Advanced POWER Virtualization on IBM System p5 - Previous ...

More documents

Recommendations

Info

partition with 0.50 processing units was assigned two virtual processors and processing resources were available, the partition could use an additional 1.50 processing units. The minimum number of processing units you can have for each virtual processor depends on the server model. The maximum number of processing units that you can have for each virtual processor is always 1.00. This means that a logical partition cannot use more processing units than the number of virtual processors that it is assigned, even if the logical partition is uncapped. Additionally, the number of processing units cannot exceed the Total Managed system processing units. Virtual processor folding Starting with maintenance level 3, AIX 5L V5.3 provides an improved management of virtual processors. This feature enhances the utilization of a shared processor pool by minimizing the use of virtual processors that are idle most of the time. The important benefit of this feature is improved processor affinity, when there is a large number of largely idle shared processor partitions, resulting in effective use of processor cycles. It increases the average virtual processor dispatch cycle, resulting in better cache utilization and reduced Hypervisor workload. The following are the functions of the virtual processor folding feature: ► Idle virtual processors are not dynamically removed from the partition. They are put to sleep or disabled, and only awoken when more work arrives. ► There is no benefit from this feature when partitions are busy. ► If the feature is turned off, all virtual processors defined for the partition are dispatched to physical processors. ► Virtual processors having attachments, such as bindprocessor or rset command attachments, are not excluded from being disabled. ► The feature can be turned off or on. The default is on. When a virtual processor is disabled, threads are not scheduled to run on it unless a thread is bound to that CPU. Note: In a shared partition, there is only one affinity node, hence only one node global run queue. The tunable parameter for this feature is vpm_xvcpus and the default value is 0, which signifies the function is on. Use the schedo command to change the tunable parameter. 36 <strong>Advanced</strong> <strong>POWER</strong> <strong>Virtualization</strong> on IBM System p5
Dedicated processors Dedicated processors are whole processors that are assigned to a single partition. If you choose to assign dedicated processors to a logical partition, you must assign at least one processor to that partition. You cannot mix shared processors and dedicated processors in the same partition. By default, a powered-off logical partition using dedicated processors will have its processors available to the shared processing pool. When the processors are in the shared processing pool, an uncapped partition that needs more processing power can use the idle processing resources. However, when you power on the dedicated partition while the uncapped partition is using the processors, the activated partition will regain all of its processing resources. If you want to prevent dedicated processors from being used in the shared processing pool, you can disable this function on the HMC by deselecting the Allow idle processor to be shared check box in the partition’s properties. Note: The option “Allow idle processor to be shared” is activated by default. It is not part of profile properties and it cannot be changed dynamically. 2.4.2 Shared processor pool overview A shared processor pool is a group of physical processors that are not dedicated to any logical partition. Micro-Partitioning technology coupled with the <strong>POWER</strong> Hypervisor facilitates the sharing of processing units between logical partitions in a shared processing pool. In a shared logical partition, there is no fixed relationship between virtual processors and physical processors. The <strong>POWER</strong> Hypervisor can use any physical processor in the shared processor pool when it schedules the virtual processor. By default, it attempts to use the same physical processor, but this cannot always be guaranteed. The <strong>POWER</strong> Hypervisor uses the concept of a home node for virtual processors, enabling it to select the best available physical processor from a memory affinity perspective for the virtual processor that is to be scheduled. Affinity scheduling is designed to preserve the content of memory caches, so that the working data set of a job can be read or written in the shortest time period possible. Affinity is actively managed by the <strong>POWER</strong> Hypervisor since each partition has a completely different context. Currently, there is one shared processor pool, so all virtual processors are implicitly associated with the same pool. Figure 2-3 on page 38 shows the relationship between two partitions using a shared processor pool of a single physical CPU. One partition has two virtual Chapter 2. <strong>Virtualization</strong> technologies on System p servers 37
Page 1:
Front cover Advanced</stron
Page 4 and 5:
Note: Before using this information
Page 6 and 7:
2.1 New features in Version 1.3 of
Page 8 and 9:
4.4.2 Configuration on the Virtual
Page 10 and 11:
6.6.4 Managing memory resources . .
Page 12 and 13: 3-9 Processing sharing mode and the
Page 14 and 15: 5-14 Status window . . . . . . . .
Page 16 and 17: xiv Advanced <stro
Page 18 and 19: xvi Advanced <stro
Page 20 and 21: any form without payment to IBM, fo
Page 22 and 23: The team that wrote this redbook Th
Page 24 and 25: xxii Advanced <str
Page 26 and 27: - Additional information about the
Page 28 and 29: management” on page 257 discusses
Page 30 and 31: 1.1.5 Virtual LAN 1.1.6 Virtual I/O
Page 32 and 33: The service processor provides the
Page 34 and 35: Adapter sparing Adapter sparing can
Page 36 and 37: 1.4.1 IBM AIX 5L for System p5 syst
Page 38 and 39: Note: i5/OS partitions are only sup
Page 40 and 41: Function / capability IBM System z9
Page 42 and 43: 1.6 The value of the Advanc
Page 44 and 45: 18 Advanced <stron
Page 46 and 47: 2.1 New features in Version 1.3 of
Page 48 and 49: ► throughput and performance. For
Page 50 and 51: ► IVM enhancements Industry leadi
Page 52 and 53: Shared Ethernet Adapter failover ca
Page 54 and 55: the partition. This allows you to c
Page 56 and 57: Figure 2-2 ASMI menu to enable the
Page 58 and 59: management interface for systems wi
Page 60 and 61: When setting up a partition, you ha
Page 64 and 65: processors and the other a single o
Page 66 and 67: Figure 2-5 shows the usage of the s
Page 68 and 69: 2.4.5 Dynamic partitioning Partitio
Page 70 and 71: Number of virtual processors In gen
Page 72 and 73: The benefit of SMT is greatest wher
Page 74 and 75: Setting SMT mode using SMIT Use the
Page 76 and 77: The Hypervisor firmware and the hos
Page 78 and 79: Partition 2 is configured with one
Page 80 and 81: Note: The Virtual I/O Server suppor
Page 82 and 83: Most ISVs consider partitions with
Page 84 and 85: When a partition is running as unca
Page 86 and 87: ► ► ► Selected IBM software p
Page 88 and 89: Shared pool = 4 CPU Sum EC = 3.0 PU
Page 90 and 91: For more information about the use
Page 92 and 93: Table 2-5 Licensing estimation for
Page 94 and 95: Maximum virtual processors used Cap
Page 96 and 97: 2.8 Virtual and Shared Ethernet int
Page 98 and 99: ► VLAN 2 - Node A-2 - Node B-3 -
Page 100 and 101: VLAN support in AIX 5L is based on
Page 102 and 103: 2. Tagged port receives a packet ta
Page 104 and 105: page 78. EtherChannel and Link Aggr
Page 106 and 107: Note: In this type of configuration
Page 108 and 109: Figure 2-18 Shared Ethernet Adapter
Page 110 and 111: Other aspects that have to be taken
Page 112 and 113:
► Only packets for the VLAN speci
Page 114 and 115:
► Packets tagged with a VLAN othe
Page 116 and 117:
► The value proposition enabling
Page 118 and 119:
Figure 2-21 Virtual SCSI architectu
Page 120 and 121:
Figure 2-23 and Figure 2-24 show th
Page 122 and 123:
4. In the Virtual I/O Server, vios1
Page 124 and 125:
The following are the steps to unco
Page 126 and 127:
► ► A logical volume on the Vir
Page 128 and 129:
I/O configuration, performance is a
Page 130 and 131:
Hardware management is done by an H
Page 132 and 133:
VMC allows IVM to provide basic log
Page 134 and 135:
Virtual TTY In order to allow LPAR
Page 136 and 137:
2.13.1 Linux device drivers for IBM
Page 138 and 139:
Mirroring Linux can mirror disks by
Page 140 and 141:
Figure 2-29 Bridging a virtual and
Page 142 and 143:
116 Advanced <stro
Page 144 and 145:
3.1 Getting started This section pr
Page 146 and 147:
lspath mkpath mkvdev mkvt rmdev rmp
Page 148 and 149:
3.1.2 Hardware resources managed Th
Page 150 and 151:
3.2 Creating a Virtual I/O Server p
Page 152 and 153:
3. Enter the partition name and ID
Page 154 and 155:
5. You have the option to change th
Page 156 and 157:
7. Select the Shared check box for
Page 158 and 159:
9. Specify the processing sharing m
Page 160 and 161:
11.Skip setting the I/O pools, as s
Page 162 and 163:
2. Click the Create adapter button
Page 164 and 165:
5. Add a virtual SCSI adapter to vi
Page 166 and 167:
7. Select Normal for your boot mode
Page 168 and 169:
A few seconds after the status wind
Page 170 and 171:
3. Select the VIO_Server1 profile a
Page 172 and 173:
You are now ready to use the newly
Page 174 and 175:
3. Specify the Maximum virtual adap
Page 176 and 177:
You can use the lsmap -all -net com
Page 178 and 179:
2. Select the check box AIX or Linu
Page 180 and 181:
2. Click Create client adapter. The
Page 182 and 183:
4. Click Create server adapter and
Page 184 and 185:
6. When the operation is complete,
Page 186 and 187:
9. When you have completed adding v
Page 188 and 189:
Server SCSI adapters have been adde
Page 190 and 191:
Using logical partitions In our bas
Page 192 and 193:
Note: Based on the lsdev -vpd comma
Page 194 and 195:
It is useful to be able to map the
Page 196 and 197:
Assuming that a NIM master is confi
Page 198 and 199:
3. Set up the NIM master IP address
Page 200 and 201:
3.4.6 Mirroring the Virtual I/O Ser
Page 202 and 203:
Once the interaction between virtua
Page 204 and 205:
Figure 3-45 shows the steps to crea
Page 206 and 207:
Ethernet server adapter and a physi
Page 208 and 209:
4.1 Providing higher serviceability
Page 210 and 211:
Figure 4-1 MPIO and LV Mirroring wi
Page 212 and 213:
Some advanced topics with respect t
Page 214 and 215:
While EC is an Cisco-specific imple
Page 216 and 217:
Alternatives for network high avail
Page 218 and 219:
Figure 4-6 IP multipathing in the c
Page 220 and 221:
SEA Failover works as follows: Two
Page 222 and 223:
Although the alternative configurat
Page 224 and 225:
► With NIB, the two internal netw
Page 226 and 227:
In Figure 4-11 on page 201, a clien
Page 228 and 229:
When Virtual I/O Server 2 returns t
Page 230 and 231:
few cycles to set up, but none for
Page 232 and 233:
4.1.7 Considerations The following
Page 234 and 235:
. On step 10, allocate an unused Et
Page 236 and 237:
7. Check the devices list on your V
Page 238 and 239:
Figure 4-17 shows the intended solu
Page 240 and 241:
. Click the Create Adapter button,
Page 242 and 243:
Tip: Mismatching SEA and SEA Failov
Page 244 and 245:
Checking which SEA is active Checki
Page 246 and 247:
In this scenario, we attach each Vi
Page 248 and 249:
2. Configure the SEA, as described
Page 250 and 251:
5. Check for the vhost adapter usin
Page 252 and 253:
8. Check again, using the lsdev com
Page 254 and 255:
Check the mappings, as shown in Exa
Page 256 and 257:
2. Check the MPIO configuration by
Page 258 and 259:
To determine which path will use th
Page 260 and 261:
7. On the first LPAR, create an enh
Page 262 and 263:
A switch can be forced with the /us
Page 264 and 265:
The configurations described in thi
Page 266 and 267:
We recommend using mirroring, strip
Page 268 and 269:
Supported multi-path configurations
Page 270 and 271:
Supported scenarios using one Virtu
Page 272 and 273:
Supported scenarios using multiple
Page 274 and 275:
4.7.2 Supported Ethernet configurat
Page 276 and 277:
In order to help clients improve th
Page 278 and 279:
Clients can find the fixes and file
Page 280 and 281:
Figure 4-35 shows the basic conside
Page 282 and 283:
256 Advanced <stro
Page 284 and 285:
5.1 Dynamic LPAR operations This se
Page 286 and 287:
5. Figure 5-3 shows you the Virtual
Page 288 and 289:
another partition. Use the lsslot -
Page 290 and 291:
2. Select the adapter to be moved a
Page 292 and 293:
5.1.3 Add memory dynamically in AIX
Page 294 and 295:
5.1.4 Removing memory dynamically T
Page 296 and 297:
The following command shows the eff
Page 298 and 299:
2. Figure 5-17 shows the current pr
Page 300 and 301:
et1 Defined 01-09 IEEE 802.3 Ethern
Page 302 and 303:
1. Use the diagmenu command to unco
Page 304 and 305:
5. Locate the blinking adapter, rep
Page 306 and 307:
► Figure 5-19 shows you the topol
Page 308 and 309:
Example 5-4 shows how to back up th
Page 310 and 311:
Example 5-6 Creating a Virtual I/O
Page 312 and 313:
5.2.5 Restoring the Virtual I/O Ser
Page 314 and 315:
For an AIX 5L NIM server recovery,
Page 316 and 317:
To view the System Plan, Select Sys
Page 318 and 319:
In Example 5-11, we look to the map
Page 320 and 321:
The last information we need is the
Page 322 and 323:
In a mirrored environment, issue th
Page 324 and 325:
Figure 5-23 shows that the virtual
Page 326 and 327:
csm.diagnostics 1.5.0.0 # Cluster S
Page 328 and 329:
os.txt.spell 5.3.0.10 # Writer's To
Page 330 and 331:
hd3 jfs2 1 2 2 open/syncd /tmp hd1
Page 332 and 333:
Adding a PCI hot plug adapter The t
Page 334 and 335:
The following is a list of empty SC
Page 336 and 337:
After the disk is replaced and brou
Page 338 and 339:
2. Get a larger disk from the Virtu
Page 340 and 341:
The updateios command does not comm
Page 342 and 343:
Note: You can use dsh based on rsh,
Page 344 and 345:
2 Checking if Redundancy with dual
Page 346 and 347:
STATUS=`lspath -l $n | grep Failed
Page 348 and 349:
Traditional performance measurement
Page 350 and 351:
CPU statistics in shared-processors
Page 352 and 353:
This is shown in Example 5-22 and E
Page 354 and 355:
5.5.4 The topas command The topas c
Page 356 and 357:
capped, and uncapped. The Mem and t
Page 358 and 359:
Tip: The output of topas -cecdisp w
Page 360 and 361:
These files can now be processed wi
Page 362 and 363:
interval and count parameters. The
Page 364 and 365:
The lparstat command POWER<
Page 366 and 367:
Note: The variable capacity weight
Page 368 and 369:
Column Measured parameter Comments
Page 370 and 371:
The sysstat command The sysstat com
Page 372 and 373:
write: wps avgserv minserv maxserv
Page 374 and 375:
5.5.7 Monitoring with PLM The Parti
Page 376 and 377:
Use the perfwb command to start the
Page 378 and 379:
Using the r and p switches after st
Page 380 and 381:
The nmon process runs in the backgr
Page 382 and 383:
The POWER Hypervis
Page 384 and 385:
5.7 Security considerations for Vir
Page 386 and 387:
The firewall can be switched on and
Page 388 and 389:
Enable mail client (dismaildmndls):
Page 390 and 391:
Network option extendednetstats (ex
Page 392 and 393:
Enable unsecure daemons (disrmtdmns
Page 394 and 395:
Minimum number of chars (mindiffhls
Page 396 and 397:
370 Advanced <stro
Page 398 and 399:
6.1 Partition Load Manager introduc
Page 400 and 401:
Notes: ► The PLM server may resid
Page 402 and 403:
When PLM is notified that a partiti
Page 404 and 405:
6.1.4 Memory management PLM manages
Page 406 and 407:
However, the managed nodes know not
Page 408 and 409:
sh and rcp During its installation,
Page 410 and 411:
2. Create a toc file with the inuto
Page 412 and 413:
. Append the PLM server public key
Page 414 and 415:
You can set up the RMC using a scri
Page 416 and 417:
Define a PLM policy as follows: 1.
Page 418 and 419:
Table 6-2 CPU-related tunables CPU
Page 420 and 421:
Memory tunables Table 6-4 shows the
Page 422 and 423:
1. Start the Web-based System Manag
Page 424 and 425:
4. Complete the information in the
Page 426 and 427:
. Fill out the values for the tunab
Page 428 and 429:
8. Specify the CPU resource policy
Page 430 and 431:
Clicking the OK button will create
Page 432 and 433:
. From the Web-based System Manager
Page 434 and 435:
2. Click the Groups tab, and then t
Page 436 and 437:
Figure 6-21 Edit Policy File partit
Page 438 and 439:
The policy_file and log_file parame
Page 440 and 441:
In its short form, the xlplm -Q com
Page 442 and 443:
6.3 Point-in-time and recurring rec
Page 444 and 445:
Select the type of operation you wi
Page 446 and 447:
6.4 Tips and troubleshooting PLM Th
Page 448 and 449:
Note: Exchanging keys is an importa
Page 450 and 451:
The Identity entry is either based
Page 452 and 453:
120200c10d3ad7a20951523b7c01bd31f42
Page 454 and 455:
► ► PLM does not directly state
Page 456 and 457:
6.6 Resource management Owners of I
Page 458 and 459:
Characteristics of WLM, PLM, and sh
Page 460 and 461:
Evaluating CPU utilization Traditio
Page 462 and 463:
Shared-processor partitions support
Page 464 and 465:
438 Advanced <stro
Page 466 and 467:
FDX FLOP FRU FTP GDPS GID GPFS GUI
Page 468 and 469:
442 Advanced <stro
Page 470 and 471:
Other publications These publicatio
Page 472 and 473:
► ► ► ► Virtual I/O Server
Page 474 and 475:
cfgdev command 163, 203, 264 cfgmgr
Page 476 and 477:
excess weight, PLM 376 extendlv com
Page 478 and 479:
Micro-Partitioning 3, 32-33 49 capp
Page 480 and 481:
processor dispatch 43, 50 processor
Page 482 and 483:
VIO 59 virtual processor 61 mainten
Page 484 and 485:
target 90, 92 virtual SCSI Server a
Page 488:
Back cover Advanced</strong
show all

Advanced POWER Virtualization on IBM System p5 - Previous ...

Create successful ePaper yourself

Delete template?

Save as template?