Ancestry has more than three million paying subscribers and a collection of more than 20 billion records. The company moved its operations to the cloud to enable it to scale with its customer base and implemented CI/CD processes to facilitate rapid feature rollout. However, with rapid growth in both its user base and the range of products it provides, Ancestry was hard-pressed to ensure that it was achieving optimum performance, efficiency, and customer experience with its cloud applications, while also efficiently spending their cloud budget. With tens of thousands of components processing petabytes of data, there is a lot of room for underutilization and wastage. That leads into the cloud optimization presentation with Amir Sharif from Opsani, who will review the way Ancestry has leveraged AI to optimize cloud operations. Learn in this session how Ancestry and some of the largest enterprises have autonomously optimized thousands of their workloads across their service delivery platform, saving countless human hours and dollars, while delivering a better customer experience.

1. © 2019-2020 Opsani. All Rights Reserved.
Continuous Optimization as a Service
© 2019-2020 Opsani. All Rights Reserved.

2. © 2019-2021 Opsani. All Rights Reserved.
Ancestry has more than 3,000,000 paying subscribers and a collection of more
than 20,000,000,000 records.
The company moved its operations to the cloud to enable it to scale with its
customer base and implemented CI/CD processes to facilitate rapid feature
rollout.
Ancestry was hard-pressed to ensure that it was achieving optimum
performance, efficiency, and customer experience with its cloud applications,
while also efficiently spending their cloud budget.
With tens of thousands of components processing petabytes of data, there is a
lot of room for underutilization and wastage.
Continuous Optimization as a Service

3. Continuous Optimization as a Service
© 2019-2021 Opsani. All Rights Reserved.
The Digital Transformation Paradox
Why Public Cloud
- Avoids heavy upfront CapEx
- Allows for PAY-GO OpEx
- Makes infra programmable via APIs
- Enables fast, easy provisioning
Why Cloud-Native
- Increases business, eng flexibility
- Kubernetes, GitOps make deploying code easier
- Containers eliminate OS underlay dependencies
- Enables fast scale-up
➔ Optimized infrastructure consumption ➔ Optimized app time-to-market
So why is your app under-performing,
over-provisioned, costing too much?

4. © 2019-2021 Opsani. All Rights Reserved.
Current State of the Art
Continuous Optimization as a Service
pre-prod prod
Status
Quo
Reactive
Tuning
product
owners
developers
code
commit
build test integrations
DevOps SRE
Dev Pre-Prod Production
related
code
real-user
monitoring
synthetic
load
APM
Performance
Lag
Over Provision

5. © 2019-2021 Opsani. All Rights Reserved.
Ancestry CI/CD: Woes to Wows
Continuous Optimization as a Service
CI/CD Tools Proliferation
● Discrete teams adopting
new processes (e.g canary
deployment)
● Each team had their own
tools
● Global changes hard to
make; could not scale best
practices
● Near-impossible governance
for top-priority goals: Cost,
Availability, Productivity
Messy, Old Pipelines
● Jenkins infrastructure
● Each team had own pipeline
● Continuous source code
changes, but not a “hollata”
CI
● Old-school, large
deployments
● Yet, there were teams with
excellent practices, like full
tests on a branch before
releasing
● Nothing standard across the
company
Comparing Canaries
● Some teams could release in
a couple of weeks
● Most teams took
significantly longer
● Some teams had 40 systems.
A few of the systems could
get updated in a sprint. The
full set of systems took “a
while.:
● ~ 70 teams. 10 systems per
on average

6. © 2019-2021 Opsani. All Rights Reserved.
Everyone’s Got One!
Continuous Optimization as a Service
pre-prod prod
Status
Quo
Reactive
Tuning
product
owners
developers
code
commit
build test integrations
DevOps SRE
Dev Pre-Prod Production
related
code
real-user
monitoring
synthetic
load
APM
Performance
Lag
Over Provision
Bespoke!

7. © 2019-2021 Opsani. All Rights Reserved.
Ancestry CI/CD: Woes to Wows
Continuous Optimization as a Service
CI/CD Tools Proliferation
● Discrete teams adopting
new processes (e.g canary
deployment)
● Each team had their own
tools
● Global changes hard to
make; could not scale best
practices
● Near-impossible governance
for top-priority goals: Cost,
Availability, Productivity
Messy, Old Pipelines
● Jenkins infrastructure
● Each team had own pipeline
● Continuous source code
changes, but not a “hollata”
CI
● Old-school, large
deployments
● Yet, there were teams with
excellent practices, like full
tests on a branch before
releasing
● Nothing standard across the
company
Comparing Canaries
● Some teams could release in
a couple of weeks
● Most teams took
significantly longer
● Some teams had 40 systems.
A few of the systems could
get updated in a sprint. The
full set of systems took “a
while.:
● ~ 70 teams. 10 systems per
on average
The Fix
● Commitment: stop
development and spend all
effort to migrate to a
common CI/CD platform
(Harness)
● Analogous to deploying a
new canary for each team.
● Now, the company has a
standardized infrastructure.
Adding new steps into the
pipeline are easy.
● Productivity up.
● Efficiency up.
● Happiness up.

8. © 2019-2021 Opsani. All Rights Reserved.
There is Only One!
Continuous Optimization as a Service
pre-prod prod
Status
Quo
Reactive
Tuning
product
owners
developers
code
commit
build test integrations
DevOps SRE
Dev Pre-Prod Production
related
code
real-user
monitoring
synthetic
load
APM
Performance
Lag
Over Provision
Enterprise Gold Standard

9. Continuous Optimization as a Service
© 2019-2021 Opsani. All Rights Reserved.
The Mutual-Exclusive Fallacy
B
u
d
g
e
t
O
v
e
r
r
u
n
s
P
o
o
r
P
e
r
f
o
r
m
a
n
c
e
Errors, Restarts
Reliability
Performance
Cost
Cloud
Services

10. © 2019-2021 Opsani. All Rights Reserved.
Current State of the Art
Continuous Optimization as a Service
pre-prod prod
Status
Quo
Reactive
Tuning
product
owners
developers
code
commit
build test integrations
DevOps SRE
Dev Pre-Prod Production
related
code
real-user
monitoring
synthetic
load
APM
Performance
Lag
Over Provision
Reactive Build
Manual Ops, Guess Work

11. Case Study
© 2019-2020 Opsani. All Rights Reserved.
Profile
Google Online Boutique is a canonical 11-tier
microservices application. The web-based
e-commerce app trains developers on
containers, microservices, and Kubernetes.
The Challenges
● “Standard” app with pre-set parameters
● 2 configuration parameters: CPU and memory for each service
● 8 possible settings for 22 different tunable parameters
● 822
, or 75 quintillion (73,786,976,294,838,200,000) search permutations for
finding the optimal solution
● 7.5 quintillion grains of sand on Earth
● ML is the only viable method for solving this optimization problem.
Continuous Optimization as a Service

12. 7.5 Quintillion Grains of Sand on Earth
Google

13. 7.5 Quintillion Grains of Sand on Earth
75 Quintillion Permutations for Google Online Boutique
Years left before Sun supernovas
Years to find the answer at 10 microsec / permutation 23 trillion
5 billion

14. © 2019-2021 Opsani. All Rights Reserved.
Ancestry CI/CD: Woes to Wows
Continuous Optimization as a Service
CI/CD Tools Proliferation
● Discrete teams adopting
new processes (e.g canary
deployment)
● Each team had their own
tools
● Global changes hard to
make; could not scale best
practices
● Near-impossible governance
for top-priority goals: Cost,
Availability, Productivity
Messy, Old Pipelines
● Jenkins infrastructure
● Each team had own pipeline
● Continuous source code
changes, but not a “hollata”
CI
● Old-school, large
deployments
● Yet, there were teams with
excellent practices, like full
tests on a branch before
releasing
● Nothing standard across the
company
Comparing Canaries
● Some teams could release in
a couple of weeks
● Most teams took
significantly longer
● Some teams had 40 systems.
A few of the systems could
get updated in a sprint. The
full set of systems took “a
while.:
● ~ 70 teams. 10 systems per
on average
The Fix
● Commitment: stop
development and spend all
effort to migrate to a
common CI/CD platform
(Harness)
● Analogous to deploying a
new canary for each team.
● Now, the company has a
standardized infrastructure.
Adding new steps into the
pipeline are easy.
● Productivity up.
● Efficiency up.
● Happiness up.

15. © 2019-2021 Opsani. All Rights Reserved.
Reactive vs. Continuous Optimization as a Service
Continuous Optimization as a Service
pre-prod prod
Status
Quo
Reactive
Tuning
Opsani
COaaS
product
owners
developers
code
commit
build test integrations
DevOps SRE
Dev Pre-Prod Production
related
code
real-user
monitoring
synthetic
load
APM
Performance
Lag
Over Provision
Reactive Build
Manual Ops, Guess Work
product
owners
developers
code
commit
build test integrations
DevOps SRE
Dev Pre-Prod Production
related
code
autonomous
workload tuning
higher
performance
lower
costs
AIOps
= +
GitOps: Optimized Config Changes & Resourcing Real Time Autonomous Config Changes & Resourcing ⬅ Shift Left
AIOps

16. Performance
Efficiency Gains
230%
Case Study
© 2019-2020 Opsani. All Rights Reserved.
Opsani Tuning Results
Company Profile
Ancestry is a $1B+ enterprise with millions of subscribers and 27,000,000,000
records. The company is the global leader in family history and consumer
genomics, harnesses the information found in family trees, historical records, and
DNA to help people gain a new level of understanding about their lives.
The Challenges
● Successfully implement cloud cost optimization
● Determine efficient runtime settings
● Improve performance predictability
● Protect and improve user experience
● Expedite new code releases
Cloud Cost Reduction
50%+
As the company continues to grow and invest in
new products, our efficiency and performance
are more important than ever. Opsani will allow
us to manage costs, maintain optimal
performance of our cloud resources and gain
visibility in an increasingly complex environment.
“
”
Russ Barnett, Chief Architect
Continuous Optimization as a Service

17. Setup Time
20
Minutes
Tuning Time
2
Days
Throughput Increase
171%
Case Study
© 2019-2020 Opsani. All Rights Reserved.
Opsani AIOps Transformation
Results
Profile
Google Online Boutique is a canonical 11-tier
microservices application. The web-based
e-commerce app trains developers on
containers, microservices, and Kubernetes.
The Challenges
● “Standard” app with pre-set parameters
● 2 configuration parameters: CPU and memory for each service
● 8 possible settings for 22 different tunable parameters
● 822
, or 75 quintillion (73,786,976,294,838,200,000) search permutations for
finding the optimal solution
● 7.5 quintillion grains of sand on Earth
● ML is the only viable method for solving this optimization problem.
Trimmed Cloud Costs
79%
800%
Transactions per Dollar Increase by
Continuous Optimization as a Service

18. Platform Deployment Solution Architecture
Opsani
SaaS Service
Servo
App #n
App #2
App #1
Tuning Pod Tuning Pod
HPA
opsani.com/SLO opsani.com/SLO opsani.com/SLO
Namespace #1 Namespace #2 Namespace #n Opsani
Namespace
Controller
K8s API
Kubernetes Cluster
Transformative Automation
● Auto service discovery
● Auto Onboarding
● Auto SLO setting
● Auto Tuning
● Predictive Scaling
● 1 or 1,000’s of services
Continuous Optimization as a Service

19. Ingress
Metrics:
Request
rate
Error rate
Latency
Pod #1
Pod #n
Operations:
Read deployment config
Create/destroy tuning pod
Pod #2
(https)
K8s. API
Opsani
SaaS Service
Tuning Pod
Deployment
HPA
(Optional)
Servo
Customer’s Cluster -
Selected Namespace
Kubernetes Pattern
Servo
Tuning Pod
Tuning Pod
Servo
Prometheus Pull
Consumer

20. © 2019-2021 Opsani. All Rights Reserved.
COaaS
Application Namespace
● Production pod replica
● Tuning instance
Service Level Objective
● Prescribed; or
● Auto-discovered
perf metrics
config tuning
parameters
CI/CD, Monitoring, and
Collaboration Integrations
Single
Instance
Tuning
Instance
Count
Tuning
Instance
Proactive
Autoscaling
Load
Profile
Adaptive
Tuning
SLO
Opsani Integration and Value
Continuous Optimization as a Service

21. © 2019-2020 Opsani. All Rights Reserved.
Continuous Optimization as a Service
© 2019-2020 Opsani. All Rights Reserved.
Automating Excellence