This slides deck about Microservices architecture and why do we need it. Architecture patterns which we need to follow doing Microservices architecture: Microservice, API Gateway, Service Discovery, Stateless/Shared-Nothing, Configuration/Service Consumption, Fault Tolerance (Circuit Breaker), Request Collapsing. And a bit about API Versioning
2. 2CONFIDENTIAL
2
• Why do we need it
• Architecture patterns
AGENDA
• Microservice
• API Gateway
• Service Discovery
• Stateless/Shared-Nothing
• Configuration/Service Consumption
• Fault Tolerance
• Request Collapsing
• API Versioning
5. 5CONFIDENTIAL
5
MICROSERVICES VS MONOLITH
Simple code base
Modularity with exact borders
Change circles decoupled
Efficient scaling
Newcomers adopting faster
Per service team responsibility
No technology lock
MONOLITH MICROSERVICES
Complex code base
Hard to maintain modularity
Change circles tightly coupled
Inefficient scaling
Scaring for newcomers
Hard to scale development team
Tied to chose technology
6. 6CONFIDENTIAL
6
MICROSERVICES VALUES
• Heavily relies on Continues Delivery principles
• Fine-grained domain capability
• Team autonomy with responsibility
• Independent release process
• Independent scaling
• System resilience
• Technology variation
7. 7CONFIDENTIAL
7
MICROSERVICES VALUES VS COMPLEXITY
Team autonomy
Time to market
Scaling
Componentization
Technology variation
Cross teams communication
Continues Deployment
Fault tolerance
Versioning
Maintenance
VALUES COMPLEXITY
11. 11CONFIDENTIAL
11
BOUNDED CONTEXT
Bounded Context is a
central pattern in
Domain-Driven
Design. It is the
focus of DDD's
strategic design
section which is all
about dealing with
large models and
teams.
13. 13CONFIDENTIAL
13
DECENTRALIZED DATA MANAGEMENT
Microservices prefer letting each service
manage its own database, either different
instances of the same database technology,
or entirely different database systems - an
approach called Polyglot Persistence.
You can use polyglot persistence in a
monolith, but it appears more frequently
with microservices.
14. 14CONFIDENTIAL
14
DESIGN FOR FAILURE
Distributed systems are
much complex than
monolith.
When we have more
systems there is more
chances to fail.
If more places when you
can fails then more often
you can deal with failures.
15. 15CONFIDENTIAL
15
KEY CONSIDERATION
Before you go into production with a microservices system, you need to ensure
that you have key prerequisites in place
• Rapid Provisioning
• Basic Monitoring
• Rapid Application Deployment
• DevOps Culture
16. 16CONFIDENTIAL
16
MICROSERVICE VS SOA
Martin Fowler
Chief Scientist at ThoughtWorks
Subset of SOA
Zhamak Dehghani
Principal Consultant at ThoughtWorks
Style of SOA
Right picture
31. 32CONFIDENTIAL
32
STORE CONFIG IN THE ENVIRONMENT
The twelve-factor app stores config in environment variables (often shortened to env vars or env). Env
vars are easy to change between deploys without changing any code; unlike config files, there is little
chance of them being checked into the code repo accidentally; and unlike custom config files, or other
config mechanisms such as Java System Properties, they are a language- and OS-agnostic standard.
http://12factor.net
http://12factor.net/config
36. 37CONFIDENTIAL
37
FAULT TOLERANCE
Availability % Downtime per year Downtime per month Downtime per week Downtime per day
90% ("one nine") 36.5 days 72 hours 16.8 hours 2.4 hours
95% 18.25 days 36 hours 8.4 hours 1.2 hours
97% 10.96 days 21.6 hours 5.04 hours 43.2 minutes
98% 7.30 days 14.4 hours 3.36 hours 28.8 minutes
99% ("two nines") 3.65 days 7.20 hours 1.68 hours 14.4 minutes
99.5% 1.83 days 3.60 hours 50.4 minutes 7.2 minutes
99.8% 17.52 hours 86.23 minutes 20.16 minutes 2.88 minutes
99.9% ("three nines") 8.76 hours 43.8 minutes 10.1 minutes 1.44 minutes
99.95% 4.38 hours 21.56 minutes 5.04 minutes 43.2 seconds
99.99% ("four nines") 52.56 minutes 4.38 minutes 1.01 minutes 8.66 seconds
99.995% 26.28 minutes 2.16 minutes 30.24 seconds 4.32 seconds
99.999% ("five nines") 5.26 minutes 25.9 seconds 6.05 seconds 864.3 milliseconds
99.9999% ("six nines") 31.5 seconds 2.59 seconds 604.8 milliseconds 86.4 milliseconds
99.99999% ("seven nines") 3.15 seconds 262.97 milliseconds 60.48 milliseconds 8.64 milliseconds
99.999999% ("eight nines") 315.569 milliseconds 26.297 milliseconds 6.048 milliseconds 0.864 milliseconds
99.9999999% ("nine nines") 31.5569 milliseconds 2.6297 milliseconds 0.6048 milliseconds 0.0864 milliseconds
Without taking steps to
ensure fault tolerance,
30 dependencies each
with 99.99% uptime
would result in 2+ hours
downtime/month
(99.99%30 ≈ 99.7%
uptime = 2+ hours in a
month)
http://techblog.netflix.com/2012/02/fault
-tolerance-in-high-volume.html
0.3% means that the one
million request will have
3000 failed
37. 38CONFIDENTIAL
38
FAULT TOLERANCE
The basic idea behind the circuit breaker
is very simple. You wrap a protected
function call in a circuit breaker object,
which monitors for failures. Once the
failures reach a certain threshold, the
circuit breaker trips, and all further calls
to the circuit breaker return with an
error, without the protected call being
made at all. Usually you'll also want some
kind of monitor alert if the circuit
breaker trips.
CIRCUIT BREAKER
43. 44CONFIDENTIAL
44
REQUEST COLLAPSING
In addition to the isolation
benefits and concurrent
execution of dependency
calls we have also leveraged
the separate threads to
enable request collapsing
(automatic batching) to
increase overall efficiency
and reduce user request
latencies.
Collapse multiple requests into a single execution
based on a time window and optionally a max batch
size.
This allows an object model to have multiple calls to
the command that execute/queue many times in a
short period (milliseconds) and have them all get
batched into a single backend call.
Typically the time window is something like 10ms
give or take.
44. 45CONFIDENTIAL
45
COLLAPSER FLOW
In addition to the isolation
benefits and concurrent
execution of dependency
calls we have also leveraged
the separate threads to
enable request collapsing
(automatic batching) to
increase overall efficiency
and reduce user request
latencies.
Collapse multiple requests into a single execution
based on a time window and optionally a max batch
size.
This allows an object model to have multiple calls to
the command that execute/queue many times in a
short period (milliseconds) and have them all get
batched into a single backend call.
Typically the time window is something like 10ms
give or take.
46. 47CONFIDENTIAL
47
API VERSIONING
• Adding authentication
• Adding authorization rules
• Removing a service
• API contract changes
REASONS SOLUTIONS
• URL Versioning
• Media Type Versioning
• Custom header
• Hostname
• Data parameter
47. 48CONFIDENTIAL
48
API VERSIONING
One method for indicating versioning is via the URI, typically via a path prefix:
Twitter: http://api.twitter.com/1.1/
Last.fm: http://ws.audioscrobbler.com/2.0/
Etsy: http://openapi.etsy.com/v2
Some APIs will provide the version via a query string parameter:
Amazon Simple Queue Service: ?VERSION=2011-10-01
URL
48. 49CONFIDENTIAL
49
API VERSIONING
Media type versioning provides the ability to use the same URI for multiple versions of an API, by specifying the version as part
of the Accept media type.
The Accept header can provide versioning in two different ways:
• As part of the media type name itself: application/vnd.status.v2+json. In this case, the segment v2 indicates the
request is for version 2. You can provide the version string however you desire.
• As a parameter to the media type: application/vnd.status+json; version=2. This option provides more
verbosity, but allows you to specify the same base media type for each version.
Many REST advocates prefer media type versioning as it solves the "one resource, one URI" problem cleanly, and allows
adding versioning support after-the-fact. The primary argument against it is the fact that the version is not visible when
looking at the URI.
MEDIA TYPE
49. 50CONFIDENTIAL
50
API VERSIONING
The above two versioning types are the most common; however, other types exist:
• Custom header. As an example,
• X-API-Version: 2
• GData-Version: 2.0
• X-MS-Version: 2011-08-18
• etc.
• Hostname. Facebook, when migrating from the first API version, switched from the host http://api.facebook.com to
http://graph.facebook.com.
• Data parameter. This could be a query string parameter for GET requests, as noted above, but a content body parameter for
other request methods.
OTHER METHODOLOGIES
50. 51CONFIDENTIAL
51
API VERSIONING
• Typical approach. Include API version. Backwards compatible.
• Some folks use custom content type allowing each resource to have a version.
• Netflix: strive to be version-less from client perspective
• Netflix: try to be version-less even if not wholly possible
• Adding new data types or new URIs don’t require new version
• Structural changes to large unknown developers better to incomplete than inaccurate. Hold onto ideas before pushing it out.
• Netflix assumes 7-10 year life of a TV
• Netflix can generally move partners from version to version as needed
NETFLIX APPROACH
51. 52CONFIDENTIAL
52
API VERSIONING
It seems that there are a number of people recommending using Content-Negotiation (the HTTP
“Accept:” header) for API versioning.
However, none of the big public REST APIs I have looked at seem to be using this approach. They almost
exclusively put the API version number in the URI.
52. 53CONFIDENTIAL
53
API VERSIONING
Twitter URI
Atlassian URI
Google Search URI
Github API URI/Media Type in v3
Intention is to remove versioning in favour of
hypermedia – current
application/vnd.github.v3
Azure Custom Header x-ms-version: 2011-08-18
Facebook URI/optional versioning graph.facebook.com/v1.0/me
Bing Maps URI
Netflix URI parameter
http://api.netflix.com/catalog/titles/series/
70023522?v=1.5
53. 54CONFIDENTIAL
54
API VERSIONING
Google data API (youtube/spreadsheets/others)
URI parameter or custom
header “GData-Version: X.0” or “v=X.0”
Flickr No versioning?
Digg URI http://services.digg.com/2.0/comment.bury
Delicious URI https://api.del.icio.us/v1/posts/update
Last FM URI http://ws.audioscrobbler.com/2.0/
LinkedIn URI
http://api.linkedin.com/v1/people/~/connec
tions
Foursquare URI
https://api.foursquare.com/v2/venues/40a55
d80f964a52020f31ee3?oauth_token=XXX&v=YY
YYMMDD
54. 55CONFIDENTIAL
55
API VERSIONING
paypal parameter &VERSION=XX.0
Twitpic URI http://api.twitpic.com/2/upload.format
Etsy URI http://openapi.etsy.com/v2
Tropo URI https://api.tropo.com/1.0/sessions
Tumblr URI api.tumblr.com/v2/user/
openstreetmap URI and response body http://server/api/0.6/changeset/create
Ebay URI (I think)
http://open.api.ebay.com/shopping?version=
713
55. 56CONFIDENTIAL
56
API VERSIONING
Wikipedia no versioning I think?
Bitly URI https://api-ssl.bitly.com/v3/shorten
Disqus URI
https://disqus.com/api/3.0/posts/remove.js
on
Yammer URI /api/v1
Drop Box URI
https://api.dropbox.com/1/oauth/request_to
ken
Amazon Simple Queue Service (Soap) URI Parameter and WSDL URI &Version=2011-10-01
MONOLITH
Complex code base
Hard to maintain modularity – cyclic dependencies, code duplication, broken layers, broken SOLID principle
Single responsibility principle
Open/closed principle
Liskov substitution principle
Interface segregation principle
Dependency inversion principle
Change circles tightly coupled – any team which a delivered functionality faster than others tied by other teams due to common release circle, need to wait stable code, common regression, etc
Inefficient scaling
Scaring for newcomers
Hard to scale development team – hard to move people between teams – hard to deep dive into new code base
Long term commitment for a chose technology – can’t try to use any new technology just because it’s too risky
MICROSERVICE
Simple code base – but challenging in cross communication setup
Modularity with exact borders – no cross dependencies between different code base, just over defined API
Change circles decoupled – you can release when your team is ready and you can release faster
Efficient scaling – in case we are heavy in writing – we can separate this functionality to separate service and scale it independently
Newcomers adopting faster
Per service team responsibility
No technology lock – easy to experiment without applying any risk to all functionality
Subset of SAOMartin Fowler, chief scientist at ThoughtWorkshttps://youtu.be/wgdBVIX9ifA?t=854
Style of SAOZhamak Dehghani, principal consultant at ThoughtWorkshttps://youtu.be/1aaw7iYS_VM?t=173
Configuration – if it does change then it is configuration, if it does not change then it is code.
Configuration – if it does change then it is configuration, if it does not change then it is code.
503 Service Unavailable
504 Gateway TimeoutconnectionTimeoutThe number of milliseconds this Connector will wait, after accepting a connection, for the request URI line to be presented. Use a value of -1 to indicate no (i.e. infinite) timeout. The default value is 60000 (i.e. 60 seconds)
maxThreads(int) The max number of active threads in this pool, default is 200