Predictive intelligence from machine learning has the potential to change everything in our day to day experiences, from education to entertainment, from travel to healthcare, from business to leisure and everything in between. Modern ML frameworks are batch by nature and cannot pivot on the fly to changing user data or situations. Many simple ML applications such as those that enhance the user experience, can benefit from real-time robust predictive models that adapt on the fly.
Join this session to learn how common practices in machine learning such as running a trained model in production can be substantially accelerated and radically simplified by using Redis modules that natively store and execute common models generated by Spark ML and Tensorflow algorithms. We will also discuss the implementation of simple, real-time feed-forward neural networks with Neural Redis and scenarios that can benefit from such efficient, accelerated artificial intelligence.
Real-life implementations of these new techniques at a large consumer credit company for fraud analytics, at an online e-commerce provider for user recommendations and at a large media company for targeting content will also be discussed.
5. Redise Cloud Private
Redis Labs Products
Redise Cloud Redise Pack ManagedRedise Pack
SERVICES SOFTWARE
Fully managed Redise service in
Virtual Private Clouds (VPC) within
Amazon (AWS), Microsoft Azure,
Google Cloud,
IBM Softlayer
Fully managed Redise service on
hosted servers within public clouds:
Amazon (AWS), Microsoft Azure,
Google Cloud, IBM Softlayer,
Heroku, Cloud Foundry &
RedHat OpenShift
Downloadable Redise software for
any enterprise datacenter or cloud
environment
Fully managed Redise software in
any private data center
&& &
6. Mature and Stable Technology & Products
250K+ 600+
100 + 1350+
1,000+
DATABASES RUN
OVER 3 YEARS
NEW DATABASES
CREATED EVERY DAY
MAN-YEARS OF ENTERPRISE
REDIS TECHNOLOGY
DEVELOPMENT
CLOUD NODE FAILURE AND
OUTAGES EVENTS
SURVIVED WITH NO DATA
LOSS
GRANTED AND PENDING
PATENTS
DEDICATED REDIS
ENGINEERS
16. Redis in Machine Learning
REDIS-ML
• Models can be stored, retrieved and
updated natively with Redis-ML
• Accelerates predictions serving for
complex ML model by x100
• Unified and simplified ML serving
operation – HA, Scale, Perf and more
Save Machine Learning
ModelRead Training Data
Analytics
Apps
Serve Predictions and
Recommendations
Train the Machine Learning Model
1
2
3
4
22. ML Models Serving Challenges
o But then…
1. ML Models are becoming bigger in size as they get more precise and complex!
2. Serving recommendations in mission critical apps - Scaling, Performance, HA & DR…
3. How do you manage multiple model types (Random Forest, Gradient Boosted Trees,
Logistic Regression, etc.)
4. How do you manage multiple versions of each model
5. How do you upgrade a model across so many machines
6. What if the training and serving apps are written in different languages
7. And so on….
Homegrown
24. Large/Accurate Models are Expensive to Serve !
Item Calculation Total
Random Forest
ops/sec
20K (ads/sec) x
1K (forests) x
15K(trees) x
7 x 0.5 (levels)
1.05 trillion ops/sec
Max ops/sec on the
strongest AWS instance
vcore
2.6Ghz x
0.9 (OS overhead) x
0.1 (10 lines of code per ops) x
0.1 (Java overhead)
23.4 million ops/sec
# of vcores needed 1.1 trillion / 23.4 million 44,872 vCores
# of c4.8xlarge
instances needed
44,872 / 36 1,247 Instances
Total Cost
in Reserved
Instances
1,247 x 9213 ~$11.5M/year
31. Neural Redis
o A very simple to use API.
o Automatic data normalization.
o Online training of neural networks in different threads.
o Ability to use the neural network while the system is training it (we
train a copy and only later merge the weights).
o Fully connected neural networks using the RPROP (Resilient back
propagation) learning algorithm.
Get Started!
> redis-server --loadmodule /path/to/neuralredis.so
https://github.com/antirez/neural-redis
37. Resources
o Getting Started with Redis and Redise
• https://hub.docker.com/r/redislabs/redis/
o Getting Started with Redis and Machine Learning
• https://redislabs.com/modules/machine-learning/
o Other Resources
• Redis-ML : https://github.com/RedisLabsModules/redis-ml
• Spark-Redis-ML : https://github.com/RedisLabs/spark-redis-ml
• Neural Redis : https://github.com/antirez/neural-redis
• Databricks Notebook – Spark & Redis : http://bit.ly/sparkredisml