Interested in Accelerating Object Stores in a Cluster Using SSDs? Here are the results from five different workloads using KIOXIA RM7 Value SAS SSDs in a MinIO® Cluster

Many of our customers have large volumes of unstructured data and use object stores as their data storage architecture.  This approach of sectioning data into units called objects, where each object includes the data, metadata and a unique identifier that applications use for easy access and retrieval, allows for the storage and management of large volumes of unstructured data.  Two key benefits include virtually unlimited scalability and low costs when storing large data volumes for data lakes, native cloud applications, data analytics, log files and machine learning (ML), to name a few.

Our lab decided to test throughput and latency of our KIOXIA RM7-R Series Value SAS SSDs in an object storage architecture.  A MinIO cluster was created on four servers deployed with eight SSDs per server, for a total cluster of 32 SSDs.  Two additional servers and a MinIO Warp^™ workload generator were used for load generation of the object store performance tests.

Five different workloads were tested that included AI, web server, cloud storage, restore and backup to emulate common use cases of object stores.  These five workloads included up to four RESTful¹ operation types, GET, STAT, PUT and DELETE.  Each workload was run three times, and their averages were recorded for the cluster’s GET/PUT throughput performance and read/write drive latency².

The distribution of operations in each workload was intended to simulate realistic use cases for object stores.  Object stores are well suited for AI workloads that need to store and read large amounts of data for training models, and the workload simulates data reading, with an occasional checkpoint.  The web server workload simulates accessing website content with logs being written or new content added.  The cloud storage workload is intended to represent multiple users connecting to it, uploading, retrieving and deleting files, while the metadata for files is used for data analytics.  The backup and restore workloads simulate backing up and restoring data from local systems quickly to meet recovery time objectives and recovery point objectives.

The test results demonstrated that the four servers worked as a cluster, and with the KIOXIA RM7-R Series Value SAS SSDs, they were able to deliver high cluster throughput and low read/write latency per the charts below:

Of the five workloads tested, the highest GET throughput delivered by the MinIO cluster was more than 23,500 megabytes³ per second (MB/s) on average.  The highest PUT throughput was more than 13,000 MB/s on average.  The maximum network bandwidth that was achievable by the load generators was 25 gigabytes³ per second (GB/s).  Four of the five workloads achieved a cluster throughput of about 23.5 GB/s.  The highest read latency was 1.07 milliseconds (ms) on average, while the highest write latency was 5.9 ms on average.  Most of the workloads saw read and write latency under 2 ms.  If you’re looking to accelerate object stores in a cluster, this tested configuration represents an excellent MinIO cluster solution that also delivers low drive latency.

The full testing and results, including average cluster CPU utilization and average drive utilization, are available here.

KIOXIA RM7 Series Value SAS SSDs are available in two 2.5-inch⁴ configurations: RM7-R Series for read-intensive, 1 DWPD⁵, up to 7.68 terabyte³ (TB) capacity applications, and RM7-V Series for mixed-use, 3 DWPD, up to 3.84 TB capacity applications.  The performance specifications⁶ include:

• Sequential Read = up to 1,100 MB/s
• Sequential Write = up to 1,050 MB/s
• Random Read = up to 190,000 IOPS
• Random Write = up to 55,000 IOPS

Additional information is available at the KIOXIA ‘Life After SATA’ site.

FOOTNOTES:

¹ RESTful systems are stateless and separate the concerns of client and server.  Those based on HTTP rely on predefined methods.  Common predefined methods include GET, PUT, DELETE and POST.

² Read and write speed may vary depending on the host device, read and write conditions, and file size.

³ Definition of capacity – Kioxia Corporation defines a megabyte (MB) as 1,000,000 bytes, a gigabyte (GB) as 1,000,000,000 bytes and a terabyte (TB) as 1,000,000,000,000 bytes. A computer operating system, however, reports storage capacity using powers of 2 for the definition of 1Gbit = 2³⁰ bits = 1,073,741,824 bits, 1GB = 2³⁰ bytes = 1,073,741,824 bytes and 1TB = 2⁴⁰ bytes = 1,099,511,627,776 bytes and therefore shows less storage capacity. Available storage capacity (including examples of various media files) will vary based on file size, formatting, settings, software and operating system, and/or pre-installed software applications, or media content. Actual formatted capacity may vary.

⁴ 2.5-inch indicates the form factor of the SSD and not the drive’s physical size.

⁵ DWPD: Drive Write(s) Per Day. One full drive write per day means the drive can be written and re-written to full capacity once a day, every day, for the specified lifetime.  Actual results may vary due to system configuration, usage, and other factors.

⁶ The KIOXIA RM7 Series Value SAS SSD performance specifications provided by Kioxia Corporation are accurate as of this publication.  Specifications are subject to change.  Read and write speed may vary depending on the host device, read and write conditions, and the file size.

TRADEMARKS:

MinIO and MinIO Warp are either registered trademarks or trademarks of MinIO, Inc.  All other company names, product names and service names may be trademarks of third-party companies.

DISCLAIMERS:

KIOXIA America, Inc. may make changes to specifications and product descriptions at any time.  The information presented in this blog is for informational purposes only and may contain technical inaccuracies, omissions and typographical errors.  Any performance tests and ratings are measured using systems that reflect the approximate performance of KIOXIA America, Inc. products as measured by those tests. In no event will KIOXIA America, Inc. be liable to any person for any direct, indirect, special or other consequential damages arising from the use of any information contained herein, even if KIOXIA America, Inc. are advised of the possibility of such damages.