Databases love nutanix
1 like2,181 views
The document outlines principles and best practices for optimizing database performance on Nutanix platforms, emphasizing storage strategies, virtual machine configurations, and management techniques for SQL Server and Oracle databases. Key recommendations include using SSD for high performance, separating storage for different database components, and leveraging virtualization effectively to reduce licensing costs. It also discusses scaling options and highlights resources for further learning on database performance optimization.
Databases love nutanix
- 2. 22 Principles of Database Servers Databases are an extension of storage Performance is more than just the underlying storage devices Size for performance before capacity “The Bitterness of Poor Performance Lasts Long After the Sweetness of a Cheap Price is Forgotten”
- 3. 33 “The best strategy to win an F1 GP is to have the fastest car – David Coulthard” Principles of Database Servers Virtualize, but without compromise Keep it standardized and simple “Flash! Ah-ahh! Savior of the Universe!... Flash! Ah-ahh! King of the Impossible!”
- 4. 44 Platform Selection Considerations • SSD for performance, HDD for cold storage • Utilize storage only nodes to reduce DB processor licensing costs • Large databases love lots of compute, memory and SSD • NX8150, Dell XC730xd-24, Lenovo HX7500 (and all flash platforms) Data Reduction Techniques - Your Mileage Will Vary • Use inline compression • Can be used in combination with DBs • SQL Compression and Oracle Advanced Compression Best Practices for Best Performance
- 5. 55 General Performance Guidelines • Single storage container* • Align to NUMA boundaries • Start with 2:1 vCPU, 1:1 memory • Huge pages / Large Pages • Separate IO across vSCSI adapters (except for AHV) • Use the latest hypervisor version • Use the latest Acropolis (AOS) version * To drive maximum single node performance on all flash systems requires two containers on VMware ESXi. Best Practices for Best Performance: General/VM
- 6. 66 • Max memory should be less than total memory • Set the right max degree of parallelism (MAXDOP) • Separate drives for the following • user data files, • user log files, • Temp DB Data, and • Temp DB Logs (Automatic in SQL 2016) • Security policy: Lock pages in memory, perform volume maintenance operations • Balance size of SQL VM and number of DB per VM • Use trace flags with care Best Practices for Best Performance: SQL Server
- 7. 77 ASM for storage management (1MB ASM AU) • At Least 2 vDisks per ASM Disk Group • Separate ASM Disk Groups/mount points for data files, redo logs, archive logs • Configure ASM Disk Groups with External Redundancy (Except OCR) Set FileSystemIO Options = SetAll Parallel threads per CPU = 1 Disable Shadow Clones with Oracle RAC Default Max Linux IO Size – Change to 1024KB Best Practices for Best Performance: Oracle
- 8. 88 Scale Up vs Scale Out Conundrum 0 10000 20000 30000 40000 50000 60000 70000 80000 90000 1 VM per Host 2 VM per Host 3 VM per Host 4 VM per Host OrdersPerMinute(OPM) 2 vCPU per VM SQL Performance Scaleability 0 10000 20000 30000 40000 50000 60000 70000 80000 90000 2 vCPU 4 vCPU 8 vCPU OrdersPerMinutes(OPM) One VM Per Host Scale Up Performance Benchmarks using DVD Store 2.1 on NX3450
- 9. 99 TPC-C Like Test Using Benchmark Factory for Databases Scale Up vs Scale Out Conundrum 1879 1945 1921 15 15 15 SQL01 SQL02 SQL03 3-Node Test Trans Per Sec Trans Time (ms) 1872 1969 1870 1953 1951 1938 15 15 15 15 15 15 SQL01 SQL02 SQL03 SQL04 SQL05 SQL06 6-Node Run Trans Per Sec Trans Time (ms)
- 10. 1010 Swingbench on Single Oracle DB VM Scale Up vs Scale Out Conundrum
- 11. 1111 DB Supported when deployed on supported Operating System Hypervisor Specific Details: • Oracle on VMware vSphere Support – MOS Note 249212.1 • Oracle on Microsoft Hyper-V Support – MOS Note 1563794.1 • Oracle on AHV – NTNX KB 2682 • SQL Server AlwaysOn Availability Groups – MS KB 956893, NTNX KB 2177 • SQL Server Failover Cluster – NTNX KB 2660 Database Support on Nutanix (Note: MOS = My Oracle Support)
- 12. 1212 • Isolate databases licensed per processor to a reduced number of hosts • Optimize Cores Per Socket and Clock Speed • All Flash CTO Options on All Platforms • Any Node can be a storage only node • It’s easy to be audit ready! Reduce Database Licensing Costs Storage only nodes
- 13. 1313 Efficient Dev / Test Environments VM clones Give developers their own high performance environment Master / production VM Quickly spin up local or remote VM clones for QA testing VM clones or replicas Benefit from a VM-centric approach to development and testing.
- 14. 1414 What a Difference a Year Makes 4.1.2 4.6 CompletionTime Nutanix OS Version Clone 12 x 500 GB SQL VM's With VMware vSphere on Nutanix NX3450 Clone 12 x 500 GB SQL VM's With VMware vSphere on Nutanix NX3450 70% Faster Clone Time
- 15. 1515 SQL Always On Availability Groups/FCI Oracle DataGuard, Fast Start Failover/MAA Technologies at your disposal Application Storage Hypervisor Async and Sync Replication High Availability / Live Migration
- 16. 16 Like What You Just Heard… There’s More! READ • Longwhiteclouds.com and Next.Nutanix.com: Lots of Performance, Oracle and SQL Content • SQL Server and Oracle Best Practice Guide • Virtualizing SQL Server with VMware: Doing IT Right (VMware Press 2014)
Editor's Notes
- #18: What big data apps are you running on Nutanix? Is this a new service for you or did you migrate it from another platform? ---If you migrated it, what did you migrate it from? ---What were the drivers for change? ---If it was new, what were the drivers for you to adopt big data? What other options did you consider for your app? Why did you choose Nutanix to run your app? ---Why did you virtualize rather then going bare metal? ---Did you consider cloud offerings? What were the deciding factors? What are the benefits that you've seen running your app on Nutanix?