Oracle database storage characteristics with HPE GreenLake for Block Storage

[StorageExperts]

Take a technical deep dive focused on HPE GreenLake for Block Storage using HPE Alletra MP. HPE's disaggregated block storage solution that can be integrated with Oracle database. Actual database IO characterization tests for OLTP and OLAP storage show over doubled improvements to current industry-leading arrays while positioning Oracle workloads into the hybrid cloud in the new digital age.

–By Todd Price, Storage Solutions Technical Marketing, HPE

We’re seeing a major shift from legacy implementations of databases to the new digital age of data science, data engineering, and AI. Many forms of databases are being used to achieve data models and data lakes that provide for these future technologies. Even with some of the newest database technologies like document stores, NoSQL, graph, and vector databases, to name a few, the traditional Oracle OLTP and OLAP databases are of extreme importance and integrate with newer database technologies.

For customers still running Oracle instances, a vast amount of data that will be needed for future data science and AI are currently stored in Oracle OLTP and OLAP databases.

The structured database

Customers are increasingly having to integrate newer database technologies like graph, time series and vector databasesHPE GreenLake for Block Storage powered by HPE Alletra MP with traditional rdbms and data analytics tables and views, You need to be able to address multiple IO patterns on common storage. HPE GreenLake for Block Storage is designed to tackle the issue of multiple IO patterns with ease. You’re also looking for a hybrid cloud operational model with the performance, low latency, and uptime that only on-prem solutions can provide. A new common hardware platform, HPE Alletra MP on GreenLake for Block Storage meets these solution needs.

In this blog, I take you on a technical deep dive into the capabilities of the performance of Oracle database in a medium-capacity mission-critical database using HPE GreenLake for Block Storage powered by HPE Alletra MP.

For a detailed technical reference, see the HPE GreenLake for Block Storage architecture technical paper.

Here’s a test with two Oracle storage IO workloads on HPE Alletra MP.

There are two workload types HPE tested with Oracle database 19c (19.18) on Linux. Two separate databases were created, one for OLTP of 3TB in size and one for OLAP 5TB in size. The purpose of this test was to characterize the IO on Alletra MP storage and compare it to an industry-leading array.

The databases used Oracle Automatic Storage Manager for the storage groups that hold all the database schemas to characterize the new Alletra MP for workload IO where traditional OLTP and data analytics meet new throughput centric workloads using newer database technologies. These configurations were also tested on an industry leading storage array.

OLTP: The first is a 3TB OLTP database built using the industry available tool called SLOB (Silly Little Oracle Benchmark). The data reduced workload IO block size was 8k block with a series of Oracle SELECT/UPDATE (80/20) commands. This test was run to measure IOPS at sub millisecond latencies. The results of the SLOB tests on the Alletra MP were compared with the same tests performed on the industry leading array. Here you’ll see the results for a 3TB database with 250 schemas. The latencies remained below IO 0.6 milliseconds.

OLAP: The second ia a 5TB OLAP database created using a popular tool called HammerDB and a workload called TPROC-H. This was also tested using data reduction. This test used an “out of the box” setup with all of the same tuning and server parameters. The only variable was the model of storage.  TPROC-H query completion times on the Alletra MP were compared to the other industry array. While observing query completion times, the data transfer rates during large table scans could also be monitored to demonstrated data transfer performance. The results of the OLAP tests on Alletra MP were also quite impressive when compared to another industry leading array. Here you’ll see the improvements in data transfer rates.

Looking deeper into the TPROC-H query completion times on HPE Alletra MP, the average mean percentage improvement was 59.5% and out of all 21 queries Query 2 through query 22, all queries showed improvements for all query types on the 5TB schema.

 

Oracle databases (like many other databases) require the ability to scale – and HPE GreenLake for Block Storage powered by HPE Alletra MP allows it to be done effortlessly. The disaggregated architecture that allows for scalability required by databases and performance that grows with scale. By using the HPE Data Services Cloud Console to provision from a cloud mindset, database storage provisioning can be simplified and implemented the same whether it’s an Oracle OLTP, OLAP, NoSQL, graph, or many other databases, ultimately enabling customers to achieve their business requirements.

HPE has optimized HPE Alletra MP addressing two areas for workloads: Traditional OLTP IO response times and the co-existence with newer heavier data analytics workloads with higher data transfer rates. This safely enables you to use HPE Alletra MP to run new workloads along with the traditional workloads on the same consolidated and disaggregated array technology providing ease of use through a cloud operational model and flexible purchasing model.

Data capacity and performance can be monitored and analyzed the same way regardless of workload type.

Using HPE GreenLake for Block Storage for databases and many other block device workloads provides a single cloud management environment, with the ability to provision and pay as you go making it cost effective.

Count out the seven top reason to run your Oracle database on HPE storage.

1. Eliminated complexity
2. 100% guaranteed uptime
3. High performant IO
4. Built in AI
5. Hybrid cloud ready
6. Full stack compute readiness
7. East integration with more services

To drill down, check out the blog and video: 7 top reasons to run Oracle on HPE Storage

Learn even more with these demos, videos, and other resources.

The ease of use on the hybrid cloud-enabled HPE Data Service Cloud Console (DSCC) makes provisioning easy for anyone administrating the storage. Watch this HPE GreenLake for Block Storage technical demo for a walk-through of provisioning on the DSCC:

Here are more videos to watch:

• HPE GreenLake for Blog Storage – three-minute demo
• HPE GreenLake for Block Storage – technical demo
• HPE GreenLake for Block Storage for mission-critical databases

Also check out:

• SLOB Tool (Silly Little Oracle Benchmark)
• HammerDB – TPROC-H

---

Meet Storage Experts blogger Todd Price, Storage Solutions Technical Marketing, HPE.

Todd is a Senior System Engineer VI focused primarily on Oracle database workloads as well as other relational databases like MySQL and Postgres. Todd’s career at HPE/HP spans 29+ years, from dedicated enterprise customer support to R&D engineering in both server and storage technologies. He also has 10 years of MRI medical imaging engineering and support experience.

Storage Experts
Hewlett Packard Enterprise

twitter.com/HPE_Storage
linkedin.com/showcase/hpestorage/
hpe.com/storage