Rule-based dispatching of events to a serverless services armada

© 2017 Global Healthcare Exchange, LLC. All rights reserved. GHX proprietary information. Do not copy or distribute.
Rule-based dispatching of events
to a serverless services armada
Daniel Buchholz
Dev Manager / Cloud Architect - Düsseldorf
@danbdo

GHX is all about supplier chain processes in healthcare
Hospitals
Groups
Suppliers
Mfr‘s
Distributors

What‘s was the problem, anyway?
Public Domain, https://commons.wikimedia.org/w/index.php?curid=868126
Individual catalog systems in all
(current) target markets
Each with different schema
All supposed to talk to each other
All doing a reasonably similar job
Operational efficiencies
Mappings / Conversions
Expandability
Catalog management

Legacy systems
this
that
More
stuff
Squeezed
in...
WHOA!
add
modify
delete
Legacy catalog processing

The Path Forward
consume
process
report
CSV Source type A
Source type BGS1
XML
JSON
Source type C
Suppliers
Groups
Hospitals
InboundConsumers/Processing
...
...
...
...
BI BI
...
...
...
...
...

The Path Forward
consume
process
report
CSV Source type A
Source type BGS1
XML
JSON
Source type C
Suppliers
Groups
Hospitals
InboundConsumers/Processing
...
...
...
...
BI BI
...
...
...
...
...
ETL
Queueing / Mapping / Routing
Single Source Of Truth
events

Consumers/Processing
Deep dive
consume
process
report
CSV Source type A
Source type BGS1
XML
JSON
Source type C
Suppliers
Groups
Hospitals
Inbound
...
...
...
...
BI BI
...
...
...
...
...
ETL
Queueing / Mapping / Routing
Single Source Of Truth
events

„Each with different schema“
Catalog A
Att A: abc
Att B: def
Att C: ghi
Att D: jkl
Catalog B
Att P: 4711
Att B: blah
Att D: stuff
Catalog C
Att W: key1
Att X: key2
Att Y: foo
Att Z: bar
>> Schema-less data store

„Each with different schema“
Catalog A
Att A: abc
Att B: def
Att C: ghi
Att D: jkl
Pkey: hash(A)
Catalog B
Att P: 4711
Att B: blah
Att D: stuff
Pkey: hash(P)
Catalog C
Att W: key1
Att X: key2
Att Y: foo
Att Z: bar
Pkey:hash(W/X)
>> Infer internal Pkey values
>> DynamoDB
>> Schema-less data store

ETL
files de-dup delta
Item hash
Item
events
map
Item key
Schema
very specific very generic

Dispatching item events
catalog data

Don‘t want to deploy full stack for each change
Dispatching item events
catalog data
Dispatcher
Function
„All doing a reasonably similar job“
>> Migrate tasks from legacy
>> Keep it open to changes
>> Static workflows wouldn‘t work
>> need rules engine

Dispatcher Config
catalog data
Dispatcher
Function Dispatcher
Config Store
2 - Description
1 - Name
3 - Conditions
4 – Action(s)
success / failure

Dispatcher Services
catalog data
Dispatcher
Function Dispatcher
Config Store
...Svc A Svc DSvc B Svc C
Search..
Engines
A Dispatcher – not a Broker >> fire & forgetHighly de-coupled
>> limit blast radius

Dispatcher Endpoints
catalog data
Dispatcher
Function Dispatcher
Config Store
Search..
Engines

Audits / Traces
v4 v3/v4 v2/v3 v1/v2 v0/v1
OPS teams need visibility >> full audit / full trace
DEV probably too
live audit
Causes & Effects
Svc A trace
Svc B trace
Consumer X
Consumer Y
Svc B trace
Svc C trace
Consumer Y
Svc A trace
BI A
BI B
Consumer X
Svc A trace
Svc B trace

Core Services
catalog data
Dispatcher
Function
Search..
Engines
Audit log
Trace log
Core Services

Versioning/Deployment options
80/20
Svc A Svc D Svc E
Svc E2Svc A2
In your face
Svc D2
Traffic shifting Parallel / Audit

So...
This is our architecture
Happy to hear feedback

Rule-based dispatching of events to a serverless services armada

More Related Content

Similar to Rule-based dispatching of events to a serverless services armada (20)

Recently uploaded (20)

Rule-based dispatching of events to a serverless services armada

Editor's Notes