Compare UNIX Servers. Learn about the benefits of IBM Power Systems compared to competitive UNIX servers.
IBM POWER6™ processor-based Linux, i, and UNIX servers are designed to help you implement a dynamic infrastructure. The combination of leadership performance, energy efficiency, flexible virtualization features and RAS features are designed to reduce your cost, improve service to end users and reduce risk – especially for mission critical applications and highly virtualized, consolidated operating environments.
Join the movement to realize the benefits
More than 1800 users of competitive hardware have eliminated server farms by migrating to Power Systems over the past three years.
Learn why Sun, HP and x86 users are moving to Power. View Video
Tab navigation
- Overview- selected tab,
- Offerings
- Dare to compare
- Migration path
- Get the facts
Tab navigation
- Overview- selected tab,
- Performance
- Availability
- Efficiency
- Virtualization
Virtualization: Compare PowerVM to Sun and HP
What is virtualization
Virtualization is technology that, among other things, allows clients to increase resource utilization. The concept is to create logical servers which look like physical servers to the applications. The applications think that they have dedicated resource, but actually these logical servers share resource. That sharing increases the utilization of the resources.
Why care about virtualization
Virtualization is one of the hottest topics in IT today because it helps solve some of the biggest problems IT departments face. Energy costs are going out of sight, and the proliferation of servers has caused management costs to get out of control. Consolidation can improve the use of energy by up to 90% and by reducing the number of physical servers and associated network and SAN connections, it also reduces hardware management costs.
The most important competitive advantages of PowerVM, the Power Systems virtualization technology
Higher resource utilization: PowerVM allows clients to achieve higher resource utilization because we separate physical processors from logical processors.
Flexibility and scalability: PowerVM runs on all Power systems from Blades to the Power 595, the most powerful commercial UNIX server and it supports partitions as small as .1 of a processor and as large as the entire system.
Availability: Live Partition Mobility helps clients eliminate HW related planned downtime. Just move the partition while it is running to another server so you upgrade or maintain HW without interrupting productive work.
Higher Resource Utilization
IBM, Sun, and HP each have a virtualization technology that works with UNIX and supports shared processors.
Sun Logical Domains (LDOMs) run on their CoolThread systems. With LDOMs, the logical processors are associated with threads and threads are associated with specific cores.
HP’s Integrity Virtual Machines (IVM) associates logical processors with cores when a partition is started.
You can think of this as like a queue at a bank. With LDOMs or IVM, each customer (logical processor) can only go to one teller (physical core) every time they go to the bank. If there are many clients who need the same teller, they will wait longer for service, even though other tellers are idle. In order to meet service level objectives, the bank has to assign fewer clients to each teller. That increases the number of tellers. More tellers (cores) with the same total workload leads to lower utilization.
With PowerVM, there is a single line for all clients. The client at the head of the line gets the next available teller. This makes best use of the physical cores and minimizes the wait time for the logical processors to be dispatched
Flexibility and Scalability
PowerVM supports the broadest range of servers of any UNIX, Linux, or Windows virtualization technology. PowerVM runs on the entire Power Systems line from blades to the Power 595. And it supports partitions from as small as 1/10 of a processor to as large as the entire system.
Sun systems use different virtualization technologies for different server lines and system classes. If clients reconfigure their servers by consolidating small servers to larger servers, they must reconfigure their partitions. The virtualization technology for midrange and highend servers, Dynamic Systems Domains, does not support core sharing which limits its ability to help increase resource utilization.
HP IVM and VMWare both have a limit of 4 logical processors per partition. This limits their use to small and medium partitions. The other HP virtualization technologies, nPars and vPars do not allow core sharing.
So, for example, if a client decided to consolidate several servers, each running a couple of 8-core partitions, to one larger server, with PowerVM, they would not have to redefine the partitions. They would simply move them.
With Sun, they would have to redefine LDOMs to Dynamic Systems Domains , and they would lose any core sharing in the process. With HP or x86, since the maximum was four logical processors, they would be moving from a non-virtualized environment. Only PowerVM has the scalability and flexibility to satisfy changing requirements.
Availability
Consolidation drives increased focus on availability. Increasing the number of workloads and users on a server increases the potential disruption if the server is down. Power is ready for consolidation with many built-in reliability features designed to minimize disruptions. A Yankee Group study has shown that AIX systems are the most reliable among UNIX, Linux, and Windows systems.
The increased demand for availability of consolidated systems has led to the development of new features like Live Partition Mobility in PowerVM.
No more planned outages for applications because of hardware upgrades or reconfiguration. Instead of stopping the application, just move it. Live Partition Mobility can help clients eliminate all hardware related planned interruptions. Sun and HP UNIX technologies do not have any equivalent although VMWare’s VMotion provides similar capability for x86 environments.
The bottom line
You have to consolidate to gain control of energy and management costs. The greater the resource utilization, the more you can save. The more of the workloads you can virtualize, the more you can save. But you have to deliver to service level objectives that force you to maintain very high availability when you consolidate. PowerVM delivers for you: higher resource utilization, greater flexibility and scalability to help consolidate more of their workloads and the best availability in the industry.
The detail feature comparisons between PowerVM and virtualization technologies offered by HP, Sun, and x86 vendors is in the tables below. They define the differences that lead to the tremendous advantage of PowerVM
- twisty Virtualization Capability: IBM PowerVM vs. HP
Virtualization Capability IBM PowerVM HP nPars HP vPars HP IVM Business Value OS Supported IBM AIX, Red Hat Enterprise Linux, SUSE Linux Enterprise HP-UX, HP Open VMS I64 8.2-1 and 8.3, Microsoft Windows Server 2003, Red Hat Enterprise Linux, SUSE Linux Enterprise [5] HP-UX11i HP-UX, Windows Server 2003 Enterprise and Data Center, Red Hat Enterprise Linux [25] Broad OS support enables consolidation of a wide range of applications. Partition Scalability Up to all CPUs on machine and 2 TB of memory Dependent upon OS running in nPartition (maximums range from 8 cores to all cores, 96GB to all of memory, 1 I/O chassis to all I/O chassis) [6] Up to size of nPartition (currently limited to 8 cells) [13] [14] Up to 4 CPUs (max) [25] and 1 CPU (recommended) [26] [27] and 64 GB of memory [25] Provide service to larger applications Servers Supported Entire current Power Systems line of servers HP 9000 and HP Integrity High End and Mid Range Servers. [7] Entry level servers not supported. HP 9000 and HP Integrity High End and Mid Range Servers. [15] Entry level servers are not supported. HP Integrity Servers. [28] HP PA-9000 servers are not supported. Lowers management cost because of consistency of management practice. Improves reaction time to change since applications can be moved without replanning partition configurations. Move/add/
remove active [4] (running a workload) resources between/to/from active partitions without a rebootYes (CPU, Memory, and I/O) No. [8,9] Limited. All CPU resources can be added/moved/
removed except one boot processor/partition. [16] Memory designated as "float" can be added/removed. Memory designated as "base" can be added, but not removed. [17] 1/4 to 1/2 memory/partition must be designated as "base." [18] I/O resources cannot be added or removed. [19]Limited. All CPU [29] and Memory [30] resources can be added/moved/
removed. Storage devices can be added/ moved/
removed. [30] Virtual Storage Adapters can only be added/moved/
removed if the virtual machine is powered off. [30] Virtual Network Adapters can only be removed when the virtual machine is powered off. [30] In order to modify the logical volumes attached to a virtual machine, the virtual machine must be powered off. [31]Minimize disruption from changes in demand. Share Processing Resources between partitions Yes, up to 10 partitions/CPU No[10] No[20] Yes, up to 20 partitions/CPU [32] Lowers cost because of higher resource utilization Dynamically allocate processing resources between partitions Yes, 1/100th of a processor or more evaluated and dynamically allocable every 10 milliseconds No [10] No [21] Yes [33] Improves service because resources are autonomically moved to meet service demand. Dedicate Processing Resources to a partition Yes, as little as 1 CPU / partition and up to all processors in the machine Yes, at least one cell board (minimum of 4 sockets) / partition [11] Yes, as little as 1 CPU / partition23 and up to all processors in the nPartition. [2] Only 8 vPars are allowed per nPartition. [24] No[34] Guarantee service to the most important applications Share I/O Resources between partitions Yes No [9] No [23] Yes. However, IVM's hardware agnostic implementation requires all I/O operations to pass through the Integrity VM storage subsystem. [35] In addition, all I/O operations go through a single IVM storage subsystem [36] and the VM guests do not support multipathing. [35] Lower cost because of higher resource utilization. Dedicate I/O Resources to a partition Yes, as little as 1 dedicated I/O adapter slot / partition and up to all I/O slots in the machine Yes, at least one cell board (all I/O attached to cell) / partition [11] Yes, as little as 1 dedicated I/O adapter slot / partition [22] and up to all I/O slots in the nPartition [13] Only 8 vPars are allowed per nPartition. [24] Yes, as little as 1 dedicated device / partition. [37] However, IVM's hardware agnostic implementation requires all operations to pass through the Integrity VM storage subsystem. [35] Guarantee service to the most important applications Dedicate Memory Resources to a partition Yes, as little as 16MB of RAM and up to 2 TB Yes, at least one cell board (minimum of 16-32 DIMMs) [11] Yes, as little as 64MB of RAM [22] and up to all available memory in the nPartition [13] Only 8 vPars are allowed per nPartition. [24] Yes, as little as 512 MB of RAM and up to 64 GB of memory. [24] Memory is allocated in 64 MB chunks. [38] Guarantee service to the most important applications
- twisty Virtualization Capability: IBM PowerVM vs. Sun
Virtualization Capability IBM PowerVM Sun Dynamic Domains Sun Logical Domains Business Value OS Supported IBM AIX, Red Hat Enterprise Linux, SUSE Linux Enterprise Sun Solaris 10[39] Sun Solaris 10[47] Broad OS support enables consolidation of a wide range of applications. Partition Scalability Up to all 64 cores on machine and 2 TB of memory Up to all 256 cores on machine and 2 TB of memory [40] Up to 32 cores on machine and 64 GB of memory [48] Provide service to larger applications Servers Supported Entire current Power Servers line [*] SUN SPARC Enterprise High End and Mid Range Servers. [41] Entry level servers not supported. SUN SPARC Enterprise Low End Servers. [49] High End and Mid Range servers not supported. Lowers management cost because of consistency of management practice. Improves reaction time to change since applications can be moved without replanning partition configurations. Move/add/remove active (running a workload) resources between/to/from active partitions without a reboot Yes (CPU, Memory, and I/O) Limited. DR operations are performed at a granularity of 1/4 of a motherboard. [42] Motherboards containing kernel memory require an active partition to be suspended to perform a move/remove. [43] Running processes must manually be unbound from a processor before performing a move/remove operation on the 1/4 motherboard the processor is attached to. [44] Limited. CPU dynamic reconfiguration is disabled on all domains that contain a cryptographic unit. [50] Memory and I/O reconfiguration require a reboot and only one domain may have outstanding memory or I/O reconfiguration requests before a reboot must occur. [51] Minimize disruption from changes in demand. Share Processing Resources between partitions Yes, up to 10 partitions/CPU No [45] Yes, up to 4 (UltraSPARC T1-based systems) or 8 (UltraSPARC T2-based systems) partitions / CPU. [52] Lowers cost because of higher resource utilization Dynamically allocate processing resources between partitions Yes, 1/100th of a processor or more evaluated and dynamically allocable every 10 milliseconds No [45] No. Virtual CPUs are tied to specific physical CPU threads. [53] Improves service because resources are autonomically moved to meet service demand. Dedicate Processing Resources to a partition Yes, as little as 1 CPU / partition and up to all processors in the machine Yes, at least 1/4 of a motherboard (1 processor, 8 DIMMs, and 1/4 of I/O attached to the board) / partition [46] Yes, as little as 1 CPU / partition and up to all processors in the machine [54] Guarantee service to the most important applications Share I/O Resources between partitions Yes No [45] Yes [55] Lower cost because of higher resource utilization. Dedicate I/O Resources to a partition Yes, as little as 1 dedicated I/O adapter slot / partition and up to all I/O slots in the machine Yes, at least 1/4 of a motherboard (1 processor, 8 DIMMs, and 1/4 of I/O attached to the board) / partition [46] Limited. For UltraSPARC T1-based systems, only one partition outside the control domain can have I/O dedicated to it. [56] Guarantee service to the most important applications Dedicate Memory Resources to a partition Yes, as little as 16MB of RAM and up to 2 TB Yes, at least 1/4 of a motherboard (1 processor, 8 DIMMs, and 1/4 of I/O attached to the board) / partition [46] Yes, as little as 8KB of RAM and up to 512 GB[57] Guarantee service to the most important applications
- twisty Virtualization Capability: IBM PowerVM vs. x86 vendors
Virtualization capability IBM PowerVM VMware Infrastructure [3] Enterprise on x86 Business Benefit of PowerVM Partition scalability 64 cores, 2 TB RAM 4 cores,64 GB RAM [1] Helps improve TCO from consolidating more workloads System scalability 64 cores, 2 TB RAM 32 cores, 256 GB RAM Dynamic Logical Partitioning Yes VM reboot [2] Adapt to changes without downtime Security/fault isolation CAPP/EAL 4+ CAPP/EAL 2 [3] Secured environment for mission critical applications Support for Shared Processors Yes Higher server utilization and consolidation ratios Support for dedicated I/O Yes No [4] Superior performance for I/O intensive workloads Support for Shared I/O Yes Yes Share a resource that is typically underutilized Capacity on Demand integration Reserve CoD No [1] Add capacity when/where needed, turn it off when not required Partitions per CPU 10 8 [1] Management flexibility Partitioning Granularity 1/100th of a core, 16 MB of RAM, fraction of an I/O adapter Smaller granularity prevents over allocating resources in a sub-optimal manner to workloads Support for Live Partition Mobility Yes Yes Decrease downtime in your enterprise Linux OS Support Yes Yes
VMWare notes
1 VMWare Configuration Maximums for VMware Infrastructure 3 http://pubs.vmware.com/vi301/config_max/vi3_301_201_config_max_update.2.1.html July 23, 2007
2 Source: VMware Infrastructure 3 Online Library http://pubs.vmware.com/vi301/quick_start/vi_quick_start_manage.5.26.html "You cannot edit most virtual machine properties if the virtual machine is powered on" 10/11/07
3 Sources: VMware ESX Server 2.5.0 and VirualCenter 1.2.0 Validation Report, Version 1.7 March 27, 2006 http://www.commoncriteriaportal.org/public/files/epfiles/ST_VID10056-VR.pdf IBM certification info at http://www.ibm.com/servers/aix/products/aixos/certifications/index.html
4 VMWare Configuration Maximums for VMware Infrastructure 3 http://pubs.vmware.com/vi301/config_max/vi3_301_201_config_max_update.2.1.html July 23, 2007
HP nPartition notes
5From nPartition Administrator's Guide, see Table 1-1 "nPartition Operating System Support" http://docs.hp.com/en/5991-1247B/5991-1247B.pdf (PDF, 2.14MB) HP Part Number: 5991-1247B Published: February 2007 Edition: First Edition
6 From nPartition Administrator's Guide, see Table 3-1 "Operating System Hardware Requirements" http://docs.hp.com/en/5991-1247B/5991-1247B.pdf (PDF, 2.14MB) HP Part Number: 5991-1247B Published: February 2007 Edition: First Edition
7 From nPartition Administrator's Guide, see Table 1-2 "HP Servers Supporting nPartitions" http://docs.hp.com/en/5991-1247B/5991-1247B.pdf (2.14MB)HP Part Number: 5991-1247B Published: February 2007 Edition: First Edition
8 From HP-UX 11i v3 Dynamic nPartitions - Features and Configuration Recommendations "Cells that are assigned to an nPartition and have booted to form an nPartition are active cells" http://docs.hp.com/en/10907/dynamic_nPars_WP.pdf (PDF, 214KB) 4AA0-7690ENW Published: October 2006
9 From nPartition Administrator's Guide "You must perform a reboot for reconfig of a modified nPartition after adding cells to it" and "You must immediately perform a reboot of an nPartition when you have removed an active cell from the nPartition." http://docs.hp.com/en/5991-1247B/5991-1247B.pdf (PDF, 2.14MB) HP Part Number: 5991-1247B Published: February 2007 Edition: First Edition
10 From HP Partitioning Continuum for HP-UX11i on HP 9000 and HP Integrity Servers "Because each nPar has its own CPU, memory, and I/O" http://h71028.www7.hp.com/ERC/downloads/5982-9141EN.pdf (PDF, 8.99MB) 5982-9141EN, Rev. 1 February 2006
11 From nPartition Administrator's Guide, "Either one or two cells. Each cell has up to four processor sockets and up to 16 DIMMs" and "The Superdome 64-way/128-way server is a tightly interconnected dual-cabinet server that has from 4 to 16 cells, each with four processor sockets and up to 32 DIMMs." and "An nPartition includes one or more cells assigned to it (with processors and memory) and all I/O chassis connected to those cells." http://docs.hp.com/en/5991-1247B/5991-1247B.pdf (PDF, 2.14MB) HP Part Number: 5991-1247B Published: February 2007 Edition: First Edition
HP vPar Notes
12 From page 18 of HP-UX Virtual Partitions Administrator's Guide "vPars is a Virtual Partitions product that enables you to run multiple instances of HP-UX simultaneously on one hard partition by dividing that hard partition further into virtual partitions." http://docs.hp.com/en/T1335-90078/T1335-90078.pdf (PDF, 2.73MB) Manufacturing Part Number: T1335-90078 September 2007 Edition 13
13 From page 301 of HP-UX Virtual Partitions Administrator's Guide "Virtual Partitions exist within an nPartition, but they cannot span across nPartitions." http://docs.hp.com/en/T1335-90078/T1335-90078.pdf (PDF, 2.73MB) Manufacturing Part Number: T1335-90078 September 2007 Edition 13
14 From HP-UX 11i Virtual Partitions (vPars) - Worldwide QuickSpecs "Each virtual partition can be supported in an nPartition with a meximum of 8 cells" http://h18004.www1.hp.com/products/quickspecs/12714_div/12714_div.HTML DA-12714 - Worldwide - Version 2 - June 5, 2007
15 From HP-UX Virtual Partitions Ordering and Configuration Guide, see Table 4-4 "Server Support for vPars Releases" http://docs.hp.com/en/1705/oc.pdf (PDF, 741KB) Published: September 2007 Edition: Version 07.09.17
16 From page 217 of HP-UX Virtual Partitions Administrator's Guide "These are all the other CPUs, because all CPUs, except the boot processor of each virtual partition can be dynamically migrated." http://docs.hp.com/en/T1335-90078/T1335-90078.pdf (PDF, 2.73MB) Manufacturing Part Number: T1335-90078 September 2007 Edition 13
17 From page 195 of HP-UX Virtual Partitions Administrator's Guide "Base memory cannot be deleted from a virtual partition when a virtual partition is up. When a virtual partition is up, base memory can only be added to a virtual partition. To delete base memory, the target virtual partition must be down." and from page 196 "Float memory can be added to as well as deleted from a virtual partition while the virtual partition is up or down." http://docs.hp.com/en/T1335-90078/T1335-90078.pdf (PDF, 2.73MB) Manufacturing Part Number: T1335-90078 September 2007 Edition 13
18 From page 367 of HP-UX Virtual Partitions Administrator's Guide, see Table F-1 Minimum Base Memory Requirements http://docs.hp.com/en/T1335-90078/T1335-90078.pdf (PDF, 2.73MB) Manufacturing Part Number: T1335-90078 September 2007 Edition 13
19 From HP-UX Virtual Partitions Administrator's Guide, p. 47, Table 2-4 Dynamic Migration "Dynamic I/O Migration - No" http://docs.hp.com/en/T1335-90078/T1335-90078.pdf (PDF, 2.73MB) Manufacturing Part Number: T1335-90078 September 2007 Edition 13
20 From page 20 of HP-UX Virtual Partitions Administrator's Guide "A virtual partition uses only the cores and memory that you assign to it; cores are not time-sliced across virtual partitions." http://docs.hp.com/en/T1335-90078/T1335-90078.pdf (PDF, 2.73MB) Manufacturing Part Number: T1335-90078 September 2007 Edition 13
21 From page 32 of HP-UX Virtual Partitions Administrator's Guide "Once a virtual partition is launched, the Monitor transfers ownership of the hardware to the virtual partitions. At that point, the Monitor is not involved in accessing I/O hardware, physical memory, or process to processor cycles: the individual HP-UX instances have complete ownership of their respective hardware resources." http://docs.hp.com/en/T1335-90078/T1335-90078.pdf (PDF, 2.73MB) Manufacturing Part Number: T1335-90078 September 2007 Edition 13
22 From HP-UX 11i Virtual Partitions (vPars) - Worldwide QuickSpecs "Resource granularities: processor core, 64 MB memory, I/O slot" http://h18004.www1.hp.com/products/quickspecs/12714_div/12714_div.HTML DA-12714 - Worldwide - Version 2 - June 5, 2007
23 From Booting, Installing, Recovery, and Sharing in a vPars Environment from DVD/CDROM/TAPE/Network Version 2.0 11/05/2005 "The concept of "sharing" needs to be clarified for the purposes of this document. Within a vPars environment, I/O, including CD/DVDs and Tape drives, may not be shared internally across Pars. This means that I/O is assigned to one and only one Par and other Pars can not see or access it through any internal means, such as the vPars software." http://docs.hp.com/en/5760/vPars_DVD_TAPE_NET_Usage_V2.pdf (PDF, 72.9KB) Version 2.0 11/05/05
24 From HP-UX 11i Virtual Partitions (vPars) - Worldwide QuickSpecs "There can be a maximum of 8 virtual partitions per nPartition" http://h18004.www1.hp.com/products/quickspecs/12714_div/12714_div.HTML DA-12714 - Worldwide - Version 2 - June 5, 2007
IVM Notes
25 From HP Integrity Virtual Machines (Integrity VM) version 3.0 - Worldwide QuickSpecs "Currently supports the following guest OSs: HP-UX 11i v2 and v3, Windows Server 2003 Enterprise and Datacenter SP1 and SP2 for 64 bit Itanium-based systems, Red Hat Enterprise Linux 4 Update 4" and "Each virtual machine is tuned for up to four virtual CPUs" and "512 MB - 64 GB memory per Virtual Machine" http://h18004.www1.hp.com/products/quickspecs/12715_div/12715_div.HTML DA-12715 - Worldwide - Version 3 - July 2, 2007
26 From Best Practices for Using Integrity Virtual Machines, p. 6 "Uniprocessor Virtual Machines are More Efficient than Virtual SMPs" http://docs.hp.com/en/9983/BestPractices2.2.pdf (PDF, 551KB) 4AA1-1168ENW Rev 2.2, 3/2007
27 From Top Ten Tips for Using Integrity Virtual Machines, p. 7 "#9 - Set Entitlements Conservatively ... In general, you should use the minimum or default entitlements until one or more of the VMs are not receiving adequate CPU resources for their workloads." http://docs.hp.com/en/9985/TopTenTips2.5.pdf (PDF, 107KB) 4AA1-1169ENW Rev 3.0, 8/2007
28 From Introduction to Integrity Virtual Machines "Support of low to high-end HP Integrity servers and blades: ... And future Integrity servers and blades." http://docs.hp.com/en/9987/Intro_VM_2.1.pdf (PDF, 600KB) 4AA1-1167ENW Rev 3.0, 8/2007
29 From HP Partitioning Continuum for HP-UX 11i on HP 9000 and HP Integrity servers, p. 35 "CPU resources can be added to or removed from a partition without rebooting the partition. (Dot in Virtual Machines Column)" http://h71028.www7.hp.com/ERC/downloads/5982-9141EN.pdf (PDF, 8.99MB) 5982-9141EN, Rev. 1 February 2006
30 From HP Integrity Virtual Machines A.03.00 Installation, Configuration, and Administration, p. 120 "Dynamic Memory is an optional feature of Integrity VM that allows you to change the amount of physical memory in use by a virtual machine without rebooting the virtual machine. In this release of Integrity VM, dynamic memory is available on HP-UX guests only." and p. 32 "You can modify storage devices while the virtual machine is running. It is not necessary to restart the virtual machine; however, it may be necessary to rescan for devices on the virtual machine." and p. 72 "Using these commands, the VM Host administrator dynamically adds, deletes, and modifies storage devices on virtual machines." and p. 93 "Virtual storage devices can be added or attached while the virtual machine is powered on or off. A new virtual storage adapter can be added only when the virtual machine is off." and p. 94 "Integrity VM storage devices can be deleted or detached while the virtual machine is powered on or off. An Integrity VM storage adapter can only be removed when the virtual machine is off." and p. 106 "To remove a VNIC from a virtual machine's configuration, first stop the guest using the hpvmstop command. Then use the -d option to the hpvmmodify command." http://docs.hp.com/en/T2767-90067/index.html HP Part Number: T2767-90067 Published: April 2007, Edition 3
31 From Best Practices for Using Integrity Virtual Machines "Logical volumes ... are easy to extend, import, or export ... provided the VM using those logical volumes has been shut down and powered off." http://docs.hp.com/en/9983/BestPractices2.2.pdf (PDF, 551KB) 4AA1-1168ENW Rev 2.2, 3/2007
32 From HP Partitioning Continuum for HP-UX 11i on HP 9000 and HP Integrity servers, p. 8 "Granularity: Integrity VM can be configured with as little as 5% of one CPU" http://h71028.www7.hp.com/ERC/downloads/5982-9141EN.pdf (PDF, 8.99MB) 5982-9141EN, Rev. 1 February 2006
33 From HP Partitioning Continuum for HP-UX 11i on HP 9000 and HP Integrity servers, p. 19 "Dynamic reallocation of CPU resources (Integrity VM has workload management technology built into the VM host that moves resources from idle guest operating system instances to busy operating system instances.)" http://h71028.www7.hp.com/ERC/downloads/5982-9141EN.pdf (PDF, 8.99MB) 5982-9141EN, Rev. 1 February 2006
34 From HP Partitioning Continuum for HP-UX 11i on HP 9000 and HP Integrity servers, p. 35 "100% CPU entitlement is available, but CPU is not dedicated." http://h71028.www7.hp.com/ERC/downloads/5982-9141EN.pdf (PDF, 8.99MB) 5982-9141EN, Rev. 1 February 2006
35 From HP Integrity Virtual Machines A.03.00 Installation, Configuration, and Administration, p.72 "The VM Host maintains complete control of the physical hardware and handles the virtual machine I/O operations just as it would be handled for any other user application. Thus, just as hardware is shared among normal applications running on the VM Host, virtual machine I/O is shared across the physical storage as well." and p. 77 "Multipath solutions are supported on the VM Host only, not on virtual machines," http://docs.hp.com/en/T2767-90067/index.html HP Part Number: T2767-90067 Published: April 2007, Edition 3
36 From HP Integrity Virtualization (VSE) Competitive Advantages vs. IBM System p, slide 18 "Shared I/O is handled directly by the VM host." http://pokgsa.ibm.com/~jhuston/public/HP/HPVirtapril2007.pdf (PDF, 5.84MB), 2006
37 From HP Integrity Virtual Machines A.03.00 Installation, Configuration, and Administration, p.30 "Attached I/O devices, such as tape, DVD burner, and autochanger, are not presented as virtual devices; they are presented as direct I/O device. You specify both the physical device to allocate to the virtual machine and the virtual device name that the virtual machine will use to access the device." http://docs.hp.com/en/T2767-90067/index.html HP Part Number: T2767-90067 Published: April 2007, Edition 3
38 From HP Integrity Virtual Machines A.03.00 Release Notes, p.34 "When you specify the guest memory, use a multiple of 64 MB." http://docs.hp.com/en/T2767-90047/T2767-90047.pdf (PDF, 307MB) HP Part Number: T2767-90047Published: June 2007
Dynamic Domain Notes
39 "Each Domain runs its own copy of the Solaris OS." From p. 49 of "Sun SPARC Enterprise Mx000 Servers Administration Guide", Part No. 819-3601-11, September 2007, Revision A http://dlc.sun.com/pdf/819-3601-11/819-3601-11.pdf
40 "[The Sun SPARC Enterprise M9000-64 server] supports up to 64 dual core SPARC64 VI 64-bit processors (up to 128 cores), or up to 64 quad core SPARC64 VII 64-bit processors (up to 256 cores), or combinations of SPARC64 VI and VII processors, 2 TB of memory" From p.1 of "Sun SPARC Enterprise Servers System and Resource Management", http://www.sun.com/servers/sparcenterprise/SPARCEnt-ResMan-Final.pdf (PDF, 1.86MB)
41 See Table 4-1 "Boards, Domains, and Domain ID Numbers" on p. 50 of "Sun SPARC Enterprise Mx000 Servers Administration Guide", Part No. 819-3601-11, September 2007, Revision A http://dlc.sun.com/pdf/819-3601-11/819-3601-11.pdf and http://www.sun.com/servers/index.jsp?tab=1
42 "Each of the four XSBs contains one-quarter of the total board resources: 1 CPU, 8 DIMMS, and I/O" From p. 47 of "Sun SPARC Enterprise Mx000 Servers Administration Guide", Part No. 819-3601-11, September 2007, Revision A http://dlc.sun.com/pdf/819-3601-11/819-3601-11.pdf and "DR functions on these servers are performed on an XSB." From p. 1-2 of "Sun SPARC Enterprise Mx000 Servers Dynamic Reconfiguration User's Guide", Part No. 819-5992-11, September 2007, Revision A http://dlc.sun.com/pdf/819-5992-11/819-5992-11.pdf
43 "Then, the kernel memory on the system board to be deleted is copied into memory on the selected copy-destination system board. The system is suspended while the copying is in progress." From p. 2-8 of "Sun SPARC Enterprise Mx000 Servers Dynamic Reconfiguration User's Guide", Part No. 819-5992-11, September 2007, Revision A http://dlc.sun.com/pdf/819-5992-11/819-5992-11.pdf
44 "A CPU to be deleted must meet the following conditions: No running process is bound to the CPU to be deleted. If a running process is bound to the target CPU, you must unbind or stop the process." From p. 2-5 of "Sun SPARC Enterprise Mx000 Servers Dynamic Reconfiguration User's Guide", Part No. 819-5992-11, September 2007, Revision A http://dlc.sun.com/pdf/819-5992-11/819-5992-11.pdf
45 "Each instance of the Solaris OS running in a Dynamic Domain has access to its own resources, providing complete isolation of one Solaris instance from any other instance running on the same server." From p. 1 of "Introduction to Dynamic Reconfiguration and Capacity On Demand For Sun SPARC Enterprise Servers", Part No. 820-2084-10, Revision 1.0, 4/22/07, Edition: April 2007, http://www.sun.com/blueprints/0407/820-2084.pdf (PDF, 562KB)
46 "A domain must have, at a minimum, 1 CPU, 8 DIMMs, and I/O" From p. 49 of "Sun SPARC Enterprise Mx000 Servers Administration Guide", Part No. 819-3601-11, September 2007, Revision A http://dlc.sun.com/pdf/819-3601-11/819-3601-11.pdf LDOM notes
47 "The guest domain must run an operating system that understands both the sun4v platform and the virtual devices presented by the hypervisor. Currently, this is the Solaris 10 11/06 OS at a minimum." From p. 47 of "Logical Domains (LDoms) 1.0.1 Administration Guide", Part No. 820-3268-10, September 2007, Revision A, http://dlc.sun.com/pdf/820-3268-10/820-3268-10.pdf (PDF, 1.37MB)
48 See Table 1 "Supported Servers" from p. 2 of "Logical Domains (LDoms) 1.0.1 Release Notes", Part No. 820-3269-10, September 2007, Revision A, http://dlc.sun.com/pdf/820-3269-10/820-3269-10.pdf
49 See Table 1 "Supported Servers" from p. 2 of "Logical Domains (LDoms) 1.0.1 Release Notes", Part No. 820-3269-10, September 2007, Revision A, http://dlc.sun.com/pdf/820-3269-10/820-3269-10.pdf and http://www.sun.com/servers/index.jsp?tab=1
50 "DR of virtual CPUs is completely disabled on all active logical domains that contain any cryptographic units" From p. 13 of "Logical Domains (LDoms) 1.0.1 Release Notes", Part No. 820-3269-10, September 2007, Revision A, http://dlc.sun.com/pdf/820-3269-10/820-3269-10.pdf
51 "Dynamic reconfiguration (DR) is the ability to add or remove resources while the operating system is running. The Solaris 10 OS supports only the adding and removing of virtual CPUs (vcpus). Dynamic reconfiguration of memory and input/output is not supported in the Solaris 10 OS." and "Also, when there is a delayed reconfiguration outstanding for one logical domain, reconfiguration requests for other logical domains are severely restricted and will fail with an appropriate error message" From p. 6 and 7 of "Logical Domains (LDoms) 1.0.1 Administration Guide", Part No. 820-3268-10, September 2007, Revision A, http://dlc.sun.com/pdf/820-3268-10/820-3268-10.pdf (PDF, 1.37MB)
52 "From a Logical Domains Manager perspective, each of the threads in a system appear to be virtual CPUs, and as such, can be allocated independently to any domain in the system. In reality, threads are grouped into cores; for example, there are 4 threads per core in the Sun Fire and SPARC Enterprise T1000 CMT processors." From p. 25 of "BEGINNERS GUIDE TO LDOMS: UNDERSTANDING AND DEPLOYING LOGICAL DOMAINS for Logical Domains 1.0 release", Part No 820-0832-21, Revision 2.1, 10/9/2007, Edition: July 2007, http://www.sun.com/blueprints/0207/820-0832.pdf (PDF, 1.73MB)
53 "The domain and the virtual CPU number within the domain, which correspond to a given physical CPU number, can be determined with the following procedures." From p. 76 of "Logical Domains (LDoms) 1.0.1 Administration Guide", Part No. 820-3268-10, September 2007, Revision A, http://dlc.sun.com/pdf//820-3268-10/820-3268-10.pdf (PDF, 1.37MB)
54 "Also, there are hardware resources that are provided on a per-core basis and, therefore, are shared between the threads on that core. For the purposes of this discussion, the level 1 cache is an important such resource. Because of this, if the threads within a core are allocated to two or more separate domains, this could lead to suboptimal performance of those threads, due to cache thrashing. For that reason, to get the best performance from caches and other shared processor resources, it is best to avoid allocating the threads of a core to separate domains, if at all possible." From p. 25 of "BEGINNERS GUIDE TO LDOMS: UNDERSTANDING AND DEPLOYING LOGICAL DOMAINS for Logical Domains 1.0 release", Part No 820-0832-21, Revision 2.1, 10/9/2007, Edition: July 2007, http://www.sun.com/blueprints/0207/820-0832.pdf (PDF, 1.73MB)
55 "In Logical Domains 1.0.1 software, the virtualized I/O functionality comprises support for virtual networking, storage, and consoles." From p. 5 of "Logical Domains (LDoms) 1.0.1 Administration Guide", Part No. 820-3268-10, September 2007, Revision A, http://dlc.sun.com/pdf/820-3268-10/820-3268-10.pdf (PDF, 1.37MB)
56 "I/O domain: Domain that has direct ownership of and direct access to physical I/O devices ... The number of I/O domains you can have is dependent on your platform architecture. For example, if you are using a Sun UltraSPARC® T1 processor, you can have a maximum of two I/O domains, one of which also must be the control domain." From p. 44 of "Logical Domains (LDoms) 1.0.1 Administration Guide", Part No. 820-3268-10, September 2007, Revision A, http://dlc.sun.com/pdf/820-3268-10/820-3268-10.pdf (PDF, 1.37MB)
57 "The memory can be allocated in increments as small as 8KB chunks" From p. 15 of "BEGINNERS GUIDE TO LDOMS: UNDERSTANDING AND DEPLOYING LOGICAL DOMAINS for Logical Domains 1.0 release", Part No 820-0832-21, Revision 2.1, 10/9/2007, Edition: July 2007, http://www.sun.com/blueprints/0207/820-0832.pdf (PDF, 1.73MB) and See Table 1 "Supported Servers" from p. 2 of "Logical Domains (LDoms) 1.0.1 Release Notes", Part No. 820-3269-10 September 2007, Revision A, http://dlc.sun.com/pdf/820-3269-10/820-3269-10.pdf
We're here to help
Easy ways to get the answers you need.
or call us at 1-866-883-8901
Priority code: 6N8AX80W
Power = ROI - sprawl
Announcing new Power Systems pricing that can significantly lower your total cost of acquisition and accelerate ROI.
Power rewards
Sun-set Special!
Migrate applications from Sun to IBM and earn up to 8,000 points per core through December 31, 2009.
