Bigtable is a distributed storage system for managing large structured datasets. It uses a sparse distributed multidimensional sorted map as its data model. Bigtable has been widely adopted by Google for applications requiring scalability, high performance, and high availability, serving over 60 Google products. It utilizes various technologies like the Google File System for storage and Chubby for coordination.

1. Bigtable
A Distributed Storage System for
Structured Data
Authors: Fay Chang et. al.
Presenter: Zafar Gilani
1

2. Bigtable
Outline
• Introduction
• Data model
• Implementation
• Performance evaluation
• Conclusions
2

3. Bigtable
A distributed storage system ..
• .. for managing structured data.
• Used for demanding workloads, such as:
– Throughput oriented batch processing.
– Serving latency-sensitive data to the client.
• Dynamic control instead of relational model.
• Data locality properties (revisit later briefly).
3

4. Bigtable
Bigtable has achieved several goals
• Wide applicability: used for 60+ Google
products, including:
– Google Analytics, Google Code, Google Earth,
Google Maps and Gmail.
• Scalability (explain later under evaluation).
• High performance.
• High availability.
4

5. Bigtable
Outline
• Introduction
• Data model
• Implementation
• Performance evaluation
• Conclusions
5

6. Bigtable
Data model
• Essentially a sparse, distributed, persistent
multi-dimensional sorted map.
• The map is indexed by a row key, column key
and a timestamp.
• Atomic reads and writes over a single row.
Columns
Rows
6

7. Bigtable
Row and column range
• Row range dynamically • Column keys grouped
partitioned into tablets. into column families.
• Data in lexicographic • Each family has the
order. same type.
• Allows data locality. • Allows access control
and disk or memory
accounting.
7

8. Bigtable
Row and column range
• Row range dynamically • Column keys grouped
partitioned into tablets. into column families.
• Data in lexicographic • Each family has the
order. same type.
• Allows data locality. • Allows access control
and disk or memory
accounting.
Enables reasoning about data locality
8

9. Columns
Rows
9

10. Columns
Rows
Anchor is a column family
10

11. Columns
Tablets
“anchor:bbcworld.com “anchor:weather
“com.bbc.www” “BBC” “BBC.com”
11

12. Bigtable
Outline
• Introduction
• Data model
• Implementation
• Performance evaluation
• Conclusions
12

13. Bigtable
Bigtable uses several other
technologies
• Google File System to store log and data files.
• SSTable file format to store BigTable data.
• Chubby, a distributed lock service.
For more details on these technologies, refer to section 4 of the paper.
13

14. Bigtable
Implementation
Master responsibilities:
-Assign tablets to tablet
servers
-Add/delete tablet servers
-Balance tablet server load
-GC
-Schema changes
INTERNET
CLIENT
Communicate directly
to tablet servers
MASTER
TABLET SERVERS
14

15. Bigtable
How data is stored?
A three-level hierarchy, similar to B+ trees.
15

16. Bigtable
Location hierarchy
Chubby file contains location
of the root tablet.
16

17. Bigtable
Location hierarchy
Root tablet contains all
tablet locations in Metadata
table.
17

18. Bigtable
Location hierarchy Metadata table stores
locations of actual tablets.
18

19. Bigtable
Location hierarchy
Client moves up the hierarchy (Metadata -> Root -> Chubby),
if location of tablet is unknown or incorrect.
19

20. Bigtable
How data is served?
20

21. Bigtable
Tablet serving
Persistent
21

22. Bigtable
Tablet serving
Compactions
Compactions occur
regularly, advantages:
-Shrinks memory usage.
-Reduces amount of data
read from log during
recovery. 22

23. Bigtable
Outline
• Introduction
• Data model
• Implementation
• Performance evaluation
• Conclusions
23

24. Bigtable
Benchmarks for perf evaluation
• Scan:
– Scans over values in a row range.
• Random reads from memory.
• Random reads/writes:
– R keys to be read/written spread over N clients.
• Sequential reads/writes:
– 0 to R-1 keys to be read/written spread over N
clients.
24

25. Bigtable
Performance evaluation
Scan uses single RPC call and
shows best performance.
25

26. Bigtable
Performance evaluation
Sequential reads are better
than random reads, since
each fetched block is used to
serve next requests.
26

27. Bigtable
Performance evaluation
Random read shows the worst
performance. Fetching 64KB
every 1000 bytes is expensive.
27

28. Bigtable
Performance evaluation
Not linear, but scales well.
28

29. Bigtable
Outline
• Introduction
• Data model
• Implementation
• Performance evaluation
• Conclusions
29

30. Bigtable
Conclusions
• Bigtable: highly scalable and available, without
compromising performance.
• Flexibility for Google – designed using their
own data model.
• Custom design gives Google the ability to
remove or minimize bottlenecks.
• Related work:
– Apache Hbase (open source)
– Boxwood (though targeted at a lower/FS level)
30

31. Bigtable
A Distributed Storage System for
Structured Data
Authors: Fay Chang et. al.
Presenter: Zafar Gilani
31

32. Bigtable
B+ Trees
• A tree with sorted data for:
– Efficient insertion, retrieval and removal of
records.
• All records are stored at the leaf level, only
keys stored in interior nodes.
32