PoC of IBM Informix Warehouse Accelerator and Storage Optimization Feature

Alexandre Marini

Senior Informix DBA – Orizon Brazil

[email protected]

PoC of IWA with Informix storage optimization, and its great value to

Health Insurance systems

1

Abstract This presentation will cover a PoC of Informix Warehouse Accelerator, together with implementation of storage optimization features in the Informix OLAP engine, based on health insurance systems, made to demonstrate the product capabilities to reduce enterprise costs, ease administration, and lower the report generation periods, compared to our market

competitors. The idea of this PoC was to provide our company (Orizon Brazil) a better product, with lower costs and higher speed, to increase it´s portfolio of products, with unmatched IT and information values to offer our clients.

2

Alexandre Marini - Personal Profile

– Started working with IBM Informix 1998 - Brazilian state government (4gl / DBA)

– First IBM Informix On Campus in Brazil in 2011

– Worked with MC Software in 2011

– Worked in Cleartech in 2011/2012

– Started working in Orizon in October, 2012

– My First IIUG presentation (please be patient!)

3

Agenda

• About Orizon

• A little about IWA

• A little about SOF

• Business and company needs

• Implementation of this PoC

• Results

• Conclusion

• References

4

About Orizon

5

MORE THAN 10 YEARS of history in the health care market

LEADER SHAREHOLDERS in their segments

Bradesco Seguros Group

Cielo

CASSI

1 OUT OF 3 LIFES in private health

care are touched by our systems

OVER 140 MILLION

of health transactions per year, each one completed in

less than 0.5 second

GREATEST MARKET companies are our clients

More than a system, we offer a SERVICES PLATFORM fully attached to our

customers needs

18 MILLION lives

130 THOUSAND

of connected providers

8.5 THOUSAND

Drugstores

About Orizon

6

Providers

100% electronic medical bills

Electronic Receipt validation

Clients Platform

Electronic Authorization From low to high complexity

“Autorize” Platform AUTORIZE is an electronic platform for capture and validation of requests, electronic receipts and processing of medical/ hospital care, with application of SMART ELIGIBILITY rules

A little about IWA • Designed with in-memory acceleration for

Informix DW databases, mixed or not with OLTP data

• Introduced in 11.70.xC2, March 2011 – one node only

• Last release 12.10.xC2, October 2013 – works on multiple nodes, loading from single or multiple clusters, TimeSeries acceleration, external tables acceleration

• Hardware prerequisites: Linux 64 bits Intel box with SSE3 (recommendation: separate box from Informix engine)

7

A little about SOF • Dictionary based for Informix databases • Introduced in 11.50.xC4, May 2009 – basic data

types only, table data only • Last release 12.10.xC2, October 2013 –

compression of B-tree indexes, simple large objects, automatic data compression (xC1 features)

• License as a separated pack, available for Informix Enterprise Edition

• Average of storage savings: around 70% • Rather different from other engine vendors

8

Business and company needs • Integrate all Orizon DWs, from different vendors • Improve stability and speed, bringing economy and new

capabilities for company reports generation • Informix and its best: stability, confidence, and low TCO

costs (cheaper at least 31% than SQL Server – published September 2010)

• IWA proposed, migration of all DWs to Informix 12.10, plus storage optimization to reduce storage usage and costs, so a PoC was needed to prove Orizon needs

• Purpose of this PoC is not the best performance: lab for demonstration purposes, for comparison (IWA) and storage (SOF)

• Informix 12.10: index compression, Smart object compression, NoSQL features, if needed

9

Implementation specs • Hardware is HP Intel Blade Xeon (2 sockets),

product installed into a VM with 4 cores and 16GB of memory

• Informix 12.10.FC2TL on a two node VM cluster (prim + SDS), running on a RHES 6.4 with GFS clustering

• IWA on primary Informix node, NUM_NODES=4, WORKER_SHM=9GB and COORDINATOR_SHM=1GB

• Raw devices in a HP P4500 storage, with RAID level 0

10

Implementation – numbers

• Historical health care OLAP database was created, with one fact table and 8 dimensional tables, more than 6 months of data are loaded

• Fact table with 14.9 million rows, populated from production OLTP data for real results output, demanding 496.83MB of storage

• OLAP database size is 1.01GB

• Load data mart timing (stop + load process): 1m5.815s

11

Query testings (1/6)

12

• Queries tested: simple ones, with aggregation, ranking select a13.year year, a13.month month, a12.razao_social razao_social, sum(a11.total_proce), sum(a11.total_trans) from fato_transacao a11, dim_prestador a12, dim_data a13 where a11.id_prestador = a12.id_prestador and a11.id_data = a13.id_data and (a13.year in (2011,2012,2013) and a13.month in (3,5,6,9, 10,12)) group by a13.year, a13.month, a12.razao_social • Rows retrieved: 78752

SIMPLE ONE, LONG RESULT SET

Query testings (2/6) select id_ems, a13.year year, a13.month month, sum(a11.total_proce), sum(a11.total_trans), RANK() over (order by a13.year, a13.month) as rank from fato_transacao a11, dim_prestador a12, dim_data a13 where a11.id_prestador = a12.id_prestador and a11.id_data = a13.id_data group by id_ems, year, month

13

RANKING

Query testings (3/6) select a13.year year, a13.month month, a12.desc_situacao desc_situacao, sum(a11.total_proce) from fato_transacao a11, dim_situacao a12, dim_data a13 where a11.id_situacao = a12.id_situacao and a11.id_data a13.id_data and (a13.year in (2013) and a13.month in (6, 7, 8, 9, 10, 11,12)) group by a13.year, a13.month, a12.desc_situacao

14

SIMPLE ONE HIGHER

PROJECTION

Query testings (4/6) select a11.id_ems ems, a13.year year, a13.month month,

sum(a11.total_proce) SUM_TOTAL_PROCE,

sum(a11.total_trans) SUM_TOTAL_TRANS,

RATIO_TO_REPORT(a11.total_proce) OVER() *100 AS RATIO_TOTAL_PROCE,

RATIO_TO_REPORT(a11.total_trans) OVER() *100 AS RATIO_TOTAL_TRANS from fato_transacao a11, dim_data a13 where a11.id_data = a13.id_data and (a13.year in (2011,2012,2013) and a13.month in (1,2,3,4,5,6)) group by a11.id_ems, a13.year, a13.month, a11.total_proce, a11.total_trans order by 1,2

15

RATIO

Query testings (5/6) select id_ems, a13.year year, a13.month month,

sum(a11.total_proce),

sum(a11.total_trans),

PERCENT_RANK() over (order by id_ems) as perc_rank

from fato_transacao a11,

dim_prestador a12,

dim_data a13

where a11.id_prestador = a12.id_prestador and

a11.id_data = a13.id_data

group by id_ems, year, month

16

RANKING

Query testings (6/6) select a13.year year, a13.month month, a12.razao_social razao_social, sum(a11.total_proce) , sum(a11.total_trans) from fato_transacao a11, dim_prestador a12, dim_data a13 where a11.id_prestador = a12.id_prestador and a11.id_data = a13.id_data and (a13.year in (2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023, 2024, 2025, 2026, 2027, 2028, 2029, 2030, 2031, 2032, 2033, 2034, 2035) and a13.month in (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12)) group by a13.year, a13.month, a12.razao_social • Rows retrieved: 141792

17

SIMPLE ONE FULL OLAP

PROJECTION

Screenshots

18

MicroStrategy web results

Results

• Timing comparison

19

Informix Informix + IWA Time reduction Times faster

query 1 00:04:37.86 00:00:03.42 00:04:34.44 81.25

query 2 00:03:11.63 00:00:01.47 00:03:10.16 130.36

query 3 00:07:49.37 00:00:01.16 00:07:48.00 404.63

query 4 00:01:26.11 00:00:00.93 00:01:25.18 92.59

query 5 00:03:11.28 00:00:01.48 00:03:09.80 129.24

query 6 00:04:00.41 00:00:04.01 00:03:56.40 59.95

Average

enhancements: 149.67

Obs: Time format HH:MM:SS.00

Results

• Storage comparison

20

Informix 12 Engine1 Engine2

Row size 44 47 52

Data Size (MB) 465.43 680.74 753.16

Storage costs - +31.63% +38.20%

Informix 12 storage (SOF) compared to other market databases

Conclusion • IWA - higher value to our information services

– quicker report generations - increase our product portfolio to our clients, a new perspective.

– Reports will run in seconds instead of hours

– Ease administration, on indexes/table reorgs, installation was very simple

• Informix approx. 2 years savings in storage space (OLAP size 3TB, in data, HP P4500 storage) : – US$ 117K+ (compared to a market leader product

Engine2)

– US$ 97K+ (compared to another Engine1)

21

Conclusion

• In memory technology considerations – Source data ammount does not impact result timings

• IWA licensing features

– Two brands of distribution packages

• Advanced Workgroup Edition: only PVU, 16 cores and 48GB of memory, neither include SOF nor HA/ER

• Advanced Enterprise Edition: full features

• Combination of IWA + SOF is absolutely a “state of the art” for health insurance systems.

22

Conclusion

We need no Iron Man to be our company heroes…..

23

References • Query acceleration for Business using Informix

Warehouse Accelerator (IBM RedBook): http://www.redbooks.ibm.com/Redbooks.nsf/RedbookAbstracts/SG248150.html

• Informix Warehouse Accelerator (youtube): http://youtu.be/C-dvl_EptLY

• IBM Informix 12 Compression: Helps Optimize Storage (youtube):

http://buff.ly/1gSN5di

24

References • Informix Warehouse Accelerator (IBM

DeveloperWorks blog): https://www.ibm.com/developerworks/community/blogs/2fa81a5c-cb30-4873-b775-1370151e3614/entry/introducing_informix_warehouse_accelerator9?lang=en

• Keshava Murthy blog (DeveloperWorks): https://www.ibm.com/developerworks/community/blogs/Keshav/?maxresults=15&lang=en_us

• Fred Ho blog (DeveloperWorks): https://www.ibm.com/developerworks/community/blogs/fredho66/?maxresults=15&lang=en_us

25

Alexandre Marini Senior Informix DBA

Orizon Brazil: www.orizon.com.br [email protected]

Questions?

26