3. 3
Agenda
A Quick Big Data Primer
Healthcare and Big Data
Compliance and Auditing
SQL Compliance Project
Compliance and Auditing with Big Data and BI
Big Data: Unstructured Volumes of Data
Analytics: PowerPivot, Power View
4. 4
What is Big Data?
Volume
Exceeds physical limits of vertical scalability
Velocity
Decision window small compared to data
change rate
Variety
Many different formats makes integration
expensive
Variability
Many options or variable interpretations
confound analysis
5. 5
10x
increase every
five years
85%from
new data types
Data
explosion
Volume
Velocity
Variety
Hadoop
Cloud
By 2015, organizations that
build a modern information
management system will
outperform their peers
financially by 20 percent.
– Gartner, Mark Beyer
“Information Management in the
21st Century”
6.
7. 7
Big Data Business Value
140,000-190,000
1.5 million
$300 billion
15 out of 17
€250 billion 50-60%
12. 12
Healthcare and IT
Often the laggard in technology
Yet application of IT to healthcare can radically change what we can do
Genomic Sequencing
Proteomic sequencing
Incidence Prediction
13. 13
Healthcare Big Data Example Scenarios
Clinical Trial Deviations
Originally Viagra was developed to lower blood pressure and treat Angina
Now its used to help newborn pulmonary hypertension and altitude sickness
Incidence Prediction
Missed 4 or more visits, twice as likely to have an asthmatic incident
Particular Cardiac monitor sine wave points to highly likelihood of heart attack
Campaigns
Social media and advertising campaigns to understand user behavior and sentiment
Patient Satisfaction
Social media and advertising campaigns to understand user behavior and sentiment
14. 14
BIDMC Auditing Scenario
Auditing is critical component HIPAA in ensuring patient privacy
1 Billion rows+ of audit data
146 mission critical clinical applications
Comprehensive audits yield 300-500k transactions/day
HIPAA requires audit system with 20 years of data
Auditing Project
Available to community as part of Compliance SDK
Updating for SQL Server 2012, HDInsight, Power View, and MobileBI*
Creating an enterprise tool for consolidated storage, reporting and alerting of all application audit
data - that's cool!
John Halamka’s Cool Technology of the Week
(Wellsphere Top Health Blogger, Health Impact Award)
16. 16
Auditing Sensitive Information
16
Querying Audit Information
Use PowerPivot / Power View / Analysis Services to Query the data.
Security InformationPolicy Information
Process Audit Information
Use SSIS to process SQL2008 All-Actions Audit Information and other CG application
audit log data; potentially can use Management Performance DW framework.
Caregroup Environment
File Server
SQL Audit
Connect/Logic
SSIS
CG Application Data
Intersystems
Cache
SQL2005
Oracle
SQL2008 All-Actions Audit Data
SQL 2008 / 2012 R2
SSRS 2008 /
Power View
Policy Analysis
Policy Reports
Policy Best
Practices
Security Analysis
Security Reports
Compliance
Reports
Feedback Action Loop
Update systems to keep them
compliant and secure
23. 23
Hadoop / Auditing: File sizes
Currently testing gz vs. raw
E.g. 12MB raw text file vs. 633Kb gz file (~20x compression)
20x smaller size, ~same query time
Approx same map / reduce task utilization
File Size is 250MB-1GB
SSIS package takes care of the size
Future testing: avro, protobuf
23
Query Duration (s)
select count(*) from sql_audit_asv_raw 56.066
select count(*) from sql_audit_asv_gz 58.994
24. 24
Hadoop / Auditing: Formats
For ease of processing, replace carriage returns within embedded SQL
statements, e.g.
select col1, col2
from tableA
to
select col1, col2 from tableA
This allows you to create a Hive table using CR as row delimiter (i.e.
does not have things like SQL quoted identifiers)
24
Centralizing Logs
Allows you to have one system process all audit logs from your servers
Easier manageability
Set files to 250MB in size (less files, but not too large to process)
Optimized for Hadoop General Rule of Thumb: 250MB-1GB file sizes
Can also centralize processing
… and centralize reporting
Compliance SDK contains the full project
Organized by Server, Database, DDL, and DML actions