• Overview
  • MySQL Solutions
    • ACID-Compliant MySQL Enterprise with InnoDB
    • Reduce Sharding and Consolidate Slaves
    • Cost-Effectively Scale with Blazing Performance
    • Consolidate with Integrated Multi-Instance
    • Benchmarking MySQL
  • NoSQL Solutions
  • Caching Solutions
  • General

Benchmarking MySQL

Benchmarks are very useful in evaluating a MySQL solution's performance, but no one quite believes that any standard benchmark is right for their particular workload. It's true that "your mileage may vary" but standard benchmarks are a good place to start, and it's important to know what to really look for when you want to run benchmarks that reflect your own very demanding workload. One thing we're sure of is that the Schooner MySQL Appliance routinely beats other MySQL solutions — whether on traditional servers augmented with flash memory or with huge amounts of DRAM — in head-to-head comparisons across various benchmarks.

A good place to start is with the DBT2 benchmark, a standard test for OLTP transactional performance. Use this to compare the throughput increase for the Schooner Appliance relative to a traditional MySQL server, where both systems under test are equipped with equivalent processors, Solid-State Drives (SSDs, aka flash memory), and RAID5. The Schooner Appliance easily achieves 5 times the number of transactions per minute (TPM). The performance improvements are even more significant for the Schooner Appliance when compared to a traditional MySQL server using hard disk drives.

That gets people's attention, because we can prove it again and again.

In addition to standard benchmarks, various tests of real-world workloads and schema have shown that the Schooner Appliance delivers order-of-magnitude improvement in performance across different environments. Whether it is a read- or write-intensive workload, a large working set, or a large number of concurrent sessions, the Schooner Appliance comes out ahead — by a lot.

Unless you're really sure that you'll always have a wimpy workload, make the MySQL implementation that you're testing earn its keep — put some serious stress on it. The results below are for a typical performance-sensitive website that relies on MySQL. Results are shown for increasing the database size to ensure the database does not fit completely in memory, and that there is an increasing number of concurrent sessions to test the parallelism of the system. You can test under less stringent conditions, but if you do you're making it too easy. The Schooner Appliance will scale with your needs far more effectively and gracefully than anything else on the market.

Results show the Schooner Appliance's scalability advantage:

• Our performance scales linearly with increasing database size for read, write, and mixed workloads.
• Our performance scales linearly with increasing number of concurrent sessions (even up to 20K connections).
• For large working sets that do not fit entirely in memory, the Schooner Appliance continues to display high performance. In fact, with a buffer pool that is not warmed up, the Schooner Appliance performs exceptionally well.
• Additionally, while some SSDs are known to have poor performance when utilization is high, the Schooner Appliance keeps its high performance even when the SSDs are close to 100% utilized.

A note on the test setup: To ensure that the tests load the MySQL server, no bottlenecks should be introduced in the test environment. Ensure that the test clients do not saturate before they load the server and that there is enough network bandwidth to handle the load.

All we ask for is an unfair trial, with the most demanding workload you can think of benchmarking. We'll win, and so will you when you select the Schooner Appliance for MySQL.