pebblesdb building key value stores using fragmented log
play

PebblesDB: Building Key-Value Stores using Fragmented Log - PowerPoint PPT Presentation

Jul 15, 2023 •339 likes •1.53k views

PebblesDB: Building Key-Value Stores using Fragmented Log Structured Merge Trees Pandian Raju 1 , Rohan Kadekodi 1 , Vijay Chidambaram 1,2 , Ittai Abraham 2 1 The University of Texas at Austin 2 VMware Research What is a key-value store?

  1. FLSM structure In-memory Memory Storage 2 …. 37 23 …. 48 Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 15 40 70 95 Level 2 2 …. 8 16 …. 32 15 …. 23 96 …. 99 45 …. 65 70 …. 90 Note how files are logically grouped within guards 40

  2. FLSM structure In-memory Memory Storage 2 …. 37 23 …. 48 Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 15 40 70 95 Level 2 2 …. 8 16 …. 32 15 …. 23 96 …. 99 45 …. 65 70 …. 90 Guards get more fine grained deeper into the tree 41

  3. How does FLSM reduce write amplification? 42

  4. How does FLSM reduce write amplification? In-memory 30 …. 68 Memory Storage 2 …. 37 23 …. 48 Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 15 40 70 95 Level 2 2 …. 8 16 …. 32 15 …. 23 96 …. 99 45 …. 65 70 …. 90 Max files in level 0 is configured to be 2 43

  5. How does FLSM reduce write amplification? 15 In-memory 15 …. 68 2 …. 68 2 …. 14 Memory Storage 2 …. 37 23 …. 48 30 …. 68 Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 15 40 70 95 Level 2 2 …. 8 16 …. 32 15 …. 23 96 …. 99 45 …. 65 70 …. 90 Compacting level 0 44

  6. How does FLSM reduce write amplification? 15 In-memory 2 …. 14 15 …. 68 Memory Storage Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 15 40 70 95 Level 2 2 …. 8 16 …. 32 15 …. 23 96 …. 99 45 …. 65 70 …. 90 Fragmented files are just appended to next level 45

  7. How does FLSM reduce write amplification? In-memory 15 …. 68 Memory Storage Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 2 …. 14 15 …. 68 15 40 70 95 Level 2 2 …. 8 16 …. 32 15 …. 23 96 …. 99 45 …. 65 70 …. 90 Guard 15 in Level 1 is to be compacted 46

  8. How does FLSM reduce write amplification? 40 In-memory 15 …. 68 40 …. 68 15 …. 39 Memory Storage Level 0 15 70 Level 1 1 …. 12 77 …. 87 82 …. 95 2 …. 14 15 40 70 95 Level 2 2 …. 8 16 …. 32 96 …. 99 15 …. 23 45 …. 65 70 …. 90 Files are combined, sorted and fragmented 47

  9. How does FLSM reduce write amplification? 40 In-memory 40 …. 68 15 …. 39 Memory Storage Level 0 15 70 Level 1 1 …. 12 77 …. 87 82 …. 95 2 …. 14 15 40 95 70 Level 2 2 …. 8 16 …. 32 15 …. 23 45 …. 65 96 …. 99 70 …. 90 Fragmented files are just appended to next level 48

  10. How does FLSM reduce write amplification? FLSM doesn’t re-write data to the same level in most cases How does FLSM maintain read performance? FLSM maintains partially sorted levels to efficiently reduce the search space 49

  11. Selecting Guards • Guards are chosen randomly and dynamically • Dependent on the distribution of data 50

  12. Selecting Guards • Guards are chosen randomly and dynamically • Dependent on the distribution of data Keyspace 1 1e+9 51

  13. Selecting Guards • Guards are chosen randomly and dynamically • Dependent on the distribution of data Keyspace 1 1e+9 52

  14. Selecting Guards • Guards are chosen randomly and dynamically • Dependent on the distribution of data Keyspace 1 1e+9 53

  15. Operations: Write Put(1, “abc”) Write (key, value) In-memory Memory Storage 2 …. 37 23 …. 48 Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 15 40 70 95 Level 2 2 …. 8 16 …. 32 15 …. 23 96 …. 99 45 …. 65 70 …. 90 FLSM structure 54

  16. Operations: Get Get(23) In-memory Memory Storage 2 …. 37 23 …. 48 Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 15 40 70 95 Level 2 2 …. 8 16 …. 32 15 …. 23 96 …. 99 45 …. 65 70 …. 90 FLSM structure 55

  17. Operations: Get Get(23) In-memory Memory Storage 2 …. 37 23 …. 48 Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 15 40 70 95 Level 2 2 …. 8 16 …. 32 15 …. 23 96 …. 99 45 …. 65 70 …. 90 Search level by level starting from memory 56

  18. Operations: Get Get(23) In-memory Memory Storage 2 …. 37 23 …. 48 Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 15 40 70 95 Level 2 2 …. 8 16 …. 32 15 …. 23 96 …. 99 45 …. 65 70 …. 90 All level 0 files need to be searched 57

  19. Operations: Get Get(23) In-memory Memory Storage 2 …. 37 23 …. 48 Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 15 40 70 95 Level 2 2 …. 8 16 …. 32 15 …. 23 96 …. 99 45 …. 65 70 …. 90 Level 1: File under guard 15 is searched 58

  20. Operations: Get Get(23) In-memory Memory Storage 2 …. 37 23 …. 48 Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 15 40 70 95 Level 2 2 …. 8 16 …. 32 15 …. 23 96 …. 99 45 …. 65 70 …. 90 Level 2: Both the files under guard 15 are searched 59

  21. High write throughput in FLSM • Compaction from memory to level 0 is stalled • Writes to memory is also stalled Write (key, value) In-memory Memory Storage Level 0 2 …. 98 23 …. 48 1 …. 37 18 …. 48 If rate of insertion is higher than rate of compaction, write throughput depends on the rate of compaction 60

  22. High write throughput in FLSM • Compaction from memory to level 0 is stalled • Writes to memory is also stalled Write (key, value) FLSM has faster compaction because of lesser I/O and In-memory Memory hence higher write throughput Storage Level 0 2 …. 98 23 …. 48 1 …. 37 18 …. 48 If rate of insertion is higher than rate of compaction, write throughput depends on the rate of compaction 61

  23. Challenges in FLSM • Every read/range query operation needs to examine multiple files per level • For example, if every guard has 5 files, read latency is increased by 5x (assuming no cache hits) Trade-off between write I/O and read performance 62

  24. Outline • Log-Structured Merge Tree (LSM) • Fragmented Log-Structured Merge Tree (FLSM) • Building PebblesDB using FLSM • Evaluation • Conclusion 63

  25. PebblesDB • Built by modifying HyperLevelDB ( ± 9100 LOC) to use FLSM • HyperLevelDB, built over LevelDB, to provide improved parallelism and compaction • API compatible with LevelDB, but not with RocksDB 64

  26. Optimizations in PebblesDB • Challenge (get/range query): Multiple files in a guard • Get() performance is improved using file level bloom filter 65

  27. Optimizations in PebblesDB • Challenge (get/range query): Multiple files in a guard • Get() performance is improved using file level bloom filter Is key 25 Definitely not Bloom filter present? Possibly yes 66

  28. Optimizations in PebblesDB • Challenge (get/range query): Multiple files in a guard • Get() performance is improved using file level bloom filter 15 70 Level 1 1 …. 12 15 …. 39 82 …. 95 77 …. 97 Maintained Bloom Filter Bloom Filter Bloom Filter Bloom Filter in-memory 67

  29. Optimizations in PebblesDB • Challenge (get/range query): Multiple files in a guard • Get() performance is improved using file level bloom filter 15 70 Level 1 1 …. 12 15 …. 39 82 …. 95 77 …. 97 Maintained Bloom Filter Bloom Filter Bloom Filter Bloom Filter in-memory PebblesDB reads same number of files as any LSM based store 68

  30. Optimizations in PebblesDB • Challenge (get/range query): Multiple files in a guard • Get() performance is improved using file level bloom filter • Range query performance is improved using parallel threads and better compaction 69

  31. Outline • Log-Structured Merge Tree (LSM) • Fragmented Log-Structured Merge Tree (FLSM) • Building PebblesDB using FLSM • Evaluation • Conclusion 70

  32. Evaluation Real world workloads - YCSB Crash recovery Micro-benchmarks CPU and memory Low memory usage Small dataset Aged file system NoSQL applications 71

  33. Evaluation Real world workloads - YCSB Crash recovery Micro-benchmarks CPU and memory Low memory usage Small dataset Aged file system NoSQL applications 72

  34. Real world workloads - YCSB • Yahoo! Cloud Serving Benchmark - Industry standard macro-benchmark • Insertions: 50M, Operations: 10M, key size: 16 bytes and value size: 1 KB 2.5 2 Throughput ratio wrt HyperLevelDB 1.5 1 0.5 0 Load A Run A Run B Run C Run D Load E Run E Run F Total IO Load A - 100 % writes Run D - 95% reads (latest), 5% writes Run A - 50% reads, 50% writes Load E - 100% writes Run B - 95% reads, 5% writes Run E - 95% range queries, 5% writes Run C - 100% reads Run F - 50% reads, 50% read-modify-writes 73

  35. Real world workloads - YCSB • Yahoo! Cloud Serving Benchmark - Industry standard macro-benchmark • Insertions: 50M, Operations: 10M, key size: 16 bytes and value size: 1 KB 2.5 2 35.08 Kops/s 33.98 Kops/s 22.41 Kops/s 57.87 Kops/s 34.06 Kops/s 32.09 Kops/s Throughput ratio wrt 25.8 Kops/s 5.8 Kops/s 952.93 GB HyperLevelDB 1.5 1 0.5 0 Load A Run A Run B Run C Run D Load E Run E Run F Total IO Load A - 100 % writes Run D - 95% reads (latest), 5% writes Run A - 50% reads, 50% writes Load E - 100% writes Run B - 95% reads, 5% writes Run E - 95% range queries, 5% writes Run C - 100% reads Run F - 50% reads, 50% read-modify-writes 74

  36. Real world workloads - YCSB • Yahoo! Cloud Serving Benchmark - Industry standard macro-benchmark • Insertions: 50M, Operations: 10M, key size: 16 bytes and value size: 1 KB 2.5 2 35.08 Kops/s 33.98 Kops/s 22.41 Kops/s 57.87 Kops/s 34.06 Kops/s 32.09 Kops/s Throughput ratio wrt 25.8 Kops/s 5.8 Kops/s 952.93 GB HyperLevelDB 1.5 1 0.5 0 Load A Run A Run B Run C Run D Load E Run E Run F Total IO Load A - 100 % writes Run D - 95% reads (latest), 5% writes Run A - 50% reads, 50% writes Load E - 100% writes Run B - 95% reads, 5% writes Run E - 95% range queries, 5% writes Run C - 100% reads Run F - 50% reads, 50% read-modify-writes 75

  37. Real world workloads - YCSB • Yahoo! Cloud Serving Benchmark - Industry standard macro-benchmark • Insertions: 50M, Operations: 10M, key size: 16 bytes and value size: 1 KB 2.5 2 35.08 Kops/s 33.98 Kops/s 22.41 Kops/s 57.87 Kops/s 34.06 Kops/s 32.09 Kops/s Throughput ratio wrt 25.8 Kops/s 5.8 Kops/s 952.93 GB HyperLevelDB 1.5 1 0.5 0 Load A Run A Run B Run C Run D Load E Run E Run F Total IO Load A - 100 % writes Run D - 95% reads (latest), 5% writes Run A - 50% reads, 50% writes Load E - 100% writes Run B - 95% reads, 5% writes Run E - 95% range queries, 5% writes Run C - 100% reads Run F - 50% reads, 50% read-modify-writes 76

  38. Real world workloads - YCSB • Yahoo! Cloud Serving Benchmark - Industry standard macro-benchmark • Insertions: 50M, Operations: 10M, key size: 16 bytes and value size: 1 KB 2.5 2 35.08 Kops/s 33.98 Kops/s 22.41 Kops/s 57.87 Kops/s 34.06 Kops/s 32.09 Kops/s Throughput ratio wrt 25.8 Kops/s 5.8 Kops/s 952.93 GB HyperLevelDB 1.5 1 0.5 0 Load A Run A Run B Run C Run D Load E Run E Run F Total IO Load A - 100 % writes Run D - 95% reads (latest), 5% writes Run A - 50% reads, 50% writes Load E - 100% writes Run B - 95% reads, 5% writes Run E - 95% range queries, 5% writes Run C - 100% reads Run F - 50% reads, 50% read-modify-writes 77

  39. Real world workloads - YCSB • Yahoo! Cloud Serving Benchmark - Industry standard macro-benchmark • Insertions: 50M, Operations: 10M, key size: 16 bytes and value size: 1 KB 2.5 2 35.08 Kops/s 33.98 Kops/s 22.41 Kops/s 57.87 Kops/s 34.06 Kops/s 32.09 Kops/s Throughput ratio wrt 25.8 Kops/s 5.8 Kops/s 952.93 GB HyperLevelDB 1.5 1 0.5 0 Load A Run A Run B Run C Run D Load E Run E Run F Total IO Load A - 100 % writes Run D - 95% reads (latest), 5% writes Run A - 50% reads, 50% writes Load E - 100% writes Run B - 95% reads, 5% writes Run E - 95% range queries, 5% writes Run C - 100% reads Run F - 50% reads, 50% read-modify-writes 78

  40. NoSQL stores - MongoDB • YCSB on MongoDB, a widely used key-value store • Inserted 20M key-value pairs with 1 KB value size and 10M operations 2.5 2 Throughput ratio wrt WiredTiger 1.5 1 0.5 0 Load A Run A Run B Run C Run D Load E Run E Run F Total IO Load A - 100 % writes Run D - 95% reads (latest), 5% writes Run A - 50% reads, 50% writes Load E - 100% writes Run B - 95% reads, 5% writes Run E - 95% range queries, 5% writes Run C - 100% reads Run F - 50% reads, 50% read-modify-writes 79

  41. NoSQL stores - MongoDB • YCSB on MongoDB, a widely used key-value store • Inserted 20M key-value pairs with 1 KB value size and 10M operations 2.5 2 20.73 Kops/s 15.52 Kops/s 19.69 Kops/s 23.53 Kops/s 20.68 Kops/s Throughput ratio wrt 9.95 Kops/s 0.65 Kops/s 9.78 Kops/s 426.33 GB WiredTiger 1.5 1 0.5 0 Load A Run A Run B Run C Run D Load E Run E Run F Total IO Load A - 100 % writes Run D - 95% reads (latest), 5% writes Run A - 50% reads, 50% writes Load E - 100% writes Run B - 95% reads, 5% writes Run E - 95% range queries, 5% writes Run C - 100% reads Run F - 50% reads, 50% read-modify-writes 80

  42. NoSQL stores - MongoDB • YCSB on MongoDB, a widely used key-value store • Inserted 20M key-value pairs with 1 KB value size and 10M operations 2.5 2 20.73 Kops/s 15.52 Kops/s 19.69 Kops/s 23.53 Kops/s 20.68 Kops/s Throughput ratio wrt 9.95 Kops/s 0.65 Kops/s 9.78 Kops/s 426.33 GB WiredTiger 1.5 1 0.5 0 Load A Run A Run B Run C Run D Load E Run E Run F Total IO Load A - 100 % writes Run D - 95% reads (latest), 5% writes Run A - 50% reads, 50% writes Load E - 100% writes Run B - 95% reads, 5% writes Run E - 95% range queries, 5% writes Run C - 100% reads Run F - 50% reads, 50% read-modify-writes 81

  43. NoSQL stores - MongoDB • YCSB on MongoDB, a widely used key-value store • Inserted 20M key-value pairs with 1 KB value size and 10M operations 2.5 2 20.73 Kops/s 15.52 Kops/s 19.69 Kops/s 23.53 Kops/s 20.68 Kops/s Throughput ratio wrt 9.95 Kops/s 0.65 Kops/s 9.78 Kops/s 426.33 GB WiredTiger 1.5 1 0.5 0 Load A Run A Run B Run C Run D Load E Run E Run F Total IO Load A - 100 % writes Run D - 95% reads (latest), 5% writes Run A - 50% reads, 50% writes Load E - 100% writes Run B - 95% reads, 5% writes Run E - 95% range queries, 5% writes Run C - 100% reads Run F - 50% reads, 50% read-modify-writes 82

  44. NoSQL stores - MongoDB • YCSB on MongoDB, a widely used key-value store • Inserted 20M key-value pairs with 1 KB value size and 10M operations 2.5 2 20.73 Kops/s 15.52 Kops/s 19.69 Kops/s 23.53 Kops/s 20.68 Kops/s Throughput ratio wrt 9.95 Kops/s 0.65 Kops/s 9.78 Kops/s 426.33 GB WiredTiger 1.5 1 0.5 0 Load A Run A Run B Run C Run D Load E Run E Run F Total IO Load A - 100 % writes Run D - 95% reads (latest), 5% writes Run A - 50% reads, 50% writes Load E - 100% writes Run B - 95% reads, 5% writes Run E - 95% range queries, 5% writes Run C - 100% reads Run F - 50% reads, 50% read-modify-writes 83

  45. NoSQL stores - MongoDB • YCSB on MongoDB, a widely used key-value store • Inserted 20M key-value pairs with 1 KB value size and 10M operations 2.5 2 20.73 Kops/s 15.52 Kops/s 19.69 Kops/s 23.53 Kops/s 20.68 Kops/s Throughput ratio wrt 9.95 Kops/s 0.65 Kops/s 9.78 Kops/s 426.33 GB WiredTiger 1.5 PebblesDB combines low write IO of WiredTiger with 1 high performance of RocksDB 0.5 0 Load A Run A Run B Run C Run D Load E Run E Run F Total IO Load A - 100 % writes Run D - 95% reads (latest), 5% writes Run A - 50% reads, 50% writes Load E - 100% writes Run B - 95% reads, 5% writes Run E - 95% range queries, 5% writes Run C - 100% reads Run F - 50% reads, 50% read-modify-writes 84

  46. Outline • Log-Structured Merge Tree (LSM) • Fragmented Log-Structured Merge Tree (FLSM) • Building PebblesDB using FLSM • Evaluation • Conclusion 85

  47. Conclusion • PebblesDB: key-value store built on Fragmented Log-Structured Merge Trees • Increases write throughput and reduces write IO at the same time • Obtains 6X the write throughput of RocksDB • As key-value stores become more widely used, there have been several attempts to optimize them • PebblesDB combines algorithmic innovation (the FLSM data structure) with careful systems building 86

  48. https://github.com/utsaslab/pebblesdb

  49. https://github.com/utsaslab/pebblesdb

  50. Backup slides 89

  51. Operations: Seek • Seek(target): Returns the smallest key in the database which is >= target • Used for range queries (for example, return all entries between 5 and 18) Get(1) Level 0 – 1, 2, 100, 1000 Level 1 – 1, 5, 10, 2000 Level 2 – 5, 300, 500 90

  52. Operations: Seek • Seek(target): Returns the smallest key in the database which is >= target • Used for range queries (for example, return all entries between 5 and 18) Seek(200) Level 0 – 1, 2, 100, 1000 Level 1 – 1, 5, 10, 2000 Level 2 – 5, 300, 500 91

  53. Operations: Seek • Seek(target): Returns the smallest key in the database which is >= target • Used for range queries (for example, return all entries between 5 and 18) 92

  54. Operations: Seek Seek(23) In-memory Memory Storage 2 …. 37 23 …. 48 Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 15 40 70 95 Level 2 2 …. 8 16 …. 32 96 …. 99 15 …. 23 45 …. 65 70 …. 90 FLSM structure 93

  55. Operations: Seek Seek(23) In-memory Memory Storage 2 …. 37 23 …. 48 Level 0 15 70 Level 1 1 …. 12 15 …. 59 77 …. 87 82 …. 95 15 40 70 95 Level 2 2 …. 8 16 …. 32 96 …. 99 15 …. 23 45 …. 65 70 …. 90 All levels and memtable need to be searched 94

  56. Optimizations in PebblesDB • Challenge with reads: Multiple sstable reads per level • Optimized using sstable level bloom filters • Bloom filter: determine if an element is in a set Is key 25 Definitely not Bloom filter present? Possibly yes 95

  57. Optimizations in PebblesDB • Challenge with reads: Multiple sstable reads per level • Optimized using sstable level bloom filters • Bloom filter: determine if an element is in a set 15 70 Level 1 1 …. 12 15 …. 39 82 …. 95 77 …. 97 Maintained Bloom Filter Bloom Filter Bloom Filter Bloom Filter in-memory True Get(97) 96

  58. Optimizations in PebblesDB • Challenge with reads: Multiple sstable reads per level • Optimized using sstable level bloom filters • Bloom filter: determine if an element is in a set 15 70 Level 1 1 …. 12 15 …. 39 82 …. 95 77 …. 97 Bloom Filter Bloom Filter Bloom Filter Bloom Filter False True Get(97) 97

  59. Optimizations in PebblesDB • Challenge with reads: Multiple sstable reads per level • Optimized using sstable level bloom filters • Bloom filter: determine if an element is in a set 15 70 Level 1 1 …. 12 15 …. 39 82 …. 95 77 …. 97 Bloom Filter Bloom Filter Bloom Filter Bloom Filter PebblesDB reads at most one file per guard with high probability 98

  60. Optimizations in PebblesDB • Challenge with seeks: Multiple sstable reads per level • Parallel seeks: Parallel threads to seek() on files in a guard Thread 1 Thread 2 15 70 Level 1 1 …. 12 15 …. 39 77 …. 97 82 …. 95 Seek(85) 99

  61. Optimizations in PebblesDB • Challenge with seeks: Multiple sstable reads per level • Parallel seeks: Parallel threads to seek() on files in a guard • Seek based compaction: Triggers compaction for a level during a seek-heavy workload • Reduce the average number of sstables per guard • Reduce the number of active levels Seek based compaction increases write I/O but as a trade-off to improve seek performance 100

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

Pebbles DB: Building Key-Value Stores using Fragmented Log- Structured Merge Trees(II) Peter

Pebbles DB: Building Key-Value Stores using Fragmented Log- Structured Merge Trees(II) Peter

Pebbles DB: Building Key-Value Stores using Fragmented Log- Structured Merge Trees(II) Peter Sassaman Overview Sentinel Guards Guard Selection (MurmurHash) SSTables High Levels in memory (1) FLSM can be viewed

164 views • 12 slides
Tizen. Using the Smart Approach to Attract Apps the problem highly fragmented app market 3

Tizen. Using the Smart Approach to Attract Apps the problem highly fragmented app market 3

Tizen. Using the Smart Approach to Attract Apps the problem highly fragmented app market 3 825,000 850,000 145,000 * *Windows Phone Store 4 consumers developers 5 6 Xchange acts as a thin-wholesale layer between stores and apps 7

278 views • 26 slides
Key-Value Stores Key-value stores are popular. web searching, social networks, e-commerce,

Key-Value Stores Key-value stores are popular. web searching, social networks, e-commerce,

AC AC-Ke Key : Adaptive Caching for LSM-based Key-Value Stores Fenggang Wu , Ming-Hong Yang, Baoquan Zhang, David H.C. Du University of Minnesota, Twin Cities July 2020. USENIX ATC20 Key-Value Stores Key-value stores are popular. web

326 views • 16 slides
Flotection: Building Main Shutoff System Overview of System Retail stores have a requirement for

Flotection: Building Main Shutoff System Overview of System Retail stores have a requirement for

Flotection: Building Main Shutoff System Overview of System Retail stores have a requirement for night time monitoring of the Domestic Water Supply. When flow is detected, during the hours the store is closed, the system shuts off this supply

390 views • 17 slides
Sustainable Roofing M. Fenner, Group Manager Existing Stores Construction-Building Envelope,

Sustainable Roofing M. Fenner, Group Manager Existing Stores Construction-Building Envelope,

Sustainable Roofing M. Fenner, Group Manager Existing Stores Construction-Building Envelope, Target Corporation S. Graveline, Vice President Technical Services Sika Sarnafil, a Division of Sika Corporation Cool Roof Rating Council Membership

826 views • 44 slides
COLD STORES PLC - Company Profile Cold Stores manufacturers and markets a unique brand of ice

COLD STORES PLC - Company Profile Cold Stores manufacturers and markets a unique brand of ice

COLD STORES PLC - Company Profile Cold Stores manufacturers and markets a unique brand of ice cream which is the leading frozen Please use the following information to answer the questions in Part A COLD STORES PLC Managing Finance Graduate

418 views • 3 slides
Fragmented Log Structured Merge Trees (Part 1) Presented by Deepak Varghese Pebble DB

Fragmented Log Structured Merge Trees (Part 1) Presented by Deepak Varghese Pebble DB

Pebble Db Key Value Store Using Fragmented Log Structured Merge Trees (Part 1) Presented by Deepak Varghese Pebble DB Overview High performance write-optimized key-value store Built using new data structure Fragmented Log-

309 views • 8 slides
Wide-Area Storage Stores: Stores: Stores: Status Updates

Wide-Area Storage Stores: Stores: Stores: Status Updates

Dont Se$le for Eventual : Scalable Causal Consistency for Wide-Area Storage

424 views • 41 slides
Column-Stores vs. Row-Stores: How Different Are They Really? Daniel Abadi (Yale), Samuel Madden

Column-Stores vs. Row-Stores: How Different Are They Really? Daniel Abadi (Yale), Samuel Madden

Column-Stores vs. Row-Stores: How Different Are They Really? Daniel Abadi (Yale), Samuel Madden (MIT), Nabil Hachem (AvantGarde Consulting) June 12 th , 2008 Daniel Abadi -- Yale University Row vs. Column-Stores Row-Store Column-Store Last

495 views • 26 slides
20180807 Market Capitalization: 122.5bn (US$2.3 bn 1 ) as of Aug 6, 2018 We are one of the

20180807 Market Capitalization: 122.5bn (US$2.3 bn 1 ) as of Aug 6, 2018 We are one of the

20180807 Market Capitalization: 122.5bn (US$2.3 bn 1 ) as of Aug 6, 2018 We are one of the largest multi-format retailers in the Philippines 1 Forex Rate: 1USD= 53.146 PHP 2 Supermarkets Department Stores DIY Stores Convenience Stores

506 views • 36 slides
Building dev tools at the right level of abstraction Ben Davis CTO @BenCDavis

Building dev tools at the right level of abstraction Ben Davis CTO @BenCDavis

Building dev tools at the right level of abstraction Ben Davis CTO @BenCDavis ben@gatherdata.co The data engineering industry is very fragmented. Gather is a data integration tool for developers . It makes it really easy to build integration

728 views • 29 slides
Understanding Price Variation Across Stores and Supermarket Chains Stores and Supermarket Chains

Understanding Price Variation Across Stores and Supermarket Chains Stores and Supermarket Chains

Understanding Price Variation Across Stores and Supermarket Chains Stores and Supermarket Chains Some Implications for CPI Aggregation Methods Lorraine Ivancic Kevin J. Fox Aggregation: Study Motivation Availability of scanner data:

415 views • 19 slides
1 H I G H L I G H T S Group revenue of 69.2m, down 8.7% with five non-contributing stores

1 H I G H L I G H T S Group revenue of 69.2m, down 8.7% with five non-contributing stores

H A L F Y E A R R E S U L T T O 3 1 S T J U L Y 2 0 1 6 1 H I G H L I G H T S Group revenue of 69.2m, down 8.7% with five non-contributing stores closed during the period Strong LFL performance with UK/Europe stores up 6.5%.

265 views • 12 slides
Iceland Foods UK Wales Number of Iceland Stores 865 57 Number of Food Warehouse Stores 3 12

Iceland Foods UK Wales Number of Iceland Stores 865 57 Number of Food Warehouse Stores 3 12

Iceland Foods UK Wales Number of Iceland Stores 865 57 Number of Food Warehouse Stores 3 12 Number of Employees 1,790 20,215 - retail 928 19,353 - head office 862 862 Online coverage 84% 82% Numbered of Delivered Sales (p.a.)

531 views • 26 slides
Scaling Log-Structured KV-Stores featuring Monkey and Dostoevsky SIGMOD17 / SIGMOD18 Niv Dayan

Scaling Log-Structured KV-Stores featuring Monkey and Dostoevsky SIGMOD17 / SIGMOD18 Niv Dayan

Scaling Log-Structured KV-Stores featuring Monkey and Dostoevsky SIGMOD17 / SIGMOD18 Niv Dayan Log-Structured KV-Stores Log-Structured KV-Stores Why Log-Structured KV-Stores? Why Log-Structured KV-Stores? fast writes Why Log-Structured

2.34k views • 198 slides
The Quest for Non-Functional Property Optimisation in Heterogeneous and Fragmented Ecosystems: a

The Quest for Non-Functional Property Optimisation in Heterogeneous and Fragmented Ecosystems: a

The Quest for Non-Functional Property Optimisation in Heterogeneous and Fragmented Ecosystems: a Distributed Approach Mahmoud A. Bokhari, Markus Wagner and Brad Alexander givenname.familyname@adelaide.edu.au Maslows hierarchy of needs

417 views • 18 slides
Sample Amplification: Increasing Dataset Size even when Learning is Impossible Brian Axelrod

Sample Amplification: Increasing Dataset Size even when Learning is Impossible Brian Axelrod

Sample Amplification: Increasing Dataset Size even when Learning is Impossible Brian Axelrod Shivam Garg Greg Valiant Vatsal Sharan 0 0 0 2 , 4 3 $ o r f r s u Y o What does it mean that a GAN made this image? (Does it mean

458 views • 20 slides
Single Sample t Test Assessment of differences between groups t comp & t crit t crit

Single Sample t Test Assessment of differences between groups t comp & t crit t crit

1/15/2018 Single Sample t Test Assessment of differences between groups t comp & t crit t crit values come from back of text t crit = REQUIRED t t comp values come from your brain t comp = ACTUAL t Alternative

386 views • 7 slides
1 Tyrosine Tyrosine TH [ 18 F] FDOPA DOPA DDC Amphetamine Reserpin DA Tetrabenazine DAT [

1 Tyrosine Tyrosine TH [ 18 F] FDOPA DOPA DDC Amphetamine Reserpin DA Tetrabenazine DAT [

NSUH NSUH NSUH NSUH NSUH NSUH 1 Tyrosine Tyrosine TH [ 18 F] FDOPA DOPA DDC Amphetamine Reserpin DA Tetrabenazine DAT [ 18 F] FPCIT DOPAC DA [ 123 I] CIT, FPCIT HVA [ 11 C] cocaine, d-threo-methylphenidate D2 D1 NSUH NSUH

326 views • 6 slides
ADHD PHARMACOLOGY University of Hawaii Hilo Pre -Nursing Program NURS 203 General

ADHD PHARMACOLOGY University of Hawaii Hilo Pre -Nursing Program NURS 203 General

ADHD PHARMACOLOGY University of Hawaii Hilo Pre -Nursing Program NURS 203 General Pharmacology Danita Narciso Pharm D 1 Understand what happens in filter & gain under normal circumstances and how that translates to ADHD

613 views • 24 slides
Mitigating Gender Bias Amplification in Distribution by Posterior Regularization Shengyu Jia *

Mitigating Gender Bias Amplification in Distribution by Posterior Regularization Shengyu Jia *

Mitigating Gender Bias Amplification in Distribution by Posterior Regularization Shengyu Jia * , Tao Meng * , Jieyu Zhao , Kai-Wei Chang Tsinghua University University of California, Los Angeles Credit to Mark Yatskar Credit

162 views • 13 slides
Ring Amplifiers for Switched Capacitor Circuits Benjamin Hershberg 1 , Skyler Weaver 1 , Kazuki

Ring Amplifiers for Switched Capacitor Circuits Benjamin Hershberg 1 , Skyler Weaver 1 , Kazuki

Ring Amplifiers for Switched Capacitor Circuits Benjamin Hershberg 1 , Skyler Weaver 1 , Kazuki Sobue 2 , Seiji Takeuchi 2 , Koichi Hamashita 2 , Un-Ku Moon 1 1 Oregon State University, Corvallis, OR, USA 2 Asahi Kasei Microdevices, Atsugi, Japan

817 views • 65 slides
Amplification by Shuffling: From Local to Central Differential Privacy via Anonymity Vitaly

Amplification by Shuffling: From Local to Central Differential Privacy via Anonymity Vitaly

Amplification by Shuffling: From Local to Central Differential Privacy via Anonymity Vitaly Feldman Ulfar Erlingsson Ilya Mironov Ananth Raghunathan Kunal Talwar Abhradeep Thakurta Local Differential Privacy (LDP) 1 1 For all ,

151 views • 11 slides
Privacy Amplification by Mixing and Diffusion Mechanisms Borja Balle, Gilles Barthe, Marco

Privacy Amplification by Mixing and Diffusion Mechanisms Borja Balle, Gilles Barthe, Marco

Privacy Amplification by Mixing and Diffusion Mechanisms Borja Balle, Gilles Barthe, Marco Gaboardi, Joseph Geumlek Amplification by Postprocessing x 1 , , x n y 1 y 2 M K Post-processing Mechanism (Markov operator) Amplification by

338 views • 15 slides

More recommend