All use of Scientific Computing and Data resources in publications and presentations must acknowledge Mount Sinai’s CTSA award. Click here

Benchmarking

We made performance comparisons for the SAP HANA system at AIR·MS comparing:

  • Execution time
  • Search time
  • Data compression ratio

Execution Time Performance Benchmark

Here you can find a performance comparison between AIR·MS using SAP HANA and Microsoft SQL Server.
The cohort record count for both was 985,092 records.

  AIR·MS on SAP HANA  Microsoft SQL Server
Database Systems SAP HANA 2.0 SPS05 on Azure Standard M64s (64 vCPUs with 1024 GiB memory)  Microsoft SQL Server Enterprise 2019 Azure Standard L32s v2 (32 vCPUs with 256 GiB memory) 
First Run 174 ms 268 µs 5 minutes 29 seconds 
Second Run 173 ms 384 µs  5 minutes 36 seconds 
Third Run 172 ms 193 µs  5 minutes 36 seconds 
Average execution time  173 ms 281 µs  5 minutes 33 seconds 

Example SQL code used to build patient cohort in AIR·MS

select count(*) from ( 
select 
    CDMPHI.CONDITION_OCCURRENCE.PERSON_ID, 
    CDMPHI.PERSON.GENDER_SOURCE_VALUE as gender, 
    CDMPHI.PERSON.RACE_SOURCE_VALUE as race, 
    CDMPHI.CONCEPT.CONCEPT_NAME as diagnosis, 
    CDMPHI.CONCEPT.CONCEPT_CODE as icd10, 
    CDMPHI.CONDITION_OCCURRENCE.CONDITION_START_DATE as dx_date 
from CDMPHI.CONDITION_OCCURRENCE 
join CDMPHI.CONCEPT_RELATIONSHIP on CDMPHI.CONCEPT_RELATIONSHIP.CONCEPT_ID_2 = CDMPHI.CONDITION_OCCURRENCE.CONDITION_CONCEPT_ID 
join CDMPHI.CONCEPT on CDMPHI.CONCEPT.CONCEPT_ID = CDMPHI.CONCEPT_RELATIONSHIP.CONCEPT_ID_1 
join CDMPHI.PERSON on CDMPHI.PERSON.PERSON_ID = CDMPHI.CONDITION_OCCURRENCE.PERSON_ID 
where CDMPHI.CONCEPT_RELATIONSHIP.RELATIONSHIP_ID = ‘Maps to’ 
and CDMPHI.CONCEPT.VOCABULARY_ID like ‘%ICD10%’ 
and CDMPHI.CONCEPT.CONCEPT_CODE like ‘C34%’) as count 

In-memory Search Benchmark

The following examples show execution time benchmarks for our SAP HANA system.

AIR.MS in-memory full-text search performance examples

Number of indexed notes: 86,893,767

Example 1: Show count for notes containing an EXACT search string ‘acute kidney disease‘, broken down by gender

 

0.31 second runtime!

Example 2: Show count for notes containing FUZZY search string ‘acute kidney disease‘, broken down by gender

 

0.39 second runtime!

Example 3:%