IBM PowerVM boosts efficiency and usability of IBM Power systems!
Use Cases and Deployment Scope
As a cloud provider we use many IBM Power systems. They all come with IBM PowerVM built-in to provide virtualisation services. Together we have hundreds of partitions (virtual machines or LPARs as IBM calls them) running; we can dynamically change CPU- and memory allocation, move running workloads from one physical server to another etc. Also the Virtual I/O Servers provide each LPAR access to the physical network and to disks (LUNs) on the SAN. Redundant VIOS ensures high availability and the PowerVM firmware is robust, safe and highly efficient.
Pros
- Changing resources for a VM on-the-fly, like adding CPU or memory to an LPAR, adding a virtual network interface, etc. But it is just as easy to remove (or reduce) resources, especially when running the AIX operating system.
- Live Partition Mobility allows us to move LPARs from one physical system to another, without any downtime or noticeable impact on the LPAR. This allows us to balance loads, or vacate a physical machine in the event it needs downtime for maintenance.
- Power Enterprise Pool 2.0 (also called Power Private Cloud with Shared Utility Capacity) allows you to buy a base capacity (CPU and memory) for a group of physical systems, all CPU and memory on these systems get activated. If the combined usage is less than base you pay nothing, excess usage (over the base) is metered by the minute and billed (or prepaid).
- Even when you have a lot of partitions that are all consuming large amounts of CPU, nearing the total capacity of the physical machine, the overhead of IBM PowerVM is still pretty low, allowing processing in the LPARs to continue.
Cons
- Each LPAR comes with a profile that sets a minimum, desired and maximum capacity (like for CPU). You can freely change the allocated CPU between the minimum and maximum on-the-fly. If you want to change the resource below the set minimum or above the set maximum, you need to shutdown the partition first and change the profile. It would be nice if there was a way to do that without downtime too.
- (Setting the maximum very high is fine for CPU but not for memory as it allocates a fixed percentage of the maximum (not the current) size for internal housekeeping).
- When you manage your Power system with an HMC, a lot of firmware updates can be done concurrent (on-the-fly), not requiring a power-cycle of the machine. If you use Novalink to manage your systems, this proces becomes broken and you need downtime on your physical system. You can use Live Partition Mobility to move every partition off of this system first, but still...
Return on Investment
- IBM PowerVM allows us to keep our customers' AIX and Linux partitions running continuously. Concurrent firmware upgrades and Live Partition Mobility allow us to move customer workloads to different servers if needed. This supports happy customers.
- The use of PEP2.0 can save significant money if you size your base capacity correctly. If you exceed the pools base capacity you pay, but if this doesn't happen too often, it will be much cheaper than activating (and paying) for adding CPU's to the base. So it can seriously save money, while still being able to handle unexpected peak loads.
- Running AIX on an IBM Power system with IBM PowerVM is a solid combination: AIX can use all advanced features of IBM PowerVM and the hardware, developed by the same company, with single point of support etc. Linux certainly works great on any Power system, but new features of the platform are typically supported on AIX first.
Usability
Other Software Used
IBM AIX
