IBM System p

The p in the name stands for performance, not how variously assumed for Power PC.

CPUs

The pSeries based on the power CPU, in smaller models of the PowerPC is used. In a computer, the pSeries up to 32 of these power CPUs can work together. The operating system is IBM delivers the proprietary AIX or Linux for pSeries.

PSeries computer support (from p630 onwards) the dynamic logical partitioning LPAR. The (2003) largest pSeries p690 machine can be divided into up to 32 independent partitions and manages 32 CPUs and 512 GB of memory. The resources can be dynamically moved from one LPAR to another, since AIX 5.2 is no reboot necessary.

History

2004 IBM eServer p5 introduced ( see below) and OpenPower server as the successor of pSeries.

2006, a rebranding of the pSeries to the designation was "System p" made ​​and introduced in this course also new models. These p5 - { 505, 510, 520, 550, 560, 570, 575, 590 and 595 } work with 1-64 POWER5 CPUs. An "A " in the model designation identifies POWER5 processors, a "Q" Quad - core CPUs.

The Power5 - based platforms, System i and System p are physically virtually identical in construction since the announcements of i5 and p5. The difference is the characteristics of the selected operating system OS/400 (now i5 OS), AIX, or Linux.

IBM eServer p5

In the IBM eServer p5 POWER5 CPUs can be used to support Micro-Partitioning and SMT. The IBM eServer p5 server line extending from a 1-2 way p5 505 servers to 64-way p595. p5 servers can be used with AIX and Linux and master logical partitioning, which involves installing multiple operating system instances on a server. The operating systems in the so-called partitions ( LPAR Logical Partition ) are independent of each other, and although they run on the same server, they have access only to their assigned hardware. There is a special series of p5 Server: OpenPower 710 (1-2 ways ) and OpenPower 720 (1-4 ways ) servers, which are intended exclusively for the use of Linux and do not allow the operation of AIX. All p5 and OpenPower servers based on POWER5 technology, parallel and can natively run 32- bit and 64- bit programs.

Virtualization on IBM eServer p5 and OpenPower

All servers based on p5 dominate logical partitioning, dynamic logical partitioning and micro- partitioning. Partitioning is the distribution of hardware resources to individual operating system instances, dynamic refers to the ability to move these resources without rebooting and Micro-Partitioning refers to the ability of the individual partitions to allocate fractions of processors. The virtualization technology comes from the mainframe area and is used today in various software and hardware solutions to their use.

Virtualization is at p5 servers supported by the processor technology in the firmware, called hypervisor done. The hypervisor allocates the hardware of the system and makes parts of the operating system in a logical partition (LPAR ) is available. Each LPAR thus receives a "local" firmware (a kind of BIOS), which makes it only the hardware visible, which have been assigned to her.

A CPU can be distributed to up to 10 LPARs, that is, on a 4 -way machine, up to 40 independent operating system instances are installed in logical partitions. At the present (2005 ) the largest p5 servers, the 64 -way p595 are 254 operating system instances possible. On these both AIX (5.2 and 5.3) as well as Linux can be installed simultaneously.

In contrast to most software- based virtualization solutions are available in the POWER architecture hardly any restrictions regarding the scalability of a single partition. The partition may include and use of 0.1 up to 64 processors. Dedicated and virtualized resources can be mixed within an LPAR and a server. The resulting by the virtualization overhead is difficult to prove because the hypervisor layer is always present on the p5 servers, that is, all benchmarks have also been made ​​with hypervisor.

Through virtualization and automatic load balancing a much higher overall utilization of the systems is achieved. By virtualizing the I / O resource adapters are also saved.

Requirements

For the virtualizer HMC (Hardware Management Console ) is a special computer, called required. This can take up to 32 servers and or up to 128 logical partitions Manage (LPARs ) and takes over next to the main task, divide the machine into LPARs, other functions, such as the HMC replaces the serial connections to the individual partitions, thus making additional consoles KVM switches and redundant. Under POWER5 is virtualized for small to medium-sized systems and the operation WITHOUT HMC. The functions of the HMC (with certain functional limitations ) takes in this case the so-called IVM ( Integrated Virtualization Manager).

HMC is the graphical interface for the classification of the machines and for the change of resource allocation. For the operation of the systems themselves partitioned HMC is not necessary and can be deactivated any time or replaced. In the Linux -only OpenPower systems a paid activation is necessary for virtualization.

CPU virtualization

The virtualization of the CPU can be used in two ways: CPUs can be dedicated to a partition ( operating system instance ) can be allocated ( dedicated), or there are shares in computing capacity to a partition allocated ( shared). On a server, both types of allocation can be mixed. Partitions with dedicated CPUs do not participate in automatic load balancing. All CPUs that are not allocated dedicated, remain in a so-called "shared CPU pool". This pool is used proportionately by all LPARs that are set up in "shared" mode. If the allocated cycles not used by an LPAR, they will be returned to the pool and can be used by instances that just need more CPU power.

Storage Virtualization

The memory is allocated in increments an adjustable for the entire system size (memory region size ) to individual LPARs. The memory accesses to be implemented by the hypervisor such that a partition of the storage areas of the other access can. AIX (from 5.2) can handle memory allocation with dynamic ( = on the fly ). Linux must be rebooted after changing the memory allocation.

Network Virtualization

The hypervisor performs the function of a virtual network switch. The virtual switch is VLAN -capable. Allen partitions virtual NICs can be assigned, these behave very much from the perspective of the operating system such as real physical cards. Physical and virtual NICs can be assigned simultaneously. For example, one of the partitions, which has both a physical and a virtual NIC, acting as a router or bridge for the other partitions that are only equipped with virtual NICs. Usually, this task is handled by the VIO server (see hard disk virtualization ). A physical and a virtual network adapter in the VIO server can be configured as a bridge (AIX terminology: SEA - Shared Ethernet Adapter ). If the virtual network cards of the other LPARs are in the same network and VLAN, they are transparently accessible from the outside (without specifying a gateway ).

Disk Virtualization

The hypervisor is only in a position to grant access to a PCI-X slot of a partition; single disks can not be distributed. For this purpose, a special partition is needed, called VIO server ( Virtual I / O Server). This gets the PCI-X slot assigned, in which the adapter is plugged in, to which in turn the hard disks are connected, and he is able to make parts of these plates the other LPARs as a whole virtual hard disks. There is a version of the VIO server, which is based on AIX, the same functionality can also be achieved with Linux. On the side of the client LPARs (those that use the virtual disks that are visible from VIO ) that is needed is a driver for a virtual SCSI adapter. This behaves just like an ordinary SCSI driver.