Page 77 - How Cloud Computing Is Transforming Business and Why You Cant Afford to Be Left Behind
P. 77

VIRTUALIZATION C HANGES EVERYTHING



                 sible for a practiced implementer, such as Accenture, to run
                 upward of 120 virtual machines per server, and probably
                 more. And if Accenture can do it, so can the cloud providers.
                     The x86 architecture is thus being put in a commanding
                 position to dominate cloud computing, if ways can be found
                 to automate the management of large numbers of x86 ma-

                 chines at one time.
                     Having multiplied the productivity of such servers, virtu-
                 alization weighs in again in providing the means to manage
                 large numbers of servers from a single console. This is an-
                 other match-up with the needs of the cloud. For example, if a
                 virtual machine has been assigned a “virtual” CPU, one giga-
                 byte of memory, and 33 percent of a network interface card

                 (an I/O device that plugs into a server to manage its network
                 traffic), these resources might be sufficient to keep a small
                 Web site running smoothly while being hosted in the cloud.
                 But what if something unexpected happens and traffic builds
                 on the site beyond any level envisioned, as happened with
                 Sony Music’s Michael Jackson store? It would be possible to
                 quickly double or triple the resources of the virtual machine
                 by taking them from the unused portion of the host server. Or
                 a management system could move other virtual machines off

                 the server to free up those resources, a process called live mi-
                 gration. To their end users, the virtual machines appear to
                 continue running even as they’re being moved, making it eas-
                 ier for managers to react to traffic spikes by moving VMs
                 around. As another alternative, the virtual machine manager
                 could decide to move the virtual machine that was starving for
                 more CPU to a different physical server where more cycles



                                                                      57
   72   73   74   75   76   77   78   79   80   81   82