The pandemic changed media forever — trends in how we create, atomize, store, and distribute content have accelerated as consumer patterns radically shifted.
Now, viewers expect their content to be personalized, on-demand, and on the device of their choosing but this can’t be done effectively using traditional SQL and relational models.
Discover how to quickly transform your content management systems, while reducing costs and increasing productivity with MongoDB and Nuxeo.
In this interactive session we go over:
- Top trends in the media and entertainment industry, and how to stay ahead of them
- Best practices to power gigantic repositories of complex content
- How to achieve performance improvement and cost reduction at the same time
11. From 1B, 2B, 3B
to 11B documents
Test on production
ready environment
Make it real
Scale infrastructure step by step Nuxeo Cloud: AWS + MongoDB Atlas
Encryption and backup are on
Full content model with document
types and full-text; Generate
consistent and clean data; Build a
realistic content application
2020 Benchmark Goals: MongoDB + AWS
Benchmark an application in a realistic environment
13. Atlas: Benchmark Challenges
Import 11B documents
Leverage NVMe to maximize IOPS
Scale Up to reduce memory pressure
Scale Out / Sharding to spread the load
Cost Efficiency
Tiered Storage / Application-level sharding
Scale Up for Import / Down for normal operations
Leverage MongoDB Atlas Elasticity / Transparency
15. US-west
0.8 B docs
M60 NVMe
State #1 State #26
Customer
#1
Customer
#n
US-east
State #27 State #n State #52
DocType: Workspace
MimeType: N/A
Cardinality: 52
Customer
#1
DocType: Customer
MimeType: N/A
Cardinality: 45M *2
DocType: IDCard
MimeType: jpeg
Cardinality: 1 per customer
DocType: Account
MimeType: N/A
Cardinality: ~2 per customer
DocType: Correspondence-State#n
MimeType: docx
Cardinality: 1 per customer
DocType: Statement
MimeType: PDF
Cardinality: 6 for each account
Customer
#n
Account
#1
Account
#n
0.8 B docs
M60 NVMe
US-east US-west
archives
11.8 B docs
r5x large*16
Non-Sharded Non-Sharded
Sharded
DocType: Statement
MimeType: PDF
Cardinality: 54 for each account
9.7 B docs
M80*5
archives
Sharded
Bulk Move
(Archive)
Bulk Move
(Archive)
16. Atlas: 11B docs Benchmark Results
CRUD / REST API
Response time < 220 ms
Throughput > 1,900 req/s
Navigation
Response time < 500 ms
Throughput > 800 req/s
Search (FullText + Facets)
Response time < 600 ms
Throughput > 400 req/s
Bulk Import (+3B)
Throughput > 25,000 docs/s
(1.5M / minute)
Bulk Indexing (+3B)
Throughput > 57,000 docs/s
(3B in 14h)
Bulk Import + Index (+5B)
Throughput > 15,000 docs/s
17. Scale from 0 to
11B Documents
Scale up / out progressively
M40 => M60 => M80 => M200
Maintain good performances
Leverage PaaS
transparency
Scale up for Bulk Import
Scale up for Bulk Re-Index
Cost efficiency
Scale down to reduce cost
11B docs Benchmark: Takeaways
19. Document
Repository
Context: MongoDB @ Nuxeo
Nuxeo Content Services
Audit KVS Directories
Collections Object Store
MongoDB Elastic
Indexes
Event Records
Async/Tmp
Results Store
User / Groups
Ontology
Meta-data
Hierarchy
ACLs
Binaries
20. Why MongoDB
and Nuxeo
Write throughput
Json Document. No lock escalation.
Scalability
Scale-Out. Sharding.
No Impedance Issue
More performance. Less code to maintain.
21. Nuxeo
MongoDB
Scale up and Scale out
Data-level Sharding
Single source of truth
for Nuxeo structured data
Leverage Cloud PaaS
Scale out and AutoScaling
Application-level Sharding
Single source of truth
for all your content
Leverage Cloud deployment
23. Pace of Modern Media New CMS on MongoDB Forbes.com traffic ⇡
Against legacy systems, outages
and costly changes
Mobile reader experience
Distributed content team
Viewer clickstream @scale
5% → 60% on mobile
Reader behavior drives next-best
content
Forbes
25. Major “household” Film Production Company
Roadblock
Multiple contradictory approvals
processes for licensed goods.
Benefits
Search across 350,000 assets for,
harvest, and license digital assets from
the creative process
Time Saved
Integration with existing workflow tool
opened system to 2500 users across the
licensor network. Whole processes now
managed by just two DAM librarians.
Impact
Faster time to market of licensed goods
to tie-in with movie launch campaigns.
Increased licensing revenue.
27. Thank you
Thierry Delprat | CTO | Nuxeo
Lisa McIntyre | Product Manager | Nuxeo
Sig Narvaez | Principal Architect | MongoDB
28. Industry Conversation: Topics/Questions
1. Asset access
a. Leveraging legacy content
2. IP licensing
a. Warner StarLabs (not public ref)
b. Nike New content from old content
3. Stock footage management
a. (Fox & Disney) eliminates re-shoot, increase prod, $100k/month
b. Up to 80% of a TV show can be stock
4. Content Metadata at-scale
5. Rights management / omni-channel (electronic billboards)
a. Disney Wonderland (internal name)
b. Manage derivatives / translations / rights for x-time, location, locale, etc.