HCIA Storage EXPERT Reveals Cascade Module Secrets
[Music] hello everyone in this lesson we will continue to learn about hard disk components hard disk types can be classified according to interface type size application scenario hard disk implementation and other dimensions the size can be divided into 1.8 in 2.5 in 3.5 in 5.25 in ETC 3.5 and 5.25 are the sizes of early floppy discs um the floppy disc is the earliest removable medium used in personal computers the 5.2 5in floppy disc was the most popular
product in the floppy disc era according to the implementation of the hard disk it can be divided into mechanical hard drives and solid state drives the structure of a mechanical hard drive is roughly divided into platters magnetic head walls red right heads spindles hard disk interfaces and controlled circuits Etc the platter is covered with Magnetic material magnetic particles on the dis surface are polarized to represent a binary unit of information or a bit and our read WR head is responsible for reading or writing data to the platter storage is achieved by changing the polarity of the extremely fine magnetic particles on the surface of the dis through the head and our magnetic head wall it can drive the re right head to move the head to the specified position components related to the spindle the motor and bearing are underneath it it can rotate the platter move the specified position of the platter under the read right head the control circuit is used to control the speed of the dis movement of the magnetic head wall issue commands to the Head Etc each platter on the hard dis has two read right heads used for reading and writing data on both surfaces of the plateau the head floats on the plateau using air flow and does not touch the plateau therefore it can move quickly between tracks if the head is too far from the platter then the red signal will be too weak if the distance is too low it will wear the surface of the platter therefore the surface of the platter must be very smooth and flat any foreign objects and dust will cause the head to rub against the magnetic surface and cause permanent damage to the data as for their work working principles initially our read right head docked in a special area near the spindle of the platter we call it the start stop area the spindle connects all platters and is connected to a motor the motor of the spindle rotates at a constant speed driving the platter to rotate when the spindle rotates there is a very small air gap between the red/ right head and the platter called The Flying height of the head the re right head is installed at the top of the head arm the head arm moves the head to above the position of the plateau that needs to be written or taken out the head reads and writes data on the hard disk surface in binary form the read data is stored in the flash chip of the hard disk finally passed to the program to run each platter of the hard disk has two surfaces each surface can store data becoming an effective platter each effective surface has a surface number in order from top to bottom numbered from zero in order in the hard disk system the surface number is also called the head number therefore each effective surface has a corresponding read right head as shown in the figure the track is also called track the track is a concentric ring around the spindle on the plateau data is recorded on the track the track from the outermost ring to the inner ring numbered in order from zero each surface of the hard disk has 300 to 1024 tracks the number of tracks for surface of the new large capacity hard disk will be more usually the number of tracks per inch on the platter to measure the compactness of the track Arrangement on the surface the track is invisible to the naked eye it's just some magnetized areas on the surface that are magnetized in a special form cylinder in the same hard disk all tracks with the same number on all platters form a cylinder from top to bottom call the cylinder of the hard disk the head under each cylinder numbered from zero from top to bottom data reading and writing is done by cylinder that is when the head reads and writes data first start reading and writing operations from the zero head in the same cylinder then proceed in order on different surfaces in the same cylinder only when all the heads in the same cylinder have completed reading and writing the head will move to the next cylinder therefore because selecting a head only needs to be switched electronically and selecting a cylinder must be switched by machine and seeking usually the position of the head in the hard disk is explained by the cylinder number instead of using the track number to explain sector each track it will be divided into smaller units called a sector the purpose of dividing sectors is to make data storage more organized the sector is the smallest storage unit that can be individually sought in the hard disk different hard diss can have different numbers of sectors per track usually a sector can save 512 bytes of user data but some hard diss can be formatted to larger sector sizes such as 4 KB sectors hard disk capacity equals the number of cylinders times the number of heads times the number of sectors then multiplied by 500 12b they can be divided into single stack or multi-stack the former only has one platter in the dis body the latter has multiple but the hard dis only allows one head to read data at any moment therefore no matter how many platters and heads it cannot improve the throughput and IO performance of the hard disk it can only increase capacity for factors affecting hard disk performance the first is rotation speed the rotation speed refers to the number of rotations of the hard disk platter the unit is rpm when reading and writing data the head will not move rely on the rotation of the plate to sense and read data there for the faster the platter turns the shorter the data transmission time in the case of continuous iio the number of seek words for the magnetic head wall is small therefore to improve throughput or iops value rotation speed is the first influencing Factor seek speed in our random IO situation the magnetic head wall needs to frequently change tracks the time used for data transmission compared to the time to change is very small therefore if the head must be able to change track at a high speed it will increase the random iops value single butterfly capacity the higher the capacity of a single disc prove that the larger the data volume in the same space and the data density is larger under the same rotation speed or seek speed conditions hard drives with high data density will show higher performance interface speed the current interface speed theoretically already meets the highest external transmission bandwidth that the hard disk can reach in a random I/O environment the interface speed is even less important because the bottleneck is almost on the seek speed the average access time of the hard disk is composed of the following two items the first is the average seek time it refers to the movement of the hard disk's red/ right head from the initial position to the disk surface the time required to reach a specified track is an important parameter affecting the internal data transfer rate of the hard disk the smaller this time the better the average waiting time is when the read right head is already on the track to be accessed the time waiting for the sector to be accessed to rotate under the read right head the average waiting time is usually half of the time required for the disk to rotate once therefore the faster the hard disk speed the shorter the waiting time the data transfer rate of the hard disk refers to the speed of reading and writing data from the hard disk data transfer rate includes external data transfer rate and internal data transfer rate the unit can be characterized as mb per second the internal transfer rate is also known as the sustained transfer rate refers to the maximum speed of the read i r head when reading and writing to the hard disk it does not include the impact of the time consumed by seeking and waiting for the sector to rotate to the read right head it is an ideal situation that is is it is assumed that the Read's right head does not need to switch tracks when reading and writing nor does it specifically read a certain sector only continuously Loop read and write all sectors of this track on one track the speed at this time is called the internal transfer rate of the disk the external transfer rate is also known as the burst data transfer rate or interface transfer rate it refers to the data transfer rate between the system bus and the hard disk cache it is related to the type of hard disk interface and the size of the hard disk cache the hard diss iops refers to the input/output volume per second it is one of the main indicators to measure hard disk performance it is composed of seek time rotation delay and data transfer time the shorter the seek time the faster the I/O operation the average seek time of current hard drives is generally between 3 and 15 milliseconds rotation delay refers to the time required for the dis to rotate the data sector to the read/ right head below rotation delay depends on the speed of the hard disk it is usually expressed as half of the time required for the hard disk to rotate once for a 7,200 revolutions per minute hard drive its average rotation time delay is 60 * 1,000 divided by 7,200 and then multiplied by 1/2 this data is approximately 4.17 milliseconds for a half hard disk with a speed of 15,000 revolutions per minute its average rotation delay is generally 2 milliseconds data transfer time refers to the time required to complete the transfer of the requested data it depends on the data transfer rate its value equals the data size divided by the data transfer rate in the case of random IO the read right head needs to frequently change tracks the time used for data transfer is much less than the time consumed by changing tracks they are not on the same order of magnitude therefore the data transfer time can be ignored in the transfer mode of the hard disk it can be divided into parallel transfer and serial transfer parallel transfer has high transfer efficiency but the transfer distance is not long the transfer frequency is also not high using multiple lines for long-distance transmission is more expensive than using a single line longdistance transmission requires thicker wires to reduce signal attention but bundling them into a single cable is quite difficult long-distance transmission is affected by wire resistance and other factors the data of each line arrives at the other party at different times this leads to the receiver having to wait for all the digits on the eight lines to arrive before initiating the next round of transmission serial transmission serial transmission is obviously much less efficient than parallel transmission but the transmission speed can be improved by increasing the transmission frequency overall serial transmission is actually faster than parallel transmission serial transmission is used for longdistance transmission currently interfaces like PCI have switched to serial transmission the pcie interface is a typical serial transmission method uh it single line transmission rate is as high as 2.5 GB per second hard drives can be divided into IDE SCSI SATA SAS FC hard drives Etc this method is quite common but the difference between hard drives under the classification is not just the interface there are also mechanical chassis IDE and SATA hard drives used the ATA series mechanical chassis suitable for single task processing SCSI SATA FC hard drives use the SCSI series mechanical chassis can better meet the needs of server multitasking ATA series hard drives facing large data through put Services the speed is not as good as SCSI series hard drives the reliability of SCSI series mechanical chassis is much higher than that of ATS series systems the at interface has been developed to this day several different versions have been released they are ata1 to ata7 the at interface has the following advantages and disadvantages its Advantage is that its price is relatively low compatibility is good its disadvantage is that the speed is relatively slow can only be used internally there are also very strict restrictions on the length of the interface cable in today's large Enterprises the current transmission efficiency of ATA can no longer meet the needs of users SATA is a Serial aaat uses a Serial method for data transmission the interface rate is higher than the ID interface SATA hard drives use a point-to-point connection method supports hot swapping Plug and Play SCSI appeared mainly because the original ID interface the hard disk speed is too slow the transfer rate is too low therefore highp speed SCSI hard drives appeared in fact SCSI is not specifically designed for hard drives in fact it is a bus type interface working on a bus independent of the system its Advantage is that it has a wide range and anti-interference ability can achieve a longer connection distance the disadvantage is that the price of SAS interface hard drives is also relatively High FC and the CSI interfaces are the same it was not originally a interface technology developed for hard drives but was specifically designed for networks but with the storage systems demand for Speed it was gradually applied to the hard dis system its Advantage is that it has a high bandwidth the disadvantage is that it is very expensive next let's look at the related knowledge of solid state drives actually for ssds it does not use polar materials to store data like traditional hard drives instead it uses a basic unit called a cell of nand flash to store data nand flash is a non-conscious random access storage medium its characteristic is that data does not disappear after exercise this technology can store data information quickly and compactly SSD has no high-speed rotating Parts it's performance is high power consumption and no noise and it has no internal mechanical components but this does not mean that their life span is infinite because the snan flash is a non-conscious medium the original data must must be erased before writing new data then write data but the number of erases of each cell is limited so when the number of erased data reaches the limit then this cell can no longer continue to read or write data SSD is mainly composed of control unit and storage unit currently it is mainly composed of flash memory particles the control unit includes SSD controller host interface drum Etc the storage unit includes nand Flash particles the host interface includes the protocol and physical interface for the host to access the SSD such as SATA SAS and pcie Etc the SSD controller is responsible for the read and write access and protocol conversion from the host to the backend medium surface management data caching and transactions are the core components next let's look at the storage principle of nand flash nand flash it uses floating sand transistors to store data it's working principle is to use whether there is charge stored on the floating sand or the amount of stored charge to change the voltage of the transistor the characterization of data information is realized by reading the preal voltage of the transistor the internal storage unit of nand flash includes lwn plane block page cell configuration and cell among them Aon is the smallest physical unit that can be independently encapsulated usually contains multiple plant plane has independent configuration registers block is the smallest unit that can perform eral usually composed of multiple configurations and the configuration is the smallest unit that can perform programming and reading usually the size is 16 KB cell is the smallest operation insert read unit in the configuration it corresponds to a floating sand transistor usually it can store one bit or multiple bits for us the basic operation unit of programming and reading is configuration and the basic unit of eraser is a block the Eraser process will cause a certain degree of damage to the insulating layer of the floating sand transistor after a ratier failure or programming failure the operating block will be set as cycle manufacturers Define that when the number of Cycles reaches a certain level for example within 4% it is considered that our n flash has reached the upper limit of its service life Nan flash particles can be divided into the following four categories according to the different bid data stored in the cell the first is SLC it is a single layer storage unit a cell can store one bit of data zero or 1 mlc is a multi-layer storage unit in mlc a Cell can store Two Bits of data such as 00 0 01 10 and 11 TLC TLC is a three layer storage unit in TLC a cell can store three bits of data as shown in the figure and qlc it is a four layer as the name suggests it can store four bits of data four different types of cells although the amount of data stored is different but the cost is roughly the same this is also a reason for the continuous increase in SSD capacity the original SD only has 64 GB or smaller and now the largest TLC SSD disc can reach a data of 2 tab however different types of SSD discs have different wear resistance capabilities resulting in different reliability of the hard disk the wear resistance of the SSD disk is also an important parameter for choosing an SSD dis as shown in the figure take a logic diagram of a flash chip as an example each of 4,314 * 8 that is 3 4 5 one two cells can logically form a configuration each configuration can store 4 KB of content and 218b of a transaction data configuration is also the small smallest unit of I/O in the flash ship every 128 configurations can form a block every 2048 blocks can form a region plane a whole flash ship is composed of two regions one region stores odd numbered blocks the other stores even numbered blocks two blocks can operate in parallel next let's look at the storage address mapping management first of all LBA is a logical block address can point to the address of a data block or the data block pointed to by a certain address PB is a physical block address the Project's host host accesses the SSD through LBA each LBA represents a sector that is a sector the operating system generally accesses the SSD in units of 4K the basic unit of host accessing SSD is called a user page that is the configuration of the host inside the SSD the SSD master control and Flash are accessed through Flash configur ation as the basic unit to access flash call flash configuration as physical page that is physical page every time the host writes a host configuration the SSD Master Control will find a physical page write the data of the host the SSD internally also enters such a mapping with such a mapping relationship next time the host needs to read a certain host configuration the SSD knows where to read the data from The Flash SSD knows how to read data from which position on flash as mentioned before the read and write unit of flash memory is the page and the size of the page is either 4 KB or 8 KB but our operating system reads and writes data it is done according to the sector size of htd that is 512 bytes the most Troublesome is that flash memory is removed and fast as a unit and it can't be written in for removal this causes the file system currently used by the operating system can't manage s at all you need to replace it with a more advanced and complex file to solve this problem but this will increase the burden on the operating system in order not to increase the burden on the operating system SST user software virtualize the operation of flash memory into independent sector operations of the disk this is FTL because FTL exists between the file system and the physical medium flash memory the operating system only needs to operate lb as before all the tone table work from LB to VA all are handed over to FTL to be responsible it is because of FTL NAD flash can be used as a hard disk the file system can directly use SSD as a common fast device because FTL is the most important core technology in SSD design manufacturers so no manufacturer is willing to disclose this technical information and there has been no technical specification standard in the industry the importance of FTL lies in determining the service life of an SST performance and reliability once there is a problem with FDL it will cause data read and write errors what's more serious is that the SSD dis cannot be accessed excellent FTL can not only extend the life of flash storage but also optimize read and write performance take eight channels as an example analyze how our host reads and writes SST the master control connects eight flash bands through eight channels for easy explanation only one block is drawn here for each band each small square represents a configuration assume the size is 4 KB the host writes 4 KB of data write a 4kb to the block of Channel Zero occupies a configuration if the host continues to write 16 KB of data will'll write 4kb in the block of Channel 14 the host continues to write in the end the entire block will be filled when all the blocks on the channel are full the SSD Master Control will select the next block right in the same order with the continuous writing of the host flash storage space gradually becomes smaller until it is exhausted if these garbage data are not cleared in time our host can't ride again whether our speed is increased by eight times depends on the data to be read is it evenly distributed in the block of each Channel like 32 KB of data stored in the block of Channel 14 in the picture then the read can only be increased by up to four times this is why the smaller the file the lower the transmission rate SSD performance has the following advantages the first is short response time the mechanical characteristics of traditional hard diss most of the time is wasted on seeking or mechanical delay data transmission efficiency is severely restricted there are no mechanical moving parts inside the SST hard drive eliminate the seek time and mechanical delay can respond to read and write requests more quickly High read and WR efficiency mechanical hard drives perform random read and write operations the head keeps moving causing very low read and ride efficiency SSD calculates the storage location of data through the internal controller and perform read and right operations eliminate the mechanical operation time greatly improve the read and right efficiency the advantage of less hard disk power consumption is not very obvious when there are a large number of hard diss the cost difference caused by power consumption is relatively large this has also become a factor considered by Enterprises when choosing a Sol solution SSD storage applications can follow two principles the first is the two eight principle you module G electrical interface module provides four electrical ports with a transmission rate of 1 gbit per second can be used for dual active materials 40G interface module provides two Optical ports with a transmission rate of 40 gbit mainly used for the connection of storage devices and the line connection of the application server 100ge interface module provides two two Optical ports with a transmission rate of 100 JB it mainly used for the connection between storage devices and application servers for SAS node modules and RDMA interface modules 25gb RDMA interface module provides four Optical ports with a transmission rate of 25 gbit per second used for the connection between control frames in direct connection networking and the 100gb RDMA interface module provides two interfaces with a rate of gbit used for the connection between the control frame and the switch or the connection between the control frame and the smart Hardware frame the S so in the label represents scale Out means horizontal expansion and the B label represents the backend 12 GB SAS node module provides four transmission rates of four X12 JB per second mini SAS HD node Port through the no Port connect the control frame and the Rus SAS hard disk frame is the connection point for data transmission between the control frame and the ru SI frame smart interface module supports 8 jbit 10 jbit 16 gbit 25 gbit and 32 gbit per second five Optical modules mainly used for the connection between storage devices and application servers the speed of the optical module needs to be consistent with the rate on the interface module label otherwise the storage system will alarm pcie interface module provides two pcie interfaces the speed of the tags is consistent otherwise the storage system will be in urgent need the pcie interface module provides two PCI interfaces it is the business interface between the control frame and the data switch used for exchanging control flow and data flow information between control frames as shown in the figure there are the following indicator lights module power indicator light pcie Port indicator light a handle of the module and the related pcie port for the 56gb IB interface module provides two transmission rates of 4 x14 gbi per second IB ports 16GB FC interface module provides two physical interfaces converted into 816 GB FC ports through a dedicated line bar each fiber Port has a transmission rate of 16 GB per second it is the business interface between the application server and the storage system used to receive data exchange commands issued by the application server 10gb FC interface module provides two ports with a transmission rate of 10 gby per second it is the business interface between the application server and the storage system used to receive data read and WR instructions issued by the application server currently the 16 GBF Coe interface module only supports direct connection networking finally let's look at two thinking questions you can pause the video for a moment to think finally here is our summary in this chapter we learned about the introduction of intelligent storage components including control frame hard disk frame and related interface modules among them hard drives can be divided into mechanical hard drives and solid state drives if you have any questions after class you can also download our related AP or log into our related learning website to get more information see you in the next lesson
2025-01-07 16:55
