The mm-ADT virtual machine is looking to be an economic engine for our open source developers in the data technologies landscape.

1. mm-ADT
A Virtual Machine/An Economic Machine
Dr. Marko A. Rodriguez
DataDays
Austin TX, January 2020
Database Keynote

2. It’s 2012
The Mayan Calendar predicts the end of the world.
The OSS community predicts a Golden Age of Data Technologies.
NoSQL

3. Document Database
Graph Database
Wide Column Store
Index Store
OLAP Processor

4. Open Source Software
costs money and time for
research and development.
In order to provide the world
free open source software many of our
young innovators created companies to fund
the development of their technology.

5. Apache

6. A Collaborative Ecosystem
of Shared Technologies
develops
uses uses

7. uses
develops
develops
A Collaborative Ecosystem
of Shared Technologies

8. develops
uses
A Collaborative Ecosystem
of Shared Technologies

9. develops
uses
A Collaborative Ecosystem
of Shared Technologies

10. The NoSQL
Summer of Love
2012uses
uses
uses
uses
uses
develops
develops
develops
develops
develops

11. I compare us to the bridge, dam, and
interstate builders of the 1930’s generation.
This generation of developers built the modern
world’s software data infrastructure.
By making our software freely
available, many institutions that
would otherwise not pay for
cutting-edge software have
been able to advance their
domain.
We did something very good for this world.

12. And then came the parasites…

14. Amazon’s use of
Apache TinkerPop
for their internal operations
is a perfect use of
open source software.
Amazon gets free software
for their shipping logistics.
Apache TinkerPop is
tested at extreme scale on a
world-class problem.
DataStax, who were the primary
developers at the time, is able to point
(potential) customers to a success story.
The Good

15. The Bad

16. Amazon “Acquires” OSS Companies for Free
“Does Amazon contribute to these projects (benefitting users and the creators)?”
Let’s see what the statistics say…
The Ugly

17. The top contributors do not work for Amazon
MongoDB MongoDB
MongoDB MongoDB

18. The top contributors do not work for Amazon
DataStax RReduX
RReduX <unknown>

19. The top contributors do not work for Amazon
DataStax DataStax
DataStax Oracle

20. The top contributors do not work for Amazon
DataBricks DataBricks
DataBricks DataBricks

21. The top contributors do not work for Amazon
Elastic Elastic
Elastic Elastic

22. 1 2 5 10 20 50 100
5205020010005000
cassandra
frequency
commits
1 2 5 10 20 50 100
20502005002000
elasticsearch
frequency
commits
1 2 5 10 20 50 100
502005002000 mongodb
frequency
commits
1 2 5 10 20 50 100
501002005001000
spark
frequency
commits
1 2 5 10 20 50 100
15505005000
tinkerpop
frequency
commits
α = 2.074 α = 1.69
α = 1.95
α = 2.19 α = 2.717
Smaller α means
fewer people do most of the work.
log/log-scale

23. Yes, AWS promises to contribute back to Cassandra. The
question is whether those contributions be more than self-
serving ones … there are no incentives to make real progress.
To Amazon’s credit, they already made a single major
contribution in the form of the Dynamo paper, 12 years ago.
https://www.scylladb.com/2019/12/04/managed-cassandra-on-aws-our-take/

24. Let’s honor some other
soldiers of the trade.

25. The top contributors do not work for Amazon
RedisLabs RedisLabs
CarrierDB Alibaba

26. The top contributors do not work for Amazon
Cloudera Salesforce
<unknown> Xiaomi

27. The top contributors do not work for Amazon
Percona MariaDB
MariaDB<unknown>

28. 1 2 5 10 20 50 100
5205020010005000
hbase
frequency
commits
1 2 5 10 20 50 100
5205020010005000
mariadb
frequency
commits
1 2 5 10 20 50 100
25201005002000 redis
frequency
commits
α = 1.778 α = 1.713
α = 1.682

30. It used to be…

31. Contribute nothing, take everything.
July 2016

32. Now it’s the…

33. MutuallyExclusive?

34. Amazon will continue to release products
based on Apache Software Foundation technology.
When Amazon has completed their exploitation and
little competition exists they will use legal techniques to
dismantle the Apache Software Foundation.
Open source developers will use less liberal licenses.
Innovation in our industry will come to a standstill.
DataDays will be subsumed by re:Invent.
The Apache Software Foundation had $1 million
in their coffers for legal related issues in 2017.
(prior to Amazon’s 2018 $1 million donation)
Predictions

35. All popular open source technologies will be taken by Amazon.
All popular closed source technologies will be cloned by Amazon.
Small agile data technology companies will struggle.
The majority of jobs in data technology
systems engineering will be at Amazon.
Salaries in data technology will drop preciptously.
Amazon employees will try to create labor unions.
Amazon will squash any unionization efforts.
Their employees will be stuck in legal quagmires.
Predictions
U.S. tax payers will be left to foot the bill on the most expensive
anti-trust suit the U.S. government has ever endured.

36. The systematic desertification of our industry
in the pursuit of profits as opposed to
the pursuit of innovation in our trade.

37. Oh father, love of mine,
you have built such wonderful things
with all your time.
In honor of your greatness,
we followed behind and now of age,
we give not what was mine.
May our software make your machines
sing to the heavens for all time.
A NoSQL Poem for
Amazon/IBM/Microsoft

38. The Virtual Machine

39. Modern Data System Components
Storage Processor Language

40. Synthetic Data Systems
Storage Processor Language

41. mm-ADT
Virtual Machine
Storage
Processor
Language

42. Storage
ProcessorLanguage
mm-ADT
Virtual Machine
programs
reads writes

43. Storage
Processor
Language
mm-ADT
Virtual Machine
int
bool
strreal
lst
rec
map
filter reduce
branch
f.g(h.i).j.k
Model Storage w/ Types
obj
ModelLanguagesw/Instructions
Model Processors w/ Operators
sideeffect
initial terminal
barrier

44. g.V(1).out(‘knows’).id()
int{*} <=[=graph]
[is,[get,’id’][eq,1]]
[get,’outE’]
[is,[get,’label’][eq,’knows’]]
[get,’inV’]
[get,’id’]
graph -> [graph -> vertex{*}
|vertex -> [‘id' :int,
’inE’ :edge{*},
’outE’:edge{*}]
|edge -> [‘label’:str,
’outV’ :vertex,
’inV’ :vertex]]
Storage
Processor
Language
mm-ADT
Virtual Machine
initial [=graph]
filter [is,[get,’id’][eq,1]]
map [get,’outE’]
filter [is,[get,’label’][eq,’knows’]]
map [get,’inV’]
map [get,’id’]
processor supports requisite operators
language compiles to the graph model
storage exposes graph model
[ a->b | c->d | e->f ]
pattern matching
“if lhs, then rhs”

45. mm-ADT is about creating models in terms of other models.
The base model is always mm.
The ‘domain of discourse’ of the mm-ADT VM.
Commonly used models are “standardize” by mm-ADT.
mm : multi-model
kv : key/value
rdb: relational
pg : property graph
wc : wide column
doc: documents
rdf: resource description framework
Storage providers expose their system via these models.
Language providers write compilers to these models.

46. A user’s schema is a model…
social -> [ person -> ['ssn' :int,
'friends':person{*}]]
DSLs

47. social->[person -> ['ssn' :int~x,
'friends':person{*}
A user’s schema is a model…
…that can be embedded in other models.
social->[person -> ['ssn' :int~x,
'friends':person{*}
social model => relational model => mm-ADT model
DSLs
social -> [ person -> ['ssn' :int,
'friends':person{*}]]
social model => graph model => mm-ADT model

48. social->[person -> ['ssn' :int~x,
'friends':person{*} <=(vertex{*} <=[=graph][is,[get,'id'][eq,x]]
[get,'outE']
[is,[get,'label'][eq,'knows']]
[get,'inV'])]
<=(vertex <=[=graph][is,[get,'id'][eq,x]])
A user’s schema is a model…
social model => graph model => mm-ADT model
…that can be embedded in other models.
social->[person -> ['ssn' :int~x,
'friends':person{*} <=(row{*} <=[=rdb][get,'people'][is,[get,'id']
[eq,[=rdb][get,'links']
[is,[get,'source'][eq,x]]
[get,'sink']]])]
<=(row <=[=rdb][get,'people'][is,[get,'id'][eq,x]])
social model => relational model => mm-ADT model
DSLs
social -> [ person -> ['ssn' :int,
'friends':person{*}]]

49. The Economic Machine

50. The mm-ADT virtual machine
serves as a universal integrator for the technologies of our industry.
The mm-ADT economic machine
serves as a universal integrator for the creators in our industry.
mm-ADT

51. Independent open source developers spend their time developing software that
provides no financial compensation and, paradoxically, as the popularity of their
project increases, success can lead to burnout and financial ruin.
Entrepreneurial open source developers may form a company around their open
source project. Unfortunately, such companies are easily obviated in the market
because the core product is freely licensed and competing firms can offer the
same product without having to invest in the research and development effort.
Corporate open source developers are in the problematic situation where their
work is dependent on their sponsoring organization who also controls the
direction and pace of development as well as the interoperability and inclusion of
any novel innovations within the project.
Entrepreneurial Open Source Developers Lack a Competitive Edge
Corporate Open Source Developers Sacrifice Autonomy
Independent Open Source Developers Suffer Economically

52. The OSS Developer
…is an artist, not an employee.
…needs an agent, not a employer.
…seeks to mentor, not to manage.
…should be patronized, not invested in.
…needs to be financially
compensated for their work.

53. A Royalty-Based Open Source Model
Data Technology Developers
Think
Etsy+Apache A Non-Growth Company
contractor

54. A Royalty-Based Open Source Model
mm-ADT
mm-ADT
mm-ADT
design: Specify the component mm-ADT model.
develop: Engineer the component binding.
price: Decide the cost for licensing the component.
promote: Write blog posts, academic articles, tutorials.
maintain: Fix bugs and update over mm-ADT VM versioning.
support: Mailing list Q&A, documentation, issue tracker.
An mm-ADT component
is created by an
open source development team

55. A Royalty-Based Open Source Model
mm-ADT
Virtual Machine
MicroCyberRobo Corp

56. A Royalty-Based Open Source Model
mm-ADT
Virtual Machine
Design
MicroCyberRobo Corp

57. A Royalty-Based Open Source Model
mm-ADT
Virtual Machine
Develop
MicroCyberRobo Corp

58. A Royalty-Based Open Source Model
mm-ADT
Virtual Machine
Compose
MicroCyberRobo Corp

59. A Royalty-Based Open Source Model
mm-ADT
Virtual Machine
Purchase
MicroCyberRobo Corp

60. A Royalty-Based Open Source Model
mm-ADT
Virtual Machine
Payout
MicroCyberRobo Corp

61. A Royalty-Based Open Source Model
mm-ADT
Virtual Machine
Design Develop Compose Purchase Payout
MicroCyberRobo Corp

62. Fin.
http://mm-adt.org