Page 274 -
P. 274

245
                                                                       C
                                                                       C h a p t e r   1 2 :  h a p t e r   1 2 :    V Vi r t u a l i z a t i o n  i r t u a l i z a t i o n    245

                         However, just by the sheer nature of virtualization and the overhead of trying to locate
                      data, there will always be some form of latency. How much will depend on the following
                      factors:

                          •  What data you’re storing
                          •  Where it has been virtualized to
                          •  How efficient your system is
                          •  The quality of your equipment

                      Throughput
                      The total amount of throughput available is also a concern. Bandwidth into and out of the
                      metadata lookup application has an impact on your system’s bandwidth. This isn’t as much of
                      an issue with asymmetric systems, because the metadata lookup occurs before any information
                      is read or written.
                         In symmetric implementations, however, throughput is limited by processing power
                      and connection bandwidths.
                      Best Practices
                      A SAN can just show up and do its job: It can store your data. But you not only want it to
                      earn its living, you want it to excel. As such, there are a few things you should expect your
                      virtualized storage solution to deliver in order to maximize the level to which you can
                      consolidate.
                      Automated Tiered Storage
                      Many organizations maintain data in different classes, which adds to the complexity of
                      maintaining a SAN. Organizations have faster (more expensive) and slower (more
                      economical) storage classes in order to better balance cost and performance. Although this
                      balance is important to strike, it is hard for IT staff to manage, because it is very labor
                      intensive. Further, IT managers need to study data patterns to determine which data needs
                      to be migrated to more economical storage so that it doesn’t stay on faster, more expensive
                      media.
                         Datacenters using this class system that want to employ a virtual storage system must
                      have a mechanism to accomplish both. The system must be able to track data usage and
                      automatically move data between storage classes based on predefined rules.
                         Data can’t just be moved to the lower, slower, more economical class. It also has to be
                      able to be promoted to the faster, more expensive drives in case a piece of data becomes   PART V
                      particularly popular with your users.
                      Green Advantages  Automated tiered storage provides a number of significant ecological
                      advantages. First, by automatically moving infrequently accessed data to energy-efficient,
                      higher capacity SATA drives, your organization will reduce the number of power-hungry
                      drives and reduce the cost of powering your system. And, as we’ve said before, less heat is
                      generated, thus reducing the amount of cooling that is necessary. As a result, fewer CO
                                                                                               2
                      emissions are produced.
                         For example, over 5 years, an enterprise-class 15,000 rpm 146GB drive with a RAID 10
                      configuration might produce 32 tons of CO  emissions. This includes the power to cool the
                                                          2
   269   270   271   272   273   274   275   276   277   278   279