Introduction to IBM X-Series Server Enterprise Architecture Technology
IBM Enterprise Type X Architecture (EXA) features and benefits:
The IBM Enterprise X-Architecture demonstrates how a cleverly conceived approach to evolution can create innovative features. The Enterprise Type X architecture uses three industry-standard server technology components—processor, memory, and I/O—and is further enhanced with advanced features designed to take standard systems to the next level.
The Enterprise Type X architecture brings features to the industry's standard servers that were previously only available to mainframe and other high-end system users. These new features, combined with existing X-type architecture technologies, create revolutionary scalability, economy, unmatched flexibility, and new levels of availability and performance. Key features that delight customers by simplifying management, reducing costs, and improving availability include:
o XpandOnDemand scalability, system segmentation, PCI–X I/O subsystem, Active PCI–X
o I/O
o Memory ProteXion
- Chipkill memory
- Memory mirroring
- Hot-added/hot-swappable memory (coming soon)
o XceL4 server accelerator cache
In the following content, we will introduce in detail the four aspects of server scalability, L4 cache, memory technology, and I/O.
Enterprise X-type architecture: XpandOnDemand
Thanks to its flexible modular design, the Enterprise X architecture creates a revolutionary new economy for servers: customers no longer need to purchase as many servers as they can upfront to ensure future capacity growth. You can pay as you grow. We call this innovative XpandOnDemand scalability.
Enterprise X-type architecture technology uses an enhanced, high-performance 4–way SMP standard building block called an SMP expansion module. Using these 4-way modules as scalable enterprise nodes, IBM SMP expansion modules allow efficient expansion from 4-way to 8-way to 12-way — and even 32–way systems, connecting them together via a single high-speed SMP expansion port. Therefore, if the customer eventually needs more processing capabilities, a spare 4–lane module can be added to create an 8–socket server combined with simple wiring. If these 8-socket servers do not provide enough slots and bays, they can further increase the I/O slot capacity by plugging in external remote I/O expansion units (described later) and remote storage units such as the IBM EXP500.
Enterprise Type X Architecture SMP expansion modules include processors, memory, I/O support, cache, storage, and other devices that can be run separately like other servers. Each module can run an operating system that is different from the others, or multiple modules can be assigned to an operating system version through system segmentation if needed. With system segmentation, a system can be configured as a memory system that shares 16–processors, or split into multiple segments. Ultimately, when all EXA features are supported, a segment is as small as a processor.
Modules are connected to each other by dedicated high-speed interconnect devices called SMP expansion ports, sharing resources for near-linear scalability, allowing users to adapt to run multiple nodes as a large conglomerate unit, or as two or more smaller units—or even rearrange the configuration later as needed.
EXA technology also provides access between all processors and all memory, independent of their respective nodes, thus reducing connectivity. With each additional node, you can also add chipsets, front-end buses, PCI buses, and other resources to share data traffic. More nodes equal more system bandwidth. Imagine the conflicts and resource issues you encounter in a traditional 16– or 32–way SMP system.
Similarly, supporting a cluster of servers connected through failover is as simple as connecting two, three, or four 4-way nodes. You can use the same system extension port routing between nodes for cluster interconnecting. For scalable clusters, a high-speed interconnect can be created without a complex Ethernet setup, as it already exists via SMP expansion ports. In addition, the Ethernet PCI–X slot is open to other I/O.
SMP Extension Module Technology: XceL4 Server Accelerator Cache
An advanced feature supported by Enterprise Type X Architecture (EXA) is a massive Level 4 (XceL4 Server Accelerator Cache) system cache that ensures the proper functioning of SMP expansion module memory performance technology, with 64 MB of 400 MHz DDR (Double Data Transfer Rate) high-speed ECC memory per SMP expansion module in Itanium based servers compared to 32 MB in Xeon systems.
By using high-speed DDR memory between the processor and the main memory, the XceL4 cache can greatly improve the performance of the processor and I/O devices. How much performance has been improved? In an industry where vendors boast a performance advantage of more than 2% over competitors, XceL4 caching can increase throughput across all servers by up to 15% to 20%.
Intel 32–bit and 64–bit processors contain relatively small scale (128 K to 4 MB, depending on the processor) of Level 1, Level 2, and (using Itanium) Level 3 built-in cache memory. The amount of built-in cache is limited by the space available inside the processor module. The larger the cache memory, the more often the processor will look for the data it needs, and the less it will have to access slower main memory. (Processor speed is increasing at a rate much faster than the speed of main memory; The number of times that the main memory must be accessed increases every year. )
Large memory capacity
Active Memory is a breakthrough in the mass memory technology of enterprise X-type architectures, designed to increase capacity, performance, and reliability. One such technology is the ability to support large memory capacities.
While some servers are still limited by the number of memory slots they can install, others are limited by the maximum memory that the chipset the server is using can support. For these reasons, most servers have a memory limit of 16 GB RAM or less. The Enterprise Type X architecture breaks this barrier, allowing up to 256 GB of RAM (64 GB in a server based on a 32–bit Intel Xeon MP processor) in a server based on a 64–bit Itanium-based server.
Memory ProteXion
Memory ProteXion helps protect against sudden failures caused by hard memory errors. It works somewhat similarly to hot spare disk sectors in the Windows NTFS file system, and if the operating system detects bad sectors on the disk, it will write data to the spare sector for this purpose. Memory ProteXion (also known as redundant bit tuning on other systems) was originally developed for IBM mainframes and has been used for many years on zSeries and iSeries servers.
Servers protected by Memory ProteXion are nearly 200 times less likely to fail than a server using standard ECC memory. The ECC (Error Detection and Correction) DIMM contains 144 bits, but only 140 bits are used for data, and the remaining four bits are unused. Memory ProteXion simply rewrites data to some of these spare bits, rather than quickly disabling DIMMs. This approach allows Memory ProteXion to correct four consecutive bit errors per DIMM—eight consecutive bit errors per memory controller (a server may have multiple controllers). This advanced technology can help reduce server downtime, resulting in a more robust client-server computing platform. This is especially important in large database environments, where transactions/rollbacks, re-indexing, and data synchronization between servers can result in hours of loss before a crashed database is back up and running. If a memory controller is running outside of the standby bit, it continues to act as a second line of defense for Chipkill memory.
Chipkill ECC memory (now the third generation of the industry's standard computers) only works when a server suffers so many errors in a short period of time that Memory ProteXion can't solve it.
Memory mirroring
The third line of defense against server downtime due to memory failures is memory mirroring. In this technology, memory is managed in a very similar way to disk mirroring in a RAID configuration. In this case, the exact mapping of the data on the main memory stick is mirrored to the spare or backup memory module. The result is that if one memory stick fails, the mirrored memory stick becomes the main memory stick. After replacing the failed memory stick, the data in the memory of the main memory stick is mirrored to the new memory stick.
PCI–X I/O system and Active PCI–X
The latest PC I/O buses allow for multiple 64–bit 66 MHz PCI bus segments, supporting 400 to 500 MBps per segment. This bandwidth is not enough to support emerging 10 Gbps (gigabytes per second)—or higher—I/O environments.
Without other performance improvements, PCI will quickly become a bottleneck preventing these high-speed networks from connecting servers at maximum network speeds. I/O bottlenecks have prevented industry-standard servers from becoming a balanced system architecture, a feature of high-speed Intel-based servers and mainframe systems. Therefore, to address these performance issues, the industry has developed an enhanced bus called PCI–X, which is designed to extend the life of PCI until next-generation serial I/O architectures such as InfiniBand are ready.
PCI–X allows all current 32–bit and 64–bit 66 MHz PCI adapters to function properly in the PCI–X bus. The PCI–X adapter takes full advantage of the new 100 MHz and 133 MHz bus rates, which allow a single 64–bit adapter to deliver up to 1 gigabyte of data per second. In addition, PCI–X supports twice as many PCI 66 MHz 64-bit adapters in a single bus.
Active PCI–X allows you to add or replace Active PCI and Active PCI–X supported cards without shutting down the server. The Active PCI–X features designed to improve overall server availability are categorized as follows:
Hot-swappable allows you to replace a faulty or impending adapter without rebooting
Hot Add provides easy upgrades that allow you to add new adapters while the server is running (IBM was the first in the industry to offer this feature)
Failover allows the backup adapter to be responsible for running all the services being processed in the event of a primary adapter failure
Technical Questions About the 8658-51Y 5100X230 Server:
1.8658 11Y----21Y—61Y-6RY and other NF 5100/X230 motherboards are all the same, this kind of server IBM is due to
There is a problem with the production design, and its CPU first slot VRM error, which will burn the CPU and motherboard in severe cases.
2. In order to solve this problem, IBM later had a 5100 improved board called FRU: 59P5869
You can not burn the CPU VRM, that is, the first slot of the CPU, you can load the CPU normally: Some major customers are IBM Send Basket Fast
The engineer replaced the motherboard with the FRU:59P5869 improved board.
3. There is another way: Lankuai's engineer approach (practiced) to move the CPU to the second CPU slot
Add a VRM CPU terminal board from the original second CPU slot to the first slot of the CPU, and so on
It avoids the loss of burning the first CPU. That is, the server can only go up to one CPU
The second CPU slot. This fits FRU: 09N7844 06P6165 25P3289 i.e. non-modified plates.
4. This is also the reason why the IBM 5100/X230 is prone to problems, but there is also a solution.
So a good CPU should never go to the first slot of the CPU.
Detailed explanation of the Ipssend command and configuration method
Ipssend is a tool for configuring arrays on the command line, the command file itself is very small, easy to download from the Internet, which can solve the problem of some users losing server raid, server guide discs and not being able to download about 500Mb of disc iso image files from the Internet.
Main commands:
1.create - The function of this command is to create a logical drive on top of an existing array or a new array.
Note: This command cannot create a logical drive for RAID level-x0.
Command Format : IPSSEND CREATE controller LOGICALDRIVE NEWARRAY/ARRAYID size raidlevel {channel sid}
l controller refers to the ID number of the RAID controller (1-12)
l NEWARRAY means to create a new array (if you don't want to create a new array, you can leave it out)
l size and raidlevel are the levels of the size and array of logical drives to be created, respectively
Example: (Default controller is 1, hard disk id starts from 0, logical drive size is 100Mb)
1. A hard disk does raid 0: ipssend create 1 logicaldrive newarray 100 0 1 0. The last 1 0 refers to the corresponding {channel sid}
2. Do raid 0 on two hard drives: ipssend create 1 logicaldrive newarray 100 0 1 0 1 1. The last 1 0 1 1 refers to the corresponding {channel sid}
3. Two hard drives do RAID 1: ipssend create 1 logicaldrive newarray 100 1 1 0 1 1. The last 1 0 1 1 refers to the corresponding {channel sid}
4. Three hard drives do RAID 5: ipssend create 1 logicaldrive newarray 100 5 1 0 1 1 1 2. The last 1 0 1 1 1 2 refers to the corresponding {channel sid} command that will define this newarray as array a.
5. If you want to create another logicaldrive input command based on example 4:
ipssend create 1 logicaldrive a 100 5 1 0 1 1 1 1 2. The last 1 0 1 1 1 2 refers to the corresponding {channel sid}
2.delete - This command deletes an array that already exists. At the same time, the data on the logical drive will be lost.
Note: This command cannot delete the logical drive of RAID level-x0
Command format : IPSSEND DELETE controller ARRAY arrayed
l controller refers to the ID number of the RAID controller (1-12)
l arrayID is the array that exists (A-H)
Example: (Assuming controller is 1 and arrayID is a)
ipssend delete 1 array a
3. devinfo - This command lists the status and size of the physical drive.
Command format: IPSSEND DEVINFO controller channel sid
l controller refers to the ID number of the RAID controller (1-12)
l channel refers to SCSI channel (1-4)
l SID refers to SCSI ID number (0-15)
For example: ipssend devinfo 1 1 0
It is shown as follows:
Found 1 IBM ServeRAID controller(s).
Device information has been initiated for controller 1...
Device is a Hard disk
Channel : 1
SCSI ID : 0
PFA (Yes/No) : No
State : Ready (RDY)
Size (in MB)/(in sectors): 34715/71096368
Device ID : IBM-ESXSST336732B84G3ET0YAHS
FRU part number : 06P5778
Command completed successfully.
4. drivever - This command lists the vendor ID, firmware version, and serial number of the physical drive.
Command format: IPSSEND DRIVEVER controller channel sid
l controller refers to the ID number of the RAID controller (1-12)
l channel refers to SCSI channel (1-4)
l SID refers to SCSI ID number (0-15)
Ipssend drivever 1 1 0
It is shown as follows:
Found 1 IBM ServeRAID controller(s).
SCSI Inquiry DCDB has been initiated for controller 1...
Device Type : Hard disk
Channel : 1
SCSI ID : 0
Vendor : IBM-ESXS
Revision Level : B84G
Serial Number : 3ET0YAHS
Command completed successfully.
5. getconfig - This command lists information about the controller, logical drive, and physical
Command format: IPSSEND GETCONFIG controller AD/LD/PD/AL
Controller refers to the ID number of the RAID controller (1-12)
l AD displays controller information
l LD displays information about logical drives
l PD displays information about physical devices
l AL displays all the above information
Example: (Default controller is 1)
ipssend getconfig 1 al
6. setconfig - This command changes the configuration of the controller, such as resuming the default value and copying the array information from the hard disk
Command format: IPSSEND SETCONFIG controller DEFAULT/IMPORTDRIVE
Example:
Revert a controller to an exit setting:
ipssend setconfig 1 default
Copy array information from hard disk:
ipssend setconfig 1 importdrive
7.scandrives – scans all hard drives on the controller
Command format: IPSSEND SCANDRIVES controller
l controller refers to the ID number of the RAID controller (1-12)
Usage: (Assuming controller is 1)
ipssend scandrives 1
8. backup - backup array information
Command format: IPSSEND BACKUP controller filename
l controller refers to the ID number of the RAID controller (1-12)
Examples of usage:
ipssend backup 1 backupfile
9. restore--Restore the backed-up array information
Command format: IPSSEND RESTORE controller filename
l controller refers to the ID number of the RAID controller (1-12)
Examples of usage:
ipssend restore 1 backupfile
About IBM's RAID card downgrade BIOS method
This is a program flashman.pro file in the IBM upgrade disk, you need to change the following program to downgrade the RAID BIOS, and use IBM RAID discs to downgrade the RAID BIOS. The way to do this is to download the 4.84 BIOS upgrade first
Program.4.84 BIOS/firmare upgrade disk. The flashman.pro file reads:
ServeRAID family firmware and BIOS download utility profile
Disk Release: 4.84.01
.
Format =
[------ BIOS -------] [---- Firmware -----] [------ Boot -------]
:Adapter Name,Image Name,Rev#,Dsk#,Image Name,Rev#,Dsk#,Image Name,Rev#,Dsk#,
.
-----------------------------------------------------------------------------
.
Type:ServeRAID,A:
.
Unknown Adapter
:?,raid.img,99,1,codeblk.cph,99,2,bootblk.cph,0.00.00,1,
.
Copperhead Adapter
:ServeRAID,raid.img,4.84.01,1,codeblk.cph,2.25.01,2,bootblk.cph,0.00.00,1,
.
ServeRAID on planar image (Navajo)
:ServeRAID1C1,raid.img,4.84.01,1,codeblk.nvj,2.88.13,2,bootblk.nvj,0.00.00,1,
.
Copperhead-Refresh Adapter
:ServeRAID II,raid.img,4.84.01,1,codeblk.rf,2.88.13,2,bootblk.rf,0.00.00,1,
.
Copperhead-Refresh on planer (Kiowa)
:ServeRAID2C2,raid.img,4.84.01,1,codeblk.rf,2.88.13,2,bootblk.rf,0.00.00,1,
.
Clarinet Adapter
:ServeRAID-3H,raid.img,7.84.01,1,codeblk.cln,7.84.01,1,bootblk.cln,0.00.00,1,
.
Clarinet-Lite Adapter (Oboe)
:ServeRAID-3L,raid.img,7.84.01,1,codeblk.cln,7.84.01,1,bootblk.cln,0.00.00,1,
.
Trombone Adapter
:ServeRAID-4H,raid.img,7.84.01,1,codeblk.trb,7.84.01,2,bootblk.trb,0.00.00,1,
.
Morpheus Adapter
:ServeRAID-4M,raid.img,7.84.01,1,codeblk.neo,7.84.01,1,bootblk.mor,0.00.00,1,
.
Morpheus-Lite Adapter
:ServeRAID-4L,raid.img,7.84.01,1,codeblk.neo,7.84.01,1,bootblk.mor,0.00.00,1,
.
Neo Adapter
:ServeRAID-4Mx,raid.img,7.84.01,1,codeblk.neo,7.84.01,1,bootblk.neo,4.84.01,1,
.
Neo-Lite Adapter
:ServeRAID-4Lx,raid.img,7.84.01,1,codeblk.neo,7.84.01,1,bootblk.neo,4.84.01,1,
This method is to change 4lx, raid.img, 4.84.01, 1 (change to 7.84.01, 1), codedblk, neo, 4.84.01, 1 ((change to 7.84.01, 1) and other unchanged, when the BIOS is upgraded, it is found that 6.10 is not high enough to upgrade to the new 7.84 BIOS, and actually generate 4.84. This is called light rise and dark fall.
After restarting, the RAID card will report an error, which is normal, CATL+1 enters the RAID card and initializes again.
It's OK to reiterate.
Use a 4.84 upgrade BIOS disk from the Internet. Open the flashman.pro file in Notepad and change it.
If it falls. BIOS still can't do RAID or the hard disk is broken, connect the SCSI cable of the hard disk backplane to the SCSI interface of the motherboard, CATL+A scan the hard disk to see if it passes evenly, or some OEM hard drives can't make RAID It's too bad, so there is no need to do RAID. Of course, having an original IBM hard drive as a RAID 0 is the best verification.
I'll help you here, the key is up to you to judge for yourself. There are problems
Call me again. I have a lot of RAID discs from RAID 3.0 |
Posted on 2/16/2015 9:14:01 PM
|
|
|
