当前位置:网站首页>Storage, computing and distributed computing (collection and sorting is suitable for Xiaobai)
Storage, computing and distributed computing (collection and sorting is suitable for Xiaobai)
2022-07-25 10:24:00 【Puffiness is also strong】
- Server availability
- BMC Integrated management module (iBMC) Be able to continuously monitor system parameters 、 Trigger alarm , And take recovery measures , In order to avoid downtime to the greatest extent
- Memory module type
- RDIMM
Dual in line memory module with registers ,RDIM A register is added to the memory module for transmission , It is located in CPU And memory particles , Because of the high efficiency of registers , So the capacity and frequency are easier to improve ; Maximum capacity :32G, The frequency is the highest :3200MT/s
- UDIMM
Unbuffered dual inline memory module , Its address and control signals do not pass through the buffer , Directly to DIMM Upper DRAM chip , So the delay is smaller at the same frequency , And because of the manufacturing process , therefore UDIM Low capacity and frequency ; Maximum capacity 4G, The frequency is the highest :2133MT/s
- LRDIMM
Low load dual in line memory module , Simple buffer , Buffering reduces the power load on the lower motherboard , No effect on performance , Reduced memory bus load , Further improve the memory support capacity ; Maximum capacity :64G, frequency :3200MT/s
- RDIMM
- Hard disk interface type
- SAS
Serial connection SCSI, Enterprise application disk , The fastest mechanical hard disk ,SAS Hard disk compatible SATA Interface ,SAS Hard disk pays more attention to reliability ,10000/15000RPM( turn / branch ), Transmission rate 300MB/s, In general, it is based on 600G To calculate the storage capacity ( Non written )
- SATA
SATA Serial hard disk , For ordinary users , It is the mainstream of computer mechanical hard disk ,SATA Hard disk is not compatible SAS Interface ,SATA Hard disk tends to be cheap and high-capacity ,7200RPM( turn / branch ), The transmission rate is 150MB/S, In general, it is based on 200G To calculate the capacity ( Non written )
- NVMe
Use NVMe Protocol SSD ,NVme Protocol nonvolatile memory host controller interface , Is a logical interface specification ; The transmission rate is 600MB/s, In general, it is based on 960G To calculate the capacity ( Non written )
- RAID
Redundant array of independent disks- Raid0
At least two hard disks are required , Split the data into two hard disks , Read at the same time when reading , Write while writing , Improve the continuous reading and writing ability of the disk , When one of the disks is damaged , Data loss
Calculation formula :N Fast disk composition , The logical capacity is N Sum of disk capacity
- Raid1
At least two hard disks are required , Copy a complete data to two hard disks , This array is the safest , Space utilization and speed are the lowest
Calculation formula : Two disks make up , The logical capacity is N Half of the sum of block disk capacity
- Raid5
You need at least three hard drives , Each hard disk stores a part of the recovery code , The disaster tolerance capability of disk is increased , In performance with raid0 near , Disaster tolerance capability is a disk
Calculation formula :N It's made up of blocks , The logical capacity is N-1 The sum of the capacities of the disks
- Raid6
At least four hard drives are required , And raid5 identical , Each disk stores two copies of the recovery code , The disaster tolerance ability has been improved from one to two hard disks
Calculation formula :N Fast disk composition , The logical capacity is N-2 The sum of the capacities of the disks
- Raid50
Need at least 9 Block disk , First, make the disk into three pieces Raid5, Then three more Raid5 Combine into one Raid0, It has higher disaster tolerance capability
Calculation formula : If each RAID5 from N It's made up of blocks , share M individual RAID5 Make up the RAID50, Then the logical capacity is (N-1)*M Sum of disk capacity .
- Raid60
Need at least 8 Block hard disk , Every four hard disks are combined into two Raid6, Two more Raid6 Form a Raid0, It has higher disaster tolerance ability , And higher reading and writing speed , The disaster tolerance capacity is two hard disks
Calculation formula : Fake each RAID6 from N It's made up of blocks , share M individual RAID6 Make up the RAID60, Then the logical capacity is (N-2)*M Sum of disk capacity .
- Raid-tp
stay raid6 The reliability and safety are increased on the basis of , Every CKG( Logical block group ) The verification data in occupies three CHUNK( Logic block ) Space , Allow any three data CHUNK( Logic block ) fault . If there are four or more CHUNK( Logic block ) Simultaneous failure , Then the whole RAID Level failure
- Raid0
- LUN virtualization
LUN Virtualization can integrate multiple Raid Groups are divided into logical units , And mapped to multiple hosts ,LUN Virtualization can achieve cross Raid Group
- Block virtualization (Raid2.0)
Cut the hard disk into multiple logical blocks , Then the logic blocks are composed LUN( Each logical block comes from a different hard disk ), be based on LUN do Raid Group , each LUN It's just one. Raid Group
- The upper LUN virtualization + Bottom block virtualization (Raid2.0+)
Divide the hard disk into several logical blocks of the same size (chunk):CK It's usually 64M, Some of them are hot spare blocks ; The rest of the logic blocks are data blocks , The hard disk domain only supports the same type of hard disk. Each hard disk has a part of data blocks combined into a data block group, which is also called a logical block group (chunk group):CKG, From different hard drives , Then quickly group the data Raid, then Raid Group progress LUN Virtualization is also known as EXCENT:EXCENT It's usually 4MB
- Raid2.0+ Fast refactoring
because Raid2.0+ characteristic , After a single hard disk fails , Will be idle CKG Medium CK Traditional reconstruction time used to reconstruct fault data :30M/s;Raid2.0+ Fast refactoring time :600M/s, If the fault disk is allocated CK Less , Refactoring is faster
- SAS
- The Internet
- Ethernet cable
The transmission distance of the network cable cannot exceed 100 rice , Effective transmission distance 90 rice , Maximum through repeater 500 rice- GE Electric port
10 Gigabit network cable interface
- Five category
Class V network cable 100Mbps( Hundred trillion ) Transmission rate ,CAT.5
- Super five categories
Super class 5 network cable barely supports 1000Mbps Transmission rate ,CAT.5e
- Six categories
Category VI network cable 250Mhz Transmission frequency 1000Mbps( Gigabit ) Transmission rate ,CAT.6
- Super six categories
Cat 6 network cable 500Mhz Transmission frequency ,1000Mbps( Gigabit ) Transmission rate ,CAT.6a
- Seven category
Class VII network cable 500Mhz Transmission frequency ,10Gbps( Ten thousand trillion ) Transmission rate ,CAT.7
- GE Electric port
- Optical fiber
- GE Smooth mouth
10 Gigabit fiber interface
- Fiber optic interface
- FC type
- SC type
- LC type
- ST type
- Optical module package type
SFP、XFP、XENPAK、BIDI SFP、GBIC、ESFP( common SFP、SFP+、QSFP、QSFP+)- SFP
Small pluggable optical module , The maximum transmission rate shall not exceed 5Gbps within
- SFP+
Enhanced SFP modular , The transmission rate is 10Gbps
- QSFP
Four channel small pluggable module , The transmission rate is 5Gbps
- QSFP+
Enhanced four channel small pluggable module , The transmission rate is 40Gbps
- SFP
- Optical module
- Monomode SM
1310nm, Large transmission loss , But the dispersion is small , Generally used in 40KM Transmission within
1550nm, Low transmission loss , But the dispersion is large , Generally used in 40KM Transmission beyond
- multimode MM
850nm, Low cost, short transmission distance , Generally only support 500M
- CWDM
A low-cost solution for metropolitan area network access layer WDM Transmission technology , It is mainly used in Enterprise Networks , Campus network
- DWDM
A laser technology used to improve the bandwidth of the existing optical backbone network , Long distance trunk network , Core node of large capacity metropolitan area network
- Monomode SM
- GE Smooth mouth
- Ethernet cable
- PCIe
High speed serial computer extended bus standard- PCIe edition
Different versions PCIe The version transfer rate is different- 1.0
Transmission rate :2.5GT/s
- 2.0
Transmission rate :5GT/s
- 3.0
Transmission rate :8GT/s
- 4.0
Transmission rate :16GT/s
- 5.0
Transmission rate :32GT/s
- 1.0
- PCIe Interface classification
- x16
Universal slot , Usually a graphics card or AI Accelerator card ,AI Reasoning cards and Raid Display cards, etc
- x8
Usually processed as X16 The style of , But only half of the stitches are effective , For graphics card or AI Accelerator card ,AI Reasoning cards and Raid Display card
- x4
Usually solid state disk interface
- x1
Usually independent network card , Independent sound card ,USB3.0 Expansion card
- x16
- PCIe edition
- Power Supply
- AC
Alternating current (ac)
- DC
Direct current
- AC
边栏推荐
猜你喜欢
随机推荐
VLAN的配置及其应用(以华为eNSP为例)
Principle of struct2
Redis使用场景
将 conda 虚拟环境 env 加入 jupyter kernel
For cycle: daffodil case
Angr(四)——angr_ctf
Bug分类和定级
Angr (I) - Installation
Duplicate SSL_ Anti spoofing, spoofing attacks and deep forgery detection using wav2vec 2.0 and data enhanced automatic speaker authentication
存储、计算、分布式计算篇(收集整理适合小白)
1、 The difference between unittest framework and pytest framework
Pytorch 张量列表转换为张量 List of Tensor to Tensor 使用 torch.stack()
一、unittest框架和pytest框架的区别
MySQL solves the problem of not supporting Chinese
Ansible Deployment Guide
js加密参数定位
【无标题】
Trojaning Attack on Neural Networks 论文阅读笔记
数论---最大公约数最小公倍数
Pow(x,n)









