Premise
Oracle has been obsessed for decades with constantly pushing the performance envelope while reducing the total cost of ownership (TCO). These continuous performance gains and TCO reductions are a consistent trait across Oracle’s product lines and services. It just as true for Oracle databases, as it is for hardware platforms, applications, cloud infrastructure, and cloud services. The fundamental objectives are always best-in-class performance that expands considerably with each new major release and the lowest TCO per unit of performance.
Based on customer experiences and standard benchmarks, it’s working.
The rationale behind Oracle’s focus is to constantly increase customer value. Faster infrastructure—on-premises or in the cloud—accelerates databases. Faster databases accelerate application performance. Faster applications reduces response times to the user. Faster application response times converts into higher user productivity. Higher user productivity translates into faster-time-to-actionable-insights, faster-time-to-market, and faster-time-to-revenues. This was revealed in IBM’s ground breaking 1982 and 1997 research in the “Economic Value of Rapid Response Times”.
When application response times exceeded 3 seconds, productivity plummeted because the user lost focus and their mind wandered. When application response times dropped from 3 seconds to .3 seconds, productivity soared. Users were more than twice as productive. Enhanced productivity leads to faster-time-to-market and faster-time-to-revenues. That’s considerable customer value.
Throughput is crucial for analytics of all kinds and machine learning (ML) performance. Higher throughput shortens training time for models and time-to-actionable-insights. Faster training time equates into more timely ML predictions. Faster actionable insights and timelier ML predictions lead to faster time-to-market and faster time-to-revenues. The latter usually eclipses the cost of getting there.
Improving performance translates into much higher customer value. It’s really that simple and why Oracle is obsessed with constantly improving performance.
Best-in-class performance is highly prevalent in every product and service available from Oracle. It’s an integral part of the Oracle DNA. That leading edge performance is a central aspect of their databases, applications, platforms, Oracle Cloud Infrastructure (OCI), and SaaS applications. Oracle is constantly pushing the performance and price performance envelope in everything they do.
Research Key Findings
Oracle is the performance market leader in databases, database platforms, and database cloud services. Whether it is the Enterprise-leading Oracle Database—used by 19 of the top 20 Global 500 banks, MySQL open-source, Exadata platform, Autonomous Database services, or MySQL HeatWave, no one else is anywhere near in performance or TCO/performance.
An essential partner helping Oracle achieve that leadership is AMD with their EPYC CPUs.
How Oracle Has Gone About Achieving Bleeding-Edge Speed
Oracle realized early in their efforts that pushing the performance envelope requires more than just optimizing their database or application software. And more than simply running on faster hardware infrastructure. Neither produced the desired performance results. Nor did just combining the two together.
To get the best market-leading performance requires a deep level of integration optimization of the entire software stack and extensive synergistic integration optimization with the complete hardware infrastructure and microcode. In other words, it needs comprehensive cooperation and close collaboration between Oracle and the hardware infrastructure vendors all the way down to the component level to get the TCO/performance Oracle seeks. It’s source code-level co-engineering.
That is precisely what Oracle has been doing. One of those increasingly important cooperative hardware partners is AMD. AMD’s EPYC Zen 3 computer processor microarchitecture is a crucial performance enabler of multiple high performance Oracle products and services. Products such as Exadata Database Service on OCI, Autonomous Transaction Processing, Autonomous Data Warehouse, Autonomous JSON Database, Exadata Cloud@Customer, Dedicated Region, and the very rapidly growing MySQL HeatWave.
Why AMD EPYC 7003 Series Processors?
AMD provides the highest core count and the fastest x86 cores. Just as Oracle consistently pushes the performance envelope for its products and services, AMD is pushing the performance standard in CPUs.
AMD 3rd Gen EPYC™ Processors have optimized multi-processor performance and scalability through the latest advancements in die-to-die connectivity with the AMD Infinity Fabric™. This delivers an exceptionally high-speed chiplet interconnect. EPYC 3rd generation processors also provide exceptional memory bandwidth with 8 memory channels per socket delivering a peak 410 GB/s DRAM bandwidth. I/O bandwidth is also industry-leading with 8×16 PCIe Gen4 links, equating into 128 PCIe 4.0 lanes per socket and 160 per dual socket. Each bandwidth link can support up to 64GB/s bi-directionally for 512GB per socket.
In simple terms, the AMD EPYC 7003 series performance is indeed epic. It is compellingly faster than equivalent Intel CPUs based on industry-standard benchmarks. For general-purpose computing throughput utilizing the SPECrate® 2017 Integer benchmark, 2 EPYC 7763 sockets are 42% faster than 2 Intel Xeon 8380s. For hyperconverged infrastructure (HCI) using the VMmark® 3.11 vSAN™ 4-node benchmark, they’re 61% faster. This is the kind of CPU performance that Oracle needs to power its cloud databases.
The AMD EPYC 7003 series doesn’t just push the performance envelope; it also breaks the CPU scalability barrier. They are scalable up to 64 cores per socket – with 128 threads – and 128 cores per dual sockets for 256 threads. From an Oracle Database point of view, it’s a 256-way processor.
A very important requirement for Oracle has always been security, and the AMD EPYC 7003 series brings cutting-edge security built into the CPU silicon with AMD Infinity Guard. Infinity Guard specifically decreases potential attack surfaces as software boots, executes, and processes data. It comes with secure encrypted virtualization (SEV) architected to help safeguard VM privacy and integrity; secure nested paging (SEV-SNP), providing strong memory integrity protection capabilities; secure memory encryption (SME) to guard against main memory attacks; and AMD Shadow Stack™ hardware-enforced stack protection capabilities against malware attacks.
To put it succinctly, AMD EPYC 7003 series processors deliver the CPU performance that Oracle demands to deliver their best-in-class cloud database performance plus the scalability and security they require.
Oracle utilizes AMD EPYC’s compute density (64 cores/socket) with speedier cores to process transactions and data queries more quickly than any other database or cloud database service, while reducing costs. Oracle also leverages AMD EPYC’s on chip large L3 cache to store and access data much quicker resulting in significantly faster query response times. And MySQL Heatwave takes advantage of AMD EPYC’s industry leading memory bandwidth to deliver much greater data access across the memory channels. All of it adds up to greater Oracle platform and database cloud services performance.
The proof is in the latest unsurpassed Oracle benchmark results.
Exadata X9M on OCI
Exadata X9M on OCI delivers unprecedented OLTP latencies ≤ 19 µs, up to 22.4 million 8K Read IOPS with only 8 database servers , and up to 2.8 TB/s in scan throughput. To put this in perspective, Exadata X9M on OCI delivers 25x better IO latency than all-flash Amazon RDS and 50x better IO latency than all-flash Azure SQL, while scan rate is 384x faster than Amazon RDS and 137x faster than Azure SQL. By any measure, Exadata is the fastest Oracle Database platform in the cloud or anywhere else.
Keep in mind that cloud database service costs directly correlate to CPU processing performance. The AMD EPYC 7003 series’ faster processing results in lower costs.
- Autonomous Database
- Autonomous Transactional Database
- Autonomous Data Warehouse
- Autonomous JSON Database
- Exadata Database Service
- Exadata Cloud@Customer
- Dedicated Region Cloud@Customer
MySQL HeatWave
Oracle also takes advantage of the AMD EPYC 7003 series with the extraordinarily popular and rapidly growing MySQL HeatWave. MySQL HeatWave on OCI TCO/performance blows away the competition. One of the key reasons is the AMD EPYC 7003 series. The proof is once again in the performance and cost numbers.
Based on the industry standard 4 TB TPC-H[1] benchmark (OLAP), MySQL HeatWave on OCI delivers 1100X better performance at ½ the cost of the Amazon Aurora MySQL service. The advantages continue with dedicated OLAP cloud database services. Using the larger 30TB TPC-H2 benchmark versus Amazon Redshift, Snowflake, Google BigQuery, and Azure Synapse demonstrates extremely impressive performance leadership and even more impressive TCO/performance as reported in Wikibon’s recent TCO and TCO/performance research. However, to accurately portray apples-to-apples comparisons with MySQL HeatWave, MySQL services, ETLs, machine learning (ML), and duplicate storage must be added back into AWS, Snowflake, GCP, and Azure.
Oracle MySQL HeatWave OLAP TCO advantages.
- AWS is 642% (4x) more costly.
- Snowflake is 723% (2x) more costly.
- GCP is 512% (1x) more costly.
- Azure is 327% (3x) more costly.
Put another way, Oracle MySQL HeatWave is:
- 84% less costly than AWS.
- 86% less costly than Snowflake.
- 80% less costly than GCP.
- 69% less costly than Azure.
The MySQL HeatWave’s OLAP TCO/Performance advantage is undeniable.
- Versus Amazon Redshift—MySQL HeatWave is 73x better in TCO/performance.
- Versus Snowflake—MySQL HeatWave is 27x better in TCO/performance.
- Versus GCP BigQuery—MySQL HeatWave is 15x better in TCO/performance.
- Versus Azure Synapse—MySQL HeatWave is 36x better in TCO/performance.
MySQL HeatWave ML TCO advantages[2].
- AWS is 1,049% (5x) more costly.
- GCP is 515% (2x) more costly.
- Azure is 389% (9x) more costly.
Put another way, MySQL HeatWave is:
- 90% less costly than AWS.
- 81% less costly than GCP.
- 74% less costly than Azure.
MySQL HeatWave ML’s TCO/Performance advantage increases radically.
- Versus Amazon Redshift—MySQL HeatWave is 82x better in TCO/performance.
- Versus GCP BigQuery—MySQL HeatWave is 88x better in TCO/performance.
- Versus Azure Synapse—MySQL HeatWave is 85x better in TCO/performance.
One of the reasons behind these huge MySQL HeatWave performance and TCO/performance advantages is the performance of the AMD EPYC 7003 series CPU. Another is its scalability. MySQL HeatWave on AMD provides exceptional TCO/performance from a very small cluster all the way up to a cluster size of 64 nodes with about 1,000 cores.
Major Database Problems Solved by Oracle partnering with AMD
- Eliminates database, data warehouse, and ML performance issues thereby accelerating response times, actionable insights, time-to-market, and time-to-revenues.
- Eradicates the conspicuous decline in database performance as data proliferates. Performance decreases are exacerbated by sprawl.
- Completely reverses mediocre-to-poor user productivity caused by slow and variable response times by accelerating them consistently.
- Obliterates sub-optimal database availability from planned and unplanned disruptions/outages—including patching and upgrades that degrades user productivity by making it all non-disruptive with best-in-the-industry uptime and availability.
- Radically reduces database TCO making it much more affordable.
AMD and Oracle Conclusion
Oracle is laser-focused on delivering best-in-class cloud database performance and TCO/performance with a strong emphasis on security, scalability, simplicity, availability, and serviceability. They have a long history of doing just that. AMD also has a very similar focus and emphasis, which makes this partnership so effective.
For those IT organizations seeking higher productivity at lower costs, this partnership is definitely a strong and effective way to go. Whether it be for demanding high performance applications and/or workload consolidation, there is nowhere near as effective and efficient line of cloud database services than those powered by the AMD and Oracle partnership.
For More Information
Go to:
Oracle Autonomous Database Video Explainer
Oracle MySQL HeatWave Video Explainer
[1] Benchmark queries are derived from the TPC benchmarks, but results are not comparable to published TPC benchmarks results since these do not comply with the TPC specifications.
[2] Snowflake currently does not provide its own ML service.
