires a minimum of six1Gb NICs. Utilizing all of those NICs would sacrifice data separation and full redundancy. Customers could also need at least one Fibre Channel (FC) mezzanine card slot. The result could be a server blade with all I/O bays full, leaving no room for expansion (six Ethernet modules and two FC modules). Hardware demands of virtual machine software Consolidating servers through the use of virtual machine software has become an increasingly important tool for data center administrators. One of the costs of software-based virtualization is the additional performance overhead imposed on the physical server infrastructure to create virtualized devices and services. Performance requirements Physical servers bear the additional burden of creating virtual processors and virtual NICs as well as managing these virtual components. The degree of performance degradation is dependent largely on the number of virtualized servers supported by the physical host server and the level of management services in use. NIC limitations When servers are consolidated using virtual machine software, it is often the case that there are not enough physical NIC ports available to separate data traffic into the multiple networks described in the previous section. This means several things: Limited redundancy for non-production services 3