an overview on solid state drives architectures and
play

An overview on solid-state-drives architectures and enterprise - PowerPoint PPT Presentation

Oct 31, 2023 •225 likes •1.34k views

An overview on solid-state-drives architectures and enterprise solutions Romolo Marotta Sapienza, University of Rome Why SSD are so attractive? Capacity vs Access Time SRAM 10 8 DRAM FLASH 10 6 Capacity (MB) 10 4 10 2 1 10 -2 10 -6 10 -4

  1. How does a flash memory cell work? How to read a cell? • Reading a cell consists in V I inferring on the cell state CG • Apply an intermediate SOURCE DRAIN FG STATE=? voltage V 1 < V I < V 0 on N+ N+ CG P-well • Read the actual value I ∗ D of the current I D I D V 1 V I V I V 0 V CG 0 STATE 1 STATE 0 10 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  2. How does a flash memory cell work? How to read a cell? • Reading a cell consists in V I inferring on the cell state CG • Apply an intermediate SOURCE DRAIN _ _ _ _ _ _ _ _ FG voltage V 1 < V I < V 0 on N+ N+ CG P-well • Read the actual value I ∗ D of the current I D • If I ∗ D = 0 the bit value is 0 I D V 1 V I V I V I V 0 V CG 0 STATE 1 STATE 0 10 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  3. How does a flash memory cell work? How to read a cell? • Reading a cell consists in V I inferring on the cell state CG • Apply an intermediate SOURCE DRAIN FG voltage V 1 < V I < V 0 on N+ N+ CG P-well • Read the actual value I ∗ D of the current I D • If I ∗ D = 0 the bit value is 0 I D • If I ∗ D � = 0 the bit value is 1 V 1 V I V I V I V I V 0 V CG 0 STATE 1 STATE 0 10 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  4. Multi-Level Cell • If we partially charge FG, we need a lower threshold voltage for creating a channel • We store 2 bits by using 1 programmed state, 2 partially programmed states and 1 erased state • A flash cell storing multiple bits is a Multi-Level Cell (MLC) • A Triple-Level Cell (TLC) stores 3 bits _ _ _ _ _ _ _ _ _ _ _ _ _ _ ERASED PROGRAMMED PARTIALL Y PROGRAMMED 11 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  5. Why does a flash cell deteriorate in time? • writing a flash cell involves an erase and a program ⇒ electrons move from/into FG anode interface (SiO 2 ) < 10 nm electron traps cathode interface 12 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  6. Why does a flash cell deteriorate in time? • writing a flash cell involves an erase and a program ⇒ electrons move from/into FG • electrons collide with and damage the insulating layer creating traps ⇒ a Stress Induced Leakage Current (SILC) can flow through these traps Damaged Oxide anode interface anode interface SILC (SiO 2 ) < 10 nm electron traps cathode interface cathode interface 12 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  7. Why does a flash cell deteriorate in time? • writing a flash cell involves an erase and a program ⇒ electrons move from/into FG • electrons collide with and damage the insulating layer creating traps ⇒ a Stress Induced Leakage Current (SILC) can flow through these traps • a lot of traps can build a path from the body to FG ⇒ electrons can flow through that path ⇒ impossibility to program ⇒ the flash cell is unusable Damaged Oxide Oxide Breakdown anode interface anode interface anode interface breakdown path SILC (SiO 2 ) < 10 nm electron traps cathode interface cathode interface cathode interface 12 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  8. Why does a flash cell deteriorate in time? PE-Cycles • A flash cell can be programmed and erased a limited number of times before a breakdown ⇒ This number is called P/E-Cycles • Vendors design firmware capable of recompute the voltage thresholds for read/write operations ⇒ enterprise-MLC (eMLC) P/E Cycles 100000 30000 10000 5000 SLC eMLC MLC TLC 13 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  9. MLC vs SLC SLC: MLC: • lower density • higher density • higher cost • lower cost • faster write • erase time is similar to SLC • faster read • the level of charges in FG has to be set carefully ⇒ slower program ⇒ • higher endurance slower write • state is not 0/1 ⇒ slower read • eMLC has 3x shorter endurance • MLC has 10x shorter endurance • TLC has 20x shorter endurance 14 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  10. How are flash chips organized? Flash Cell 1 bit 15 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  11. How are flash chips organized? Flash Cell Page 1 bit 16384 + 512 bits 15 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  12. How are flash chips organized? Flash Cell Page 1 bit 16384 + 512 bits Additional bits are used to store ECC and recover from runtime read errors 15 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  13. How are flash chips organized? Flash Cell Page Block 1 bit 16384 + 512 bits 64 pages= 128KB + 4KB Additional bits are used to store ECC and recover from runtime read errors 15 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  14. How are flash chips organized? Flash Cell Page Block 1 bit 16384 + 512 bits 64 pages= 128KB + 4KB Additional bits are used to store ECC and recover from runtime read errors Some bits are used mark the block as faulty 15 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  15. How are flash chips organized? Flash Cell Page Block 1 bit 16384 + 512 bits Plane 64 pages= 128KB + 4KB Additional bits are used to store ECC and recover from 2048 blocks = 256MB + 8MB runtime read errors Some bits are used mark the block as faulty 15 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  16. How are flash chips organized? Flash Cell Page Block 1 bit 16384 + 512 bits Plane 64 pages= 128KB + 4KB Additional bits are used to store ECC Die and recover from 2048 blocks = 256MB + 8MB runtime read errors Some bits are used 2 planes = 512MB + 16MB mark the block as faulty 15 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  17. How are flash chips organized? Flash Cell Page Block 1 bit 16384 + 512 bits Plane 64 pages= 128KB + 4KB Additional bits are used to store ECC Die and recover from 2048 blocks = 256MB + 8MB Chip runtime read errors Some bits are used 2 planes = 512MB + 16MB mark the block as faulty 4 dies = 2048MB + 64MB 15 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  18. How are flash cells organized? • Flash cells are connected forming an array called string • According to the strategies used to connect multiple cells, we can distinguish at least two kind of configuration: NAND NOR Flash cells are connected in Flash cells are connected in parallel, resembling a NOR gate series, resembling a NAND gate V cc V cc V out V out A B A B 16 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  19. How are flash cells organized? • Let F be the CG side length NOR: NAND: • occupies area 10 F 2 • occupies area 4 F 2 • read/write a single cell • read/write a single page • erase a single block NOR ARCHITECTURE NAND ARCHITECTURE Bit 1 Bit 2 Bit 3 Bit 1 Bit 2 Bit 3 Page 1 Select Gate 1 Page 1 Page 2 Page 2 Page 3 Page 3 Page 8 Select Gate 2 Source For One Block 17 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  20. NOR vs NAND NOR: NAND: • fast random-byte read • no random-byte read • slower page read • slow partial page read when supported • slower write • faster page read • lower density • faster page write ⇒ good for source code • higher density ⇒ good for storage We focus on NAND flash technology 18 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  21. How is a page written? • Write-in-place strategy: 1. read the block 2. erase the block 3. program the block with the updated page 19 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  22. How is a page written? • Write-in-place strategy: 1. read the block 2. erase the block 3. program the block with the updated page • 1 page write = N page read + 1 block erase + N page write ( N = number of pages in a block) ⇒ very slow write 19 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  23. How is a page written? • Write-in-place strategy: 1. read the block 2. erase the block 3. program the block with the updated page • 1 page write = N page read + 1 block erase + N page write ( N = number of pages in a block) ⇒ very slow write ⇒ If we update the page 40 times per second (every 25ms), the block is completely broken in: 10 5 PECycles ◦ SLC = UpdateRate = 40 ps ≈ 2500 s ≈ 40 m 10 4 ◦ MLC = 40 ps ≈ 4 m ◦ TLC = 5 · 10 3 40 ps ≈ 2 m 19 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  24. How is a page written? • Write-in-place strategy: 1. read the block 2. erase the block 3. program the block with the updated page • 1 page write = N page read + 1 block erase + N page write ( N = number of pages in a block) ⇒ very slow write ⇒ If we update the page 40 times per second (every 25ms), the block is completely broken in: 10 5 PECycles ◦ SLC = UpdateRate = 40 ps ≈ 2500 s ≈ 40 m 10 4 ◦ MLC = 40 ps ≈ 4 m ◦ TLC = 5 · 10 3 40 ps ≈ 2 m ALERT! In our example the write rate is 80 KBps 19 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  25. Write Amplification • Write amplification occurs when 1 user page write leads to multiple flash writes • Write amplification make flash blocks deteriorate faster • Let F be the number of flash writes corresponding to U user writes ⇒ The write amplification A is: A = F + U = 1 + F U = 1 + A f U where A f is the write amplification factor 20 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  26. How is a page written? Relocation-on-write • Write-in-place is inadequate in terms of reliability and performance ( A f ≈ # number of pages in a block) 0 1 2 3 4 5 6 7 8 9 Page Id Operating System's view of SSD SSD 0 1 2 3 4 5 6 7 8 9 Physical Page Id 21 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  27. How is a page written? Relocation-on-write • Write-in-place is inadequate in terms of reliability and performance ( A f ≈ # number of pages in a block) • Updated pages are re-written on new locations 3 User Write: 0 1 2 3 4 5 6 7 8 9 Page Id Operating System's view of SSD SSD 0 1 2 3 4 5 6 7 8 9 Physical Page Id 21 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  28. How is a page written? Relocation-on-write • Write-in-place is inadequate in terms of reliability and performance ( A f ≈ # number of pages in a block) • Updated pages are re-written on new locations • The logical address of the update page is mapped to a different physical page 3 3 User Write: User Write: 0 1 2 3 4 5 6 7 8 9 Page Id Operating System's view of SSD SSD 0 1 2 3 4 5 6 7 8 9 Physical Page Id 21 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  29. How is a page written? Relocation-on-write • Write-in-place is inadequate in terms of reliability and performance ( A f ≈ # number of pages in a block) • Updated pages are re-written on new locations • The logical address of the update page is mapped to a different physical page • Previous pages are invalidated 3 3 3 User Write: User Write: User Write: 0 1 2 3 4 5 6 7 8 9 Page Id Operating System's view of SSD x SSD 0 1 2 3 4 5 6 7 8 9 Physical Page Id 21 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  30. How is a page written? Relocation-on-write • Write-in-place is inadequate in terms of reliability and performance ( A f ≈ # number of pages in a block) • Updated pages are re-written on new locations • The logical address of the update page is mapped to a different physical page • Previous pages are invalidated ⇒ 1 user page write = 1 page read (obtain an empty page) + 2 page write (update data + invalidate page) ⇒ faster write 3 3 3 User Write: User Write: User Write: 0 1 2 3 4 5 6 7 8 9 Page Id Operating System's view of SSD x SSD 0 1 2 3 4 5 6 7 8 9 Physical Page Id 21 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  31. Flash Translation Layer • Assign Logical Addresses to pages • Store the association between physical and logical addresses in a Translation Mapping Table • Store the number of erase operation performed on physical pages in a Erase Count Table • Tables are: ◦ maintained in SRAM (high efficient) at runtime ◦ stored on flash during shutdown to ensure durability ◦ loaded at boot-up 22 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  32. Wear-Leveling • Free pages for relocation can be retrieved from the whole SSD • Wear-leveling guarantees that the number of PE-Cycles is uniformly distributed among all blocks ⇒ Wear-leveling extends the time to live of each block and the whole SSD • Thanks to wear-leveling all blocks break at the same time 1200 1000 800 Erase 600 Count 400 200 0 0 5000 10000 15000 20000 25000 30000 35000 Block ID 32GB+Free blocks 23 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  33. Wear-Leveling • In order to guarantee that enough free pages are available for write relocation, wear-leveling needs: ◦ Over-provisioning - keep free a percentage of raw capacity ◦ Garbage collection - keep invalid pages in the same block ◦ DRAM buffers - keep valid pages in a buffer in order to write full blocks and reduce fragmentation • We can distinguish at least two kind of wear-leveling algorithms: ◦ Dynamic wear-leveling ◦ Static wear-leveling 24 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  34. Wear-Leveling Dynamic Algorithm • It is called dynamic, because it is executed every time the OS replace a block of data • A small percentage (e.g. 2%) of raw capacity is reserved as free-block pool • It chooses from the free pool the block with minimum erase count the buffer is flushed • The replaced block is erased and added to the free pool ⇒ Only frequently-updated blocks are consumed 25 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  35. Wear-Leveling Static Algorithm • Periodically scan the metadata of each block • Individuate inactive data blocks with lower erase count than free blocks • Copy their content into free-blocks and exchange them ⇒ this guarantees that static blocks participate to wear leveling 26 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  36. Wear-Leveling Impact on reliability At first approximation, wear-leveling eliminate write amplification generated by different sizes of erase and write units ⇒ The block time to fault is: Block TTF ≈ N die · N planes · N blocks · N · PECycles PageWriteRate 27 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  37. Wear-Leveling Impact on reliability At first approximation, wear-leveling eliminate write amplification generated by different sizes of erase and write units ⇒ The block time to fault is: Block TTF ≈ N die · N planes · N blocks · N · PECycles PageWriteRate Block TTF ≈ N die · N planes · N blocks · N · PageSize · PECycles PageWriteRate · PageSize 27 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  38. Wear-Leveling Impact on reliability At first approximation, wear-leveling eliminate write amplification generated by different sizes of erase and write units ⇒ The block time to fault is: Block TTF ≈ N die · N planes · N blocks · N · PECycles PageWriteRate Block TTF ≈ N die · N planes · N blocks · N · PageSize · PECycles PageWriteRate · PageSize Block TTF ≈ Capacity SSD · PECycles WriteRate 27 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  39. Wear-Leveling Impact on reliability At first approximation, wear-leveling eliminate write amplification generated by different sizes of erase and write units ⇒ The block time to fault is: Block TTF ≈ N die · N planes · N blocks · N · PECycles PageWriteRate Block TTF ≈ N die · N planes · N blocks · N · PageSize · PECycles PageWriteRate · PageSize Block TTF ≈ Capacity SSD · PECycles WriteRate • Blocks deteriorate uniformly, thus: Block TTF ≈ SSD TTF 27 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  40. Wear-Leveling Example • Take a SSD with capacity C and a write rate W • According to the flash cells used, we have different time to fault: 28 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  41. Wear-Leveling Example • Take a SSD with capacity C and a write rate W • According to the flash cells used, we have different time to fault: • C = 4 GB , W = 80 KBps = 4 GB · 10 5 ◦ SLC ⇒ SSD TTF = C · PECycles SLC 80 KBps ≈ 158 years W ◦ MLC ⇒ SSD TTF = C · PECycles MLC ≈ 15 . 8 years W ◦ TLC ⇒ SSD TTF = C · PECycles TLC ≈ 7 . 9 years W 28 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  42. Wear-Leveling Example • Take a SSD with capacity C and a write rate W • According to the flash cells used, we have different time to fault: • C = 4 GB , W = 80 KBps = 4 GB · 10 5 ◦ SLC ⇒ SSD TTF = C · PECycles SLC 80 KBps ≈ 158 years W ◦ MLC ⇒ SSD TTF = C · PECycles MLC ≈ 15 . 8 years W ◦ TLC ⇒ SSD TTF = C · PECycles TLC ≈ 7 . 9 years W • C = 128 GB , W = 4 MBps = 128 GB · 10 5 ◦ SLC ⇒ SSD TTF = C · PECycles SLC ≈ 101 years W 4 MBps ◦ MLC ⇒ SSD TTF = C · PECycles MLC ≈ 10 years W ◦ TLC ⇒ SSD TTF = C · PECycles TLC ≈ 5 years W 28 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  43. Wear-Leveling Impact on Reliability 2 As said before wear leveling make flash blocks deteriorate uniformly. Anyhow • Garbage collection increase the number of flash write • Static Wear-leveling increases the number of flash write ⇒ re-introduce write amplification factor SSD TTF ≈ Capacity SSD · PECycles (1 + A f ) WriteRate 29 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  44. SSD Architecture Durable Storage NAND NAND Flash Flash NAND NAND Flash Flash 30 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  45. SSD Architecture Durable Storage NAND NAND Flash Flash Flash Bus NAND NAND Flash Flash 30 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  46. SSD Architecture Durable Storage NAND NAND Flash Flash Flash Bus Flash Controller NAND NAND Flash Flash Read & Write Wear-leveling FTL 30 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  47. SSD Architecture TM tables EC tables SRAM Durable Storage Control Bus NAND NAND Flash Flash Flash Bus Flash Controller NAND NAND Flash Flash Read & Write Wear-leveling FTL 30 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  48. SSD Architecture Garbage Collection TM tables ECC errors EC tables CPU SRAM Durable Storage Control Bus NAND NAND Flash Flash Flash Bus Flash Controller NAND NAND Flash Flash Read & Write Wear-leveling FTL 30 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  49. SSD Architecture Garbage Collection TM tables ECC errors EC tables CPU SRAM Durable Storage Control Bus NAND NAND Flash Flash Flash Bus Flash Host Interface Controller PATA, SATA SCSI, etc NAND NAND Flash Flash Read & Write Wear-leveling FTL 30 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  50. SSD Architecture Garbage Collection TM tables ECC errors EC tables CPU SRAM Durable Storage Control Bus NAND NAND Flash Flash Flash Bus Flash Host Interface Controller PATA, SATA SCSI, etc NAND NAND Flash Flash Read & Write Wear-leveling FTL 30 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  51. SSD Architecture Garbage Collection TM tables ECC errors EC tables CPU SRAM Durable Storage Control Bus NAND NAND Flash Flash Data Bus Data Bus Flash Bus DRAM Flash Host Interface Buffer Controller PATA, SATA SCSI, etc NAND NAND Flash Flash Write Cache Read & Write Wear-leveling FTL 30 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  52. Data Reduction • Reducing the amount of user data effectively stored in flash chips allows to reduce the write rate and increase the life of flash drives • Data reduction techniques are: ◦ Compression ◦ Deduplication 31 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  53. Data Reduction Data Compression • It consists in reducing the number of bits needed to store data. • Lossless compression allows to restore data to its original state • Lossy compression permanently eliminates bits of data that are redundant, unimportant or imperceptible • CompressionRatio = UncompressedSize CompressedSize 1 ⇒ Data reduction is DR c = CompressionRatio Capacity SSD · PECycles SSD TTF ≈ WriteRate · (1 + A f ) · DR c 32 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  54. Data Reduction Data Deduplication • It looks for redundancy of sequences of bytes across very large comparison windows. • Sequences of data are compared to the history of other such sequences. • The first uniquely stored version of a sequence is referenced rather than stored again • Let DD the average percentage of deduplicable data Capacity SSD · PECycles SSD TTF ≈ WriteRate · (1 + A f ) · DR c · (1 − DD ) 33 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  55. RAID Solutions on flash technology • RAID uses redundancy (e.g. a parity code) to increase reliability • Any RAID solution increase the amount of data physically written on disks (RAID Overhead) ⇒ when adopting a RAID solution with flash technology we are reducing the lifetime of the whole storage system by a factor at most equal to the RAID overhead 34 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  56. RAID Solutions on flash technology Example • N flash disks of capacity C and cells supporting L P/E-cycles. • Write load rate equal to W. 35 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  57. RAID Solutions on flash technology Example • N flash disks of capacity C and cells supporting L P/E-cycles. • Write load rate equal to W. RAID0: • stripes data • no fault tolerance • W is uniformly distributed on disks (thanks to striping) 35 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  58. RAID Solutions on flash technology Example • N flash disks of capacity C and cells supporting L P/E-cycles. • Write load rate equal to W. RAID0: • stripes data • no fault tolerance • W is uniformly distributed on disks (thanks to striping) ⇒ TTL RAID 0 = N · C · L W 35 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  59. RAID Solutions on flash technology Example • N flash disks of capacity C and cells supporting L P/E-cycles. • Write load rate equal to W. RAID10: RAID0: • stripes data • stripes data • replicates each disk • no fault tolerance • W is uniformly distributed on disks (thanks to striping) ⇒ TTL RAID 0 = N · C · L W 35 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  60. RAID Solutions on flash technology Example • N flash disks of capacity C and cells supporting L P/E-cycles. • Write load rate equal to W. RAID10: RAID0: • stripes data • stripes data • replicates each disk • no fault tolerance • W is uniformly distributed on disks (thanks to striping) ⇒ TTL RAID 0 = N · C · L ⇒ TTL RAID 10 = N · C · L W 2 W 35 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  61. RAID Solutions on flash technology Example • N flash disks of capacity C and cells supporting L P/E-cycles. • Write load rate equal to W. RAID10: RAID0: • stripes data • stripes data • replicates each disk • no fault tolerance • W is uniformly distributed on disks (thanks to striping) ⇒ TTL RAID 0 = N · C · L ⇒ TTL RAID 10 = N · C · L W 2 W Alert! In order to increase reliability we half the time to live of flash cells 35 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  62. Modeling SSD endurance in a complex system 36 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  63. Modeling SSD endurance in a complex system SSDs 36 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  64. Modeling SSD endurance in a complex system SSDs System Workload 36 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  65. Modeling SSD endurance in a complex system SSDs SSD System System Workload 36 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  66. Does it still make sense to use RAID? • We have learned that any redundancy reduces the maximum time to live of all SSDs • The answer is YES , but why? 37 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  67. Does it still make sense to use RAID?YES! Why? CPU SRAM Control Bus NAND NAND Flash Flash Data Bus Data Bus Flash Bus DRAM Flash Host Interface Buffer Controller PATA, SATA SCSI, etc NAND NAND Flash Flash 38 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  68. Does it still make sense to use RAID?YES! Why? CPU SRAM ALL THESE COMPONENTS MAY HAVE Control Bus NAND NAND A FAUL T Flash Flash Data Bus Data Bus Flash Bus DRAM Flash Host Interface Buffer Controller PATA, SATA SCSI, etc NAND NAND Flash Flash 38 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  69. EMC XtremIO • The system building block is called XBrick: ◦ 25 800GB eMLC SSDs ◦ Two 1U Storage Controllers (redundant storage processors) • The scale up is guaranteed by adding more XBricks (up to six in a rack) that will be connected through InfiniBand ports. • The system performs inline data reduction by: ◦ deduplication ◦ compression 39 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  70. EMC XtremIO Deduplication • The system checks for deduplicated data: 1. subdivide the write stream in 4KB blocks 2. for each block in the stream 2.1 compute a digest 2.2 check in a shared mapping table the presence of the block 2.3 if present update a reference counter 2.4 else use the digest to determine the location of the block and send the block to the respective controller node • The addressing of blocks should uniformly distribute the data on all nodes 40 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  71. EMC XtremIO XtremIO Data Protection • The XtremIO system implements a proprietary data protection algorithm called XtremIO Data Protection (XDP) • Disks in a node are arranged in 23+2 columns • 1 row parity and 1 diagonal party • Each stripe is subdivided in 28 rows and 29 diagonals 41 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  72. EMC XtremIO XtremIO Data Protection • In order to compute efficiently the diagonal parity and to spread writes on all disks, XDP waits to fill in memory the emptiest stripe • When the stripe is full, commit it on disks • The emptiest stripe selection implies that free space is linearly distributed on stripes • XDP can: ◦ overcome 2 concurrent failures (2 parities) ◦ have a write overhead smaller than other RAID solutions 42 of 47 - An overview on solid-state-drives architectures and enterprise solutions

  73. EMC XtremIO XtremIO Data Protection Suppose a system that is 80% full: • The emptiest stripe is 40% full (due to the emptiest selection) • A stripe can handle 28 · 23 = 644 writes • The emptiest stripe can handle 644 · 40% ≈ 257 # parities = 28( rows ) + 29( diagonal ) = 57 # writes RAIDoverhead = # userwrites RAIDoverhead = 257 + 57 1 . 22 257 43 of 47 - An overview on solid-state-drives architectures and enterprise solutions

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend

Open-Channel Solid State Drives Matias Bjrling 2015/03/12 Vault 1 Solid State Drives

Open-Channel Solid State Drives Matias Bjrling 2015/03/12 Vault 1 Solid State Drives

Open-Channel Solid State Drives Matias Bjrling 2015/03/12 Vault 1 Solid State Drives Thousand of IOPS and low latency (&lt;1ms) Hardware continues to improve Parallel architecture Larger flash chips Replaceable for

232 views • 22 slides
Data Systems on Modern Hardware: Multi-cores, Solid-State Drives, and Non-Volatile Memories Prof.

Data Systems on Modern Hardware: Multi-cores, Solid-State Drives, and Non-Volatile Memories Prof.

CS 591: Da Data Systems Architectures Data Systems on Modern Hardware: Multi-cores, Solid-State Drives, and Non-Volatile Memories Prof. Manos Athanassoulis https://midas.bu.edu/classes/CS591A1 with slides developed at Harvard Some class

963 views • 74 slides
A Semi Preemptive Garbage A Semi Preemptive Garbage Collector for Solid State Collector

A Semi Preemptive Garbage A Semi Preemptive Garbage Collector for Solid State Collector

A Semi Preemptive Garbage A Semi Preemptive Garbage Collector for Solid State Collector for Solid State Collector for Solid State Collector for Solid State Drives Drives Junghee Lee, Youngjae Kim, Galen M. Shipman, Sarp Oral, Feiyi Wang,

540 views • 26 slides
SFS: Random Write Considered Harmful in Solid State Drives Changwoo Min 1, 2 , Kangnyeon Kim 1 ,

SFS: Random Write Considered Harmful in Solid State Drives Changwoo Min 1, 2 , Kangnyeon Kim 1 ,

SFS: Random Write Considered Harmful in Solid State Drives Changwoo Min 1, 2 , Kangnyeon Kim 1 , Hyunjin Cho 2 , Sang-Won Lee 1 , Young Ik Eom 1 1 Sungkyunkwan University, Korea 2 Samsung Electronics, Korea Outline Background Design

367 views • 34 slides
FLIN: Enabling Fairness and Enhancing Performance in Modern NVMe Solid State Drives Arash

FLIN: Enabling Fairness and Enhancing Performance in Modern NVMe Solid State Drives Arash

FLIN: Enabling Fairness and Enhancing Performance in Modern NVMe Solid State Drives Arash Tavakkol, Mohammad Sadrosadati, Saugata Ghose, Jeremie S. Kim, Yixin Luo, Yaohua Wang, Nika Mansouri Ghiasi, Lois Orosa, Juan Gmez-Luna, Onur Mutlu June

1.14k views • 92 slides
Reliably Erasing Data from Flash-Based Solid State Drives Michael Wei* Laura Grupp*, Fredrick E.

Reliably Erasing Data from Flash-Based Solid State Drives Michael Wei* Laura Grupp*, Fredrick E.

Reliably Erasing Data from Flash-Based Solid State Drives Michael Wei* Laura Grupp*, Fredrick E. Spada, Steven Swanson* * Non-Volatile Systems Laboratory Department of Computer Science and Engineering University of California, San Diego

847 views • 80 slides
Solid State Drives (SSDs) Daniel Mosse (slides are modified from Dr. Ahmed Amers CS 1550 Slides

Solid State Drives (SSDs) Daniel Mosse (slides are modified from Dr. Ahmed Amers CS 1550 Slides

Solid State Drives (SSDs) Daniel Mosse (slides are modified from Dr. Ahmed Amers CS 1550 Slides and Sherif Khattab ) Historical Disk drive specifics IBM 360KB WD 18GB Seagate ST9250410AS floppy HD 250GB HD (16MB cache)

157 views • 11 slides
LaLDPC: Latency-aware LDPC for Read Performance Improvement of Solid State Drives Yajuan Du 1,2 ,

LaLDPC: Latency-aware LDPC for Read Performance Improvement of Solid State Drives Yajuan Du 1,2 ,

LaLDPC: Latency-aware LDPC for Read Performance Improvement of Solid State Drives Yajuan Du 1,2 , Deqing Zou 1 , Qiao Li 3 , Liang Shi 3 , Hai Jin 1 , and Chun Jason Xue 2 1 1 Huazhong University of Science and Technology 2 City University of

599 views • 41 slides
The Unwritten Contract of Solid State Drives Jun He, Sudarsun Kannan, Andrea C. Arpaci-Dusseau,

The Unwritten Contract of Solid State Drives Jun He, Sudarsun Kannan, Andrea C. Arpaci-Dusseau,

The Unwritten Contract of Solid State Drives Jun He, Sudarsun Kannan, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau Department of Computer Sciences, University of Wisconsin - Madison Enterprise SSD revenue is expected to exceed enterprise

2.48k views • 164 slides
Saft 28 April 2019 Saft Solid State (3S) program overview Program with What will Solid State

Saft 28 April 2019 Saft Solid State (3S) program overview Program with What will Solid State

Saft 28 April 2019 Saft Solid State (3S) program overview Program with What will Solid State Keys to success 2 related targets deliver? High energy 1. Develop state-of-the Ionic conductivity equal or better to liquid electrolyte art

3.56k views • 5 slides
Self-Encrypting Deception: Weaknesses in the Encryption of Solid State Drives (SSDs) Carlo Meijer

Self-Encrypting Deception: Weaknesses in the Encryption of Solid State Drives (SSDs) Carlo Meijer

Self-Encrypting Deception: Weaknesses in the Encryption of Solid State Drives (SSDs) Carlo Meijer Bernard van Gastel Radboud University Nijmegen Radboud University Nijmegen Midnight Blue Labs Open University of the Netherlands whoami Carlo

1.44k views • 141 slides
FAST: Quick Application Launch on Solid-State Drives Yongsoo Joo 1 , Junhee Ryu 2 , Sangsoo Park

FAST: Quick Application Launch on Solid-State Drives Yongsoo Joo 1 , Junhee Ryu 2 , Sangsoo Park

FAST: Quick Application Launch on Solid-State Drives Yongsoo Joo 1 , Junhee Ryu 2 , Sangsoo Park 1 , and Kang G. Shin 1,3 1 Ewha Womans University, Korea 2 Seoul National University, Korea 3 University of Michigan, USA Application Launch Delay

402 views • 36 slides
Open-Channel SSDs Matjas Bjrling LinuxCon North America 2015 1 Solid State Drives H i g h

Open-Channel SSDs Matjas Bjrling LinuxCon North America 2015 1 Solid State Drives H i g h

Open-Channel SSDs Matjas Bjrling LinuxCon North America 2015 1 Solid State Drives H i g h T h r o u g h p u t + L o w L a t e n c y P a r a l l e l i s m + C o n t r o l l e r Why Open-Channel

442 views • 16 slides
Red phosphors for solid- -state state Red phosphors for solid (LED) lighting applications

Red phosphors for solid- -state state Red phosphors for solid (LED) lighting applications

Red phosphors for solid- -state state Red phosphors for solid (LED) lighting applications (LED) lighting applications Intern: Dano Pagenkopf, Santa Barbara City College, Electrical Engineering/Materials Science Mentor: Dr. Gautam Gundiah

1.21k views • 18 slides
Overview Agent Architectures Definition of agent architecture Classical Architectures for

Overview Agent Architectures Definition of agent architecture Classical Architectures for

Overview Agent Architectures Definition of agent architecture Classical Architectures for robots consists of functional components Situated Automata eliminates explicit deliberation Behavior-Based Architectures behaviors as essential

641 views • 11 slides
Inorganic Phosphors For Solid State Inorganic Phosphors For Solid State Lighting Lighting Ka Wo

Inorganic Phosphors For Solid State Inorganic Phosphors For Solid State Lighting Lighting Ka Wo

Inorganic Phosphors For Solid State Inorganic Phosphors For Solid State Lighting Lighting Ka Wo Ho Biochemistry Major Chabot College (August 2003 May 2006) University of California, San Diego Mentor: Dr. Dr. Gautam Gautam Gundiah

910 views • 15 slides
Implicit-Explicit Runge-Kutta schemes for the Boltzmann-Poisson equation for semiconductors

Implicit-Explicit Runge-Kutta schemes for the Boltzmann-Poisson equation for semiconductors

Implicit-Explicit Runge-Kutta schemes for the Boltzmann-Poisson equation for semiconductors Vittorio Rispoli In collaboration with: G. Dimarco (Univ. of Toulouse, France) L. Pareschi (Univ. of Ferrara, Italy) HYP2012 University of Padova

498 views • 34 slides
Studies with DEPFETs for BELLE II 24th IMPRS Workshop Munich, 26th November 2012 Stefan

Studies with DEPFETs for BELLE II 24th IMPRS Workshop Munich, 26th November 2012 Stefan

Neutron Irradiations and Punch-Through-Biasing Studies with DEPFETs for BELLE II 24th IMPRS Workshop Munich, 26th November 2012 Stefan Petrovics Outline I. Theoretical Background 1. DEPFET 2. Radiation Damage 3. Punch-through biasing 4.

331 views • 31 slides
Microelectronics: The Beginning of the End or the End of the Beginning? Sanjay K. Banerjee

Microelectronics: The Beginning of the End or the End of the Beginning? Sanjay K. Banerjee

ITIF Panel: Future of Moores Law: Semiconductor Innovation &amp; the High Tech Economy Microelectronics: The Beginning of the End or the End of the Beginning? Sanjay K. Banerjee Director, Microelectronics Research Center Univ. of Texas at

343 views • 6 slides
Conductivity and Semi-Conductors J = current density = I/A E = Electric field intensity = V/l

Conductivity and Semi-Conductors J = current density = I/A E = Electric field intensity = V/l

Conductivity and Semi-Conductors J = current density = I/A E = Electric field intensity = V/l where l is the distance between two points Metals: Semiconductors: Many Polymers and Glasses 1 Electrical Conduction (motion of

742 views • 59 slides
The Belle II Pixel Detector DAQ unchow 1 David M Thomas Geler 1 , Igor Konorov 3 , Wolfgang

The Belle II Pixel Detector DAQ unchow 1 David M Thomas Geler 1 , Igor Konorov 3 , Wolfgang

The Belle II Pixel Detector DAQ unchow 1 David M Thomas Geler 1 , Igor Konorov 3 , Wolfgang K uhn 1 , S oren Lange 1 , Dmytro Levit 3 , ZhenAn Liu 2 , orn Spruck 1 , Klemens Lautenbach 1 , Jingzhou Zhao 2 Bj 1 University Gieen,

266 views • 24 slides
Field Programmable Gate Arrays by Ketil Red Field Programmable Gate Array Integrated

Field Programmable Gate Arrays by Ketil Red Field Programmable Gate Array Integrated

A brief introduction to Field Programmable Gate Arrays by Ketil Red Field Programmable Gate Array Integrated circuit including a matrix of general-purpose programmable logic I I I I I I I I I I I I I I I I I I I I

439 views • 14 slides
Programable Logic Devices In the 1970s programmable logic circuits called programmable logic

Programable Logic Devices In the 1970s programmable logic circuits called programmable logic

Programable Logic Devices In the 1970s programmable logic circuits called programmable logic device (PLD) was introduced. They are based on a structure with an AND- OR array that makes it easy to implement SOP expression William Sandqvist

830 views • 64 slides
ROMs, PLAs and FPGAs October 5, 2006 Typeset by Foil T EX Why Programmable Logic?

ROMs, PLAs and FPGAs October 5, 2006 Typeset by Foil T EX Why Programmable Logic?

ROMs, PLAs and FPGAs October 5, 2006 Typeset by Foil T EX Why Programmable Logic? Programmable logic technologies: Read-Only Memory (ROM) Programmable Logic Array (PLA) Programmable Array Logic (PAL) Field Programmable

283 views • 13 slides

More recommend