This session is intended to do a deep dive into the Power BI Service and infrastructure to ensure that you are able to monitor your solution before it starts performing or when your users are already complaining.As part of the session i will give advise you on how to address the main pains causing slow performance by answering the following questions:
* What are the components of the Power BI Service?
- DirectQuery
- Live connection
- Import
* How do you identify a bottleneck?
* What should i do to fix performance?
* Monitoring
- What parts to monitor and why?
* What are the report developers doing wrong?
- how do i monitor the different parts?
* Overview of best practices and considerations for implementations
Call Girls In DLf Gurgaon ➥99902@11544 ( Best price)100% Genuine Escort In 24...
SQL Bits 2018 | Best practices for Power BI on implementation and monitoring
1. Best practices for Power BI on implementation
and monitoring
SQL Bits 2018
2. Who am I?
• Bent Nissen Pedersen, Denmark
• Business Analytics Architect, Kapacity
• @SQLBent
• bnp@kapacity.dk
• BI for 10+ years
• www.biblog.eu
3. Agenda
• Monitoring
• Bottlenecks
• Components
• Direct-query
• Live Connection
• Import
• What should i do to fix
performance?
• What are the report developers
doing wrong?
12. Usage Performance Improvement
Breakdown of Power BI request
How queries are handled and executed
Send request
Process
request and
construct
query
Query travel
time to data
source
Query is
evaluated and
result set is
returned
Result set
travel time
Result set
processed and
HTML is
generated
HTML travel
time to end
user
Network related Power BI related Data source related
13. Usage Performance Improvement
Breakdown of Power BI request
How queries are handled and executed
Send request
Process
request and
construct
query
Query travel
time to data
source
Query is
evaluated and
result set is
returned
Result set
travel time
Result set
processed and
HTML is
generated
HTML travel
time to end
user
Network related Power BI related Data source related
14. Usage Performance Improvement
Understanding live connections
Using the gateway
Microsoft Cloud
On-premise Datacenter
User opens report on PowerBI.com
Power BI Enterprise Gateway (PBI EG) collects request to query on-prem SSAS
PBI EG connects to SSAS using service account.
Connectionstring includes ”EffectiveUserName=user@domain.com”
Query is executed with required security context
1
6
2
5 4
3
1
3
2
4
6
5 PBI EG relays the result-sset to PBI
PBI returns rendered visualizations to end-user
15. Usage Performance Improvement
Data source latency
Direct query through on-premises gateway
SSAS MD
SQL Profiler
xEvents
SSAS Tabular
DAX Studio
SQL Profiler
VertiPaqAnalyzer
Tabular Editor
xEvents
SQL
SQL Profiler
SQL – Column store indexes
16. Usage Performance Improvement
Microsoft Cloud
On-premise Datacenter
Schedule triggers update of Power BI file with imported data
Power BI Enterprise Gateway (PBI EG) collects request and starts meshup
File are loaded, transformed and compressed for each dataset
Meshup result is sent to PowerBI backend
2
4
3
1
3
2
4
1
Understanding data import
Using the gateway
17. Usage Performance Improvement
Monitoring
What to do about your gateway ?
Ressources
Monitoring of CPU and memory is
recommend to be done using
performance counters
Above 80% utilization of either memory
or CPU usually cause decreased
performance
Workloads
Monitoring the workload should happe
using Power BI specific performance
counter.
Split live query and data import effort to
different gateways and change
concurrent mesh-up executions
Availability
Ensure that multiple gateways are setup
to support high Availability in a cluster
Monitor setup and ensure that at least
one gateway is available for each
cluster.
18. Usage Performance Improvement
Microsoft Cloud
1
4
3
2
Understanding live connections
Using Azure Analysis services
User opens report on PowerBI.com
PBI EG connects to SSAS using service account.
Connectionstring includes ”EffectiveUserName=user@domain.com”
Query is executed with required security context and returns data to PBI1
2
4
4 PBI returns rendered visualizations to end-user
22. Usage Performance Improvement
Monitoring
What to do about your Azure AS?
Ressources & workload
Use metrics to monitor and alert on
situations where ressources are sparse
or not avaialble.
Scale out or up to support requirements.
Availability
High availability is not directly possible
since no automated sync can be in
place between regions or datacenters.
An alias can be set up to avoid changes
of connections.
23. Usage Performance Improvement
Network
What and how?
Mitigate with ExpressRoute Premium or by rethinking architecture
Focus first on latency Data Source <-> Power BI
One client request can lead to multiple queries
Potentially move data source to VM in same Azure region as Power BI Backend cluster
Data Source ClientPower BI
Data Source ClientPower BI
24. Usage Performance Improvement
Power BI architecture
How queries are handled and executed
Requests from one report page
Query execution time factors
• Model & measure complexity
• Tables relations
• Security
• Number of users
26. What is Power BI Premium?
Usage Performance Improvement
Power BI PremiumPower BI Cluster
PBI Tentant 1
PBI Tentant 2
PBI Tentant 3
PBI Tentant 4
Personal
Workspace
Workspace
Workspace
Workspace
Workspace
Personal
Workspace
Personal
Workspace
Workspace
Workspace
Workspace
27. Power BI Premium
Log Metrics
Usage Performance Improvement
CPU Usage
As for tabular models this is used for
mashup of data and query execution.
Memory
Models are loaded into memory and
trashed whenever there is not enough
memory.
Direct Query
Limitations apply to number of live
connection queries per second.
P1 have a combined limit of 30 per
second for cloud and on-premises.
... Setup extra capacity
29. Usage Performance Improvement
Power BI in the Enterprise
Speed Monitoring and
quality surveillance
Power BI
(Self-service)
Exploration
Flexibility
Power BI
(Enterprise)
Consolidated
business logic
Training and support
Documentation
30. Governance
Usage Performance Improvement
Include admins
Ensure that the super users get use to
include administors in the setup
Monitor success
(and failure)
Engage with the owners of the most
used reporting content to ensure
lifecycle and have it lifted to the
enterprise platform.
Slow reports might be improved and
optimized to reduce performance
impact.
Gateways
Setup, update and monitor gateways for
users to utilize new functionality and
improved mashup.
31. Governance
Usage Performance Improvement
Layout
What not to do; or how you improve
Performance
You know how to debug and what to look
into.
0. Monitoring
1. Gateway load
2. Data source performance
3. Network
4. Layout
Improvements
Considerations for improving your
governance and setup to embrace
Power BI.
33. Ressources
• Power BI documentation
• Power BI Blog
• Power BI Security Whitepaper
• Demos.microsoft.com
• Power BI Template for Power BI API (Prathy)
• Analysis Services Performance whitepaper
34. Just like Jimi Hendrix …
We love to get feedback
Please complete the session feedback forms
35. SQLBits - It's all about the community...
Please visit Community Corner, we are trying this year to get more
people to learn about the SQL Community, equally if you would be
happy to visit the community corner we’d really appreciate it.
Hinweis der Redaktion
Logging is not considers as GDPR vialations since it is metadata
How many people are using Power BI towards on-premises source?
Towards Azure Analysis Services?
Mostly import?
Success -> high usage numbers
Share the message and details of used reports
Premium support to reduce cost and remove old unused content
How many are using the personal gateway ?
(Scalpe up or out)
(Scalpe up or out)
(Scalpe up or out)
Our tests showed latency down from 320 ms to 120 ms India -> North Europe
Customer was satisfied with performance without ExpressRoute, but will consider it in the future
In other words – look for slow performing components same way you are used to.
Latency is for roundtrip
We have 2 roundtrips per refresh
Business
Success -> high usage numbers
Share the message and details of used reports
Premium support to reduce cost and remove old unused content
Success -> high usage numbers
Share the message and details of used reports