All right. Now, the next processor concept I want to mention is the logical processor. This concept represents a rather complex power systems hardware technology called simultaneous multithreading or SMT. I mentioned it earlier in another video. Over what we just saw with shared processor, SMT further splits the time for threads on a processor. But this time, it's the operating system and the hypervisor making decisions about the sharing. So threads are not using the same execution units in a processor, but they can be dispatched in the same cycle on the same processor providing the thread's execution involves some wait time. So threads that are not using the same execution units, they may end up using the same processor during the same dispatch cycle. There has to be some wait time though for the threads to make this work. Now, applications that dispatch threads that use different execution units and have a fairly high cache miss rate are likely to benefit from SMT. This includes many commercial business applications. SMT comes in different modes depending on the hardware model, and sometimes the operating system. The range of possible SMT modes include zero,, which means SMT is off, two, four, and eight. Where two, four, eight represent the number of potential threads that can be simultaneously dispatched for sharing. While the actual number of shared threads on a processor for any specific dispatch will vary based on the workload, the SMT mode will determine the number of logical processors defined in the operating system. So for example, if the LPAR has two dedicated processors or vCPUs, and an SMT value of four, then the operating system sees eight processors. Two processors, vCPUs are dedicated, times the four SMT threads. Now conversely, if the LPAR has an SMT value of zero or off, then the operating system in that case would only see two processors. Okay. So I think you could see how the arithmetic applies in other cases. So the last of the processor concepts to cover are inactive and unconfigured. Inactive processors are installed but not available until the administrator decides to turn them on. This is part of that facility that I mentioned earlier, that capacity on demand in power systems. Now, there are multiple tiers in the CoD offering, activating the processors, or memory for that matter, for fixed times as trials, for variable times, etc. There's only a charge for the time that the processors are used, or the memory is used, giving the administrator the flexibility to pay for capacity, only when needed. Like, periodically, throughout the year to cover system demand peaks, scheduled, or even on demand. Then the last of the physical or of the processor concepts is the unconfigured processor. A processor ends up, in this statement the service processor, and the hypervisor work together to take a processor offline due to detected hardware problems. This is a nice feature of the hardware. An event will be posted on the HMC for the administrator's information, so that service can be notified. That wraps up my coverage of the processor concepts. So let's move on to a close out of PowerVM virtualization.
