Case Study #2 Power School
- Hardware Specs – 16 Cores, JVM 10 GB, 63 GB RAM
- Operating System – Red Hat 4.8.5-36 (Running on MS Hyperv VM)
- Java Version – 1.8 (OpenJDK 64 Bit)
- Solr Version – 6.6.5 (Single Instance without Zookeeper)
- Core Name – UnifyNumDocs – 13,092,202
- Size of Index – 25.7 GB
- 75thPcRequestTime – 165 ms
- 95thPcRequestTime – 672 msavgTimePerRequest – 142 ms
- avgRequestsPerSecond – 2
- JVM Setting – UseParNewGC & UseConcMarkSweepGC
Before: 5.2 Seconds Query Response Time
Solution
- Recommendation on JVM Setting (Add in Solr.in.sh file)
- Recommendation on Solr Cache to be modified in SolrConfig.xml
- Add docValues=true for these fields in schema.xml
- As per the recommendation G1GC should be used with JDK 1.8 and above to get better memory usage. If possible MaxHeapSize should be set as 30 GB which is little bit higher than index size of 25.7 GB. Although there is limit of setting MaxHeapSize with OpenJDK, so we recommend to use Oracle JDK which have higher limit as compare to OpenJDK.
- Given the size of index and number of documents it is recommended to use latest version of Solr and instead of single instance it should be Solr Cloud. As per the current index size and number of documents, it is recommended to create minimum 2 shards on Solr Cloud with minimum 2 instance of Solr and 3 instances of Zookeeper (should be on separate machine).
After: 3.4 Seconds 35% Reduction
Benefits Of Working With Nextbrick
- Technical Expertise
- Domain Expertise
- Quality , team of Industry leading best practices experience Data Scientists and Solr Consultants
- Speed: Fast onboarding , designing, and development , and solution delivery services
- Cost: Low cost, flexible , negotiable as per budget and requirement
- Quality: Ensure client goals are met as per acceptance and success criteria with most optimal solution
Thank You
For more information about us, Contact us at:
Let's break ice
Email Us