The Service Processor has a permanent firmware boot side also called A side and a temporary firmware boot side or known as B side. The temporary or B side is used to test the new levels of firmware. If all goes well then the B side firmware is copied back to A side. Even if the system is in power standby mode, the Service Processor keeps on running and monitoring the system for any errors. It also ensures the connection to the HMC for manageability purposes.

How To Access Service Processor
The interface to Service Processor is called Advanced System Management Interface(ASMI). All general and administrative tasks are done through the ASMI. The tasks are listed as :

1. Setting up the Service Processor.
2. Reading Service Processor error logs.
3. Reading vital product data from Service Processor.
4. Controlling the system power.

How To Access ASMI
The ASMI can be accessed via following three methods:

1. Web Browser.
2. ASCII console.
3. HMC.

How To Access ASMI Through Web Browser
The client system can run any web browser to provide access to ASMI. The SSL(Secure Socket Layer) is used with http to access the interface through browser. It’s used as https://. Internet explorer, Netscape, Opera can be used to access ASMI through https.

Starting from Initial Program Load(IPL) through full run time, the ASMI is accessible through web browser. Only relevant menus will be shown while accessing in a particular phase like IPL.

How To Access ASMI Through ASCII Console
This is a serial port on the server which uses a terminal. This provides access to ASMI.

How To Access ASMI Through HMC Interface
HMC also provides access to ASMI. Here is the short path for accessing ASMI through HMC:

1. Expand managed system for which you want to access ASMI.
2. Expand the Service Applications menu and click on “Service Focal Point“.
3. In content area click on “Service Utilities“.
4. Select the managed system from the Service Utilities.
5. There you see “Launch ASM Menu“. You can click on this.

IBM documentation (Partitioning Implementations for IBM e-server p5 Servers: SG24-7039-02), was consulted to write this article. Thanks to all of their endeavors.

What is HMC?
HMC is a Linux based desktop PC workstation which is dedicated to a number of pSeries servers. It’s is used to manage LPARs. There can be several LPARs or a single full partition on physical pSeries server system. HMC provides GUI (Graphical User Interface) and CLI (Command Line Interface) to manage LPARs.

7040-681, 7040-671, 7039-651, 7038-6M2 etc. are some of the models of HMC.

Uses of HMC
HMC is used in a number of ways to manage LPARs. Few of its uses are listed below:

1. HMC is used to start, stop, reset and shutdown an LPAR.
2. pSeries servers systems can be booted started and stopped using HMC.
3. HMC can be used to open virtual console for every partition or pSeries server connected to the HMC. It looks like original console of the server.
4. HMC can be used to create partition profiles. Each profile contains information on processor, memory and I/O resources allocated to that particular partition. It also helps to switch between various profiles of LPAR. If you want to choose different profile for an LPAR, you need to reset the LPAR.
5. HMC can be used to carryout DLPAR operations between specific LPARs. This way memory, processor etc. resources can be allocated/de-allocated dynamically without the need to restart any LPAR.
6. Physical System resources and the status of system can be displayed by HMC.
7. HMC itself can be managed through its GUI and CLI tools.
8. HMC software level can be managed by HMC itself. That’s called microcode management.
9. Some problem determination and service support are also provided by HMC which include supports like call-home and error log notifications through phone line.

 
Connection Interface For HMC
HMC connects to the server using serial interface and the protocol used is RS232. More serial ports to connect to multiple servers can be provided by Asynchronous Adapters

Creating your first LPAR on the HMC with physical resources and boot it up and accessing the remote console.

LPAR

LPAR as you all know is Logical Partition which is supported by IBM servers. During creation of LPAR CPUs, Memory and IO slots are provided to the partition from free pool of these resources. An LPAR acts as an independent server and is able to run AIX Operating System or Linux Operating Systems. In a physical server there can be many LPARs and corruption of one LPAR does not affect working of other partitions.

In case of LPAR the main security features are:

1. Protection against inter-partition data access. Means one LPAR can not access data of another LPAR. Partitions will be able to access data from another LPAR only like it will be done when these are individual machines. Means no partition can cross its boundaries.
2. If one LPAR fails, others keep running as such without fail.
3. The resources are used fairly by LPARs. No partition can use PCI bus etc indefinitely.

VIO Server

VIO server is a special case of LPAR. In this the IOs are all virtual and provided to the partitions on VIO Server. Means IO slots, CPUs and Memory are all virtual. So, how is this done? This all is done through global firmware of Partition.

The two major functions provided by VIO Server are:

1. Shared Ethernet Adapter.
2. Virtual SCSI.

Shared Ethernet Adapter (SEA)

Shared Ethernet Adapter runs in the VIO Server. Its work is to route network traffic from partitions within VIO server to the real Ethernet adapter. Hence it acts as Layer 2 switch.

The major advantage of SEA is that partitions within VIO server can communicate in-between themselves as well as to the outside world. As if now up to 18 VLANs are supported on a single interface.

Virtual SCSI Devices (VSCSI)

The partitions within VIO Server don’t have physical disks. These have virtual disks and these virtual disks are implemented through Virtual SCSI services of VIO Server. The partitions within VIO server have access to the real storage through Virtual SCSI service. The partitions see these virtual SCSI disks as real disks. SAN, SCSI and RAID are all supported for VIO SCSI disks.

VIO Server is not used much in production environments. The reason behind this is that, if suppose this LPAR acting as VIO Server gets corrupted, all of the partitions working within VIO Server go down.

For more information on LPAR and VIO server read out the book SG24-7039-02 which is freely available from redbooks.ibm.com.

The memory is categorized in four types:

1. Physical Memory.
2. Real Memory.
3. Virtual Memory.
4. Logical Memory.

Now, I discuss the types of memory one by one.

Physical Memory

The physical memory is the memory available in the system in the form of memory cards in various slots of system. The physical address space is associated with this type of memory. The Physical address space across the all the memory cards is handled by kernel VMM.

Real Memory

The Real Memory is same as Physical memory if the system is non-partitioned. But if the system is partitioned then these two types are different. Now this memory acts as physical memory for that particular partition and the address space will be different for all the partitions with in the same system.

Logical Memory

The Logical Memory comes into picture only when the system is partitioned. The operating systems like AIX, Linux etc have the following requirements:

• The address space should start from zero.
• The address space should be contiguous.

The Logical memory concept supports both of these. The logical memory represents non-contiguous memory in the form of contiguous memory.

The following benefits are achieved:

1. The partition is isolated and secured from physical memory address space.
2. It simplifies the monitoring of partitions by POWER Hypervisor.

Virtual Memory

Sometimes the memory available for operating systems is not enough to carry out all the processes. So, operating system uses the disk space as memory. Thus operating system is tricked to think that it has more memory than it actually has. Some of the virtual memory is mapped to the physical memory and rest is divided by page size (often 4kB). These pages are mapped to the disk blocks. The address translation is handled by kernel VMM.

Now to clarify all of the above terms I take very simple example:

Suppose there are 6 cards of 10 MB each. So, total memory is: 6X10 = 60MB.

Lets say we have three partitions A, B, C with respective memories assigned as 16, 24, 20 MB. Now we have following data:

Total Physical Memory: 60MB.

To recognize interface card AIX uses three notations:

ent, en and et.

All are different and are described below and the sake of completeness I am using 0 at the end:

ent0:

The notation ent0 is used to specify the hardware adapter. It has nothing to do with the TCP/IP address. The parameters associated with ent0 can be seen as below:

# lsattr –El ent0

It will show parameters related to card.

It shows adapter_names, alt_addr, auto_recovery, backup_adapter, hash_mode, mode, netaddr, noloss_failover, num_retries, retry_time, use_alt_addr, use_jumbo_frame.

en0:

en0 represents the interface associated with hardware adapter ent0. The notation en0 is used for Standard Ethernet(inet). The TCP/IP address is associated with this interface.

The parameters associated with en0 can be seen as below:

#lsattr –El en0

It’ll shows all the parameters related with the interface en0 of the adapter ent0.

It shows alias4, alias6, arp, authority, broadcast=1500, mtu, netaddr, netaddr6, netmask, prefixlen, remmtu, rfc1323, security, state, tcp_mssdflt, tcp_nodelay, tcp_recvspace, tcp_sendspace.

Rest everything is same except mtu(Maximum Transfer Unit) value. Which is 1500 as per the standard ethernet protocol.

et0:

et0 represents the interface associated with hardware adapter ent0. The notation et0 is used for IEEE 802.3 Ethernet(inet). If you are using standard ethernet protocol then it will not have TCP/IP address.

The parameters associated with et0 can be seen as below:

#lsattr –El et0

It’ll shows all the parameters related with the interface et0 of the adapter ent0.

It shows alias4, alias6, arp, authority, broadcast, mtu=1492, netaddr, netaddr6, netmask, prefixlen, remmtu, rfc1323, security, state, tcp_mssdflt, tcp_nodelay, tcp_recvspace, tcp_sendspace.

Note here as well that the MTU shown will be 1492 as per IEEE 802.3 standard. Rest all parameters will be same. Also, netaddr, netmask fields will be empty fr et0.

Minimum Capacity:

The minimum number of processing units which have to be assigned to the partition is called minimum capacity. This capacity is the committed value for the particular partition. Without this much of processing units partition can not start. This capacity can not be used to start another shared partition.

Desired Capacity:

Desired capacity is the optimum number of processing units we desire for an application to work efficiently. An LPAR is started with this value as the preferred value, but this may not always be possible. For example, if not enough resources are available in shared processor pool, then value greater than or equal to the minimum value will be used to start the partition instead of desired value.

Maximum Capacity:

This is the value we generally use during DLPAR. The purpose is to set the upper limit of moving the resources to other LPAR.

Guaranteed Capacity:

Guaranteed value is between minimum and maximum. Guaranteed value becomes important when we talk about capped and uncapped mode of processor sharing. Its normally equal to the desired value but not always. Sometimes it may be greater or less than desired value.

In capped mode the processing units given to a partition can not exceed guaranteed value, even though resources may be available in shared processing pool.

In uncapped mode the guaranteed value can be exceeded as per requirement.

So, order can be setup something like:

Minimum < Guaranteed < Desired < Maximum

Entitle Capacity:

This term comes into picture only once the partition is activated. In capped mode it’s equal to that of guaranteed value. In uncapped mode it depends upon the uncapped weight assigned to the partition. Also entitled capacity is assigned to the partitions in the order, in which they start. The entitled capacity can be removed during DLPAR.

Physical Processor
The physical processor is physical entity.

Virtual Processor
Virtual Processor is logical entity. The physical processor is divided into processing units. And one or several processing units are combined together to make virtual processor. The processing capacity is distributed in terms of virtual processors and not as processing units.

For example, if we have 2.0 physical processors and we take 0.4 processing units as one virtual processor. This will leave shared pool with 5 virtual processors.

If the processing units required for a partition are 1.2 then 3 virtual processors will be allocated for that partition.

If the processing units required for a partition are 1.7 then 5 virtual processors will be allocated for that partition.

That means the number of virtual processors is rounded to the next higher level.

In capped mode the processing units given to a partition can not exceed the assigned processing units (entitled capacity) even though there may be resources in the shared pool.

Uncapped Mode
In uncapped mode the processing units can exceed the entitled capacity of the partition if enough resources are available in the shared resource pool. At this point of time the assigned uncapped weight of the partition comes into picture.

POWER Hypervisor or simply known as Hypervisor is a piece of firmware.

What is its major benefit

It’s used to support the shared processor technology on p5 systems.

Where does it reside

The software code resides in the flash memory of service processor. It’s the underlying control mechanism which resides below Operating Systems. It owns all of the resources and creates partitions by allocating and sharing these resources.

What functions are performed by POWER Hypervisor

The Hypervisor performs the following function:

1. It initializes and configures the processors.
2. It provides virtualization support to run up to 140 partitions concurrently on

IBM p5 servers.

1. It provides virtual I/O.
2. It provides high speed communication between partitions using Virtual LAN.