The use of data analytics has become a necessity for businesses to extract insights from raw data and enable informed business decisions. These capabilities require fast and efficient access to datasets for handling complex queries. As datasets are becoming larger, moving from terabytes1 to petabytes1, using in-memory only access has limitations in vertical scaling of memory. Horizontal scaling of memory across multiple servers for distributed applications can become expensive to account for large datasets. As such, high performance storage devices that support higher capacities are required to ensure that these queries are not only completed quickly but are also cost effective.
HDDs are a legacy data storage technology, and though they use fast interfaces like SAS, they are limited by slow, spinning disks and mechanical read/write heads. These limitations mean high latency, low throughput and a lot of waiting for your analytical queries to finish. SSDs are built for speed and efficiency that take advantage of fast storage interfaces and can run advanced analytical queries, obtaining insights in near real time.
A managed storage area network (SAN) is great for data analytics because they provide the performance (IOPS, throughput), scalability and low latency needed to handle massive datasets. They can be built with HDDs, SSDs or a hybrid configuration to store and provide reliable access to large volumes of data. However, if you are relying on traditional HDDs, you’re likely not getting the performance and efficient access to data you need for data analytics. A SAN with all SSDs enable near real-time insights, faster querying and efficient data processing for complex analytical tasks and large data warehouses while also offering centralized management and data protection.
Data Analytics Lab Test: SSDs vs. HDDs
To see the differences in performance between HDDs and SSDs, our Innovation Lab compared the drive performance of a data analytics workload featuring an all-SSD configuration versus an all-HDD configuration in a managed SAN environment. An HPE MSA 2070 SAN was used with dual controllers and Modular Smart Arrays Data Protection Plus (MSA-DP+) advanced RAID technology that distributes spare capacity across all drives in a disk group to enable faster rebuild times. The managed SAN was configured with KIOXIA PM7-R Series Enterprise SAS SSDs and enterprise SAS HDDs from a leading vendor in an MSA-DP+ storage pool. We chose KIOXIA PM7 Series SSDs as they are designed for read-intensive, data-heavy workloads. They are dual-ported and can utilize a 24G SAS (SAS-4) interface. However, the comparison was run at the same interface speed2 as the Vendor A HDDs to make the comparison as fair as possible.
The HammerDB3 benchmarking tool was used for each drive configuration to run the data analytics queries against an Oracle® database. The data analytics workload was created with a scale factor of 1,000 and ran with one virtual user that performed the queries in succession against the database. It was also then run with seven virtual users that performed the multiple series of queries in parallel against the database. Each virtual user performed 22 different analytical query requests. The tests included maximum read throughput, maximum read latency, geometric mean of query completion times and completion time for entire query set.
For Data Analytics, SSDs Blow HDDs Out of the Water
When we ran the analytics queries against an Oracle database in the all-SSD configuration and all-HDD configuration, for both 1 virtual user and 7 virtual users, the SSD results were significantly better:
With 1 virtual user, the all-SSD configuration delivered4:
- 2.07x higher maximum read throughput
- 3.50x lower maximum read latency
- 2.35x faster query completions
- 2.43x faster entire query set completions
With 7 virtual users, the all-SSD configuration delivered4:
- 2.14x higher maximum read throughput
- 7.00x lower maximum read latency
- 2.06x faster query completions
- 2.47x faster entire query set completions
In short, the all-SSD configuration outperformed the all-HDD configuration in every metric tested with the given data analytics workload. These SAS SSDs fully saturated the available throughput from the HPE MSA 2070 SAN to the client server and could handle multiple users who sent queries concurrently. This performance resulted in faster querying completion times and reduced waits for data retrieval. If you’re running data analytics on big datasets and you want faster and higher performance, an all-SSD configuration in a managed SAN environment is the preferred option versus traditional HDDs.
More information on the test comparison is available here.
NOTES:
1 Definition of capacity: Kioxia Corporation defines a megabyte (MB) as 1,000,000 bytes, a gigabyte (GB) as 1,000,000,000 bytes, a terabyte (TB) as 1,000,000,000,000 bytes and a petabyte (PB) as 1,000,000,000,000,000 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, 1TB = 240 bytes = 1,099,511,627,776 bytes and 1PB = 240 bytes = 1,125,899,906,842,624 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.
2 KIOXIA PM7 Series Enterprise SAS SSDs utilize a SAS-4 interface, which can have interface speeds up to 22.5 Gbit/s, but also support 12 Gbit/s interface speeds that the HPE MSA 2070 system controllers utilize. The HPE MSA 2070 cabling configuration used to generate the reported results in the comparison was limited by two 12 Gbit/s SAS cables connected to the client server. The reported results are based on KIOXIA PM7 Series Enterprise SAS SSDs performing at an interface speed of 12 Gbit/s in dual port mode.
3 HammerDB is benchmarking and load testing software that is used to test popular databases. It simulates the stored workloads of multiple virtual users against specific databases to identify transactional scenarios and derive meaningful information about the data environment, such as performance comparisons.
4 Read and write speed may vary depending on the host device, read and write conditions and file size.
TRADEMARKS:
HPE is a registered trademark of Hewlett Packard Enterprise Company and/or its affiliates. Oracle is a registered trademark of Oracle and/or its affiliates. All other company names, product names and service names may be trademarks of third-party companies.
DISCLAIMERS:
Information in this blog, including product specifications, tested content, and assessments are current and believed to be accurate as of the date of this publication, but is subject to change without prior notice. Technical and application information contained here is subject to the most recent applicable KIOXIA product specifications.