With the popularity of all flash, strong storage performance features are becoming more and more intensified. For example, 100,000 IOPS was considered very high in 2008, but now even 10 million IOPS is a performance that many vendors claim to offer. Other features, such as storage systems latency shortened from milliseconds to microseconds, has meant that the all-flash market is more competitive than ever.
Misleading Points in Storage Performance Promotion
Despite the many advancements in the industry, many vendors use false advertisements to promote their all-flash products. Let’s look at several misleading points in storage performance promotion.
Point 1: Use performance with a 100% cache hit ratio
When promoting storage performance, many vendors often write “100% cache hit” in the remarks. This means that all data is stored in memory, instead of being stored permanently. The criterion to decide a high-performance IT architecture is to check whether it stores the most frequently accessed data to the place with the fastest response.
The preceding figure shows that hot data at L0 to L2 layers is scheduled by the operating system’s CPUs, while data at L3 to L4 layers is scheduled by the storage system.
This is to say it is almost impossible to have a 100% catch hit ratio in customers’ environments. Because all flash storage applies to not only different industries but also to varying business models within in industry, this claim should be taken with a grain of salt, and customers should thoroughly check their needs against the product.
Point 2: 100% read performance
Mainstream SSDs offer high performance in data reads, but their write performance is lackluster because each time data is written, SSDs erase data from a NAND before writing new data. This process is called “Program/Erase”, which takes 1 ms to 2 ms in MLC/TLC NAND flash, leading to huge performance differences in data reads and writes.
SSDs often reserve a large amount of over-provisioning space and the “Erase/Program” process is not required every time when data is written onto SSDs. Generally speaking, SSD performance deteriorates in read and write, and read-only scenarios. Therefore, real-life performance value must reflect the customers’ business models, instead of just using the value of 100% read performance.
Point 3: Performance stability
Many vendors only a promote reference value using a specific model and under certain pressure, rather than showing their products’ stability.
The stability of storage performance is critical, especially when service pressure changes and in complex service environments. When choosing storage SLA, customers regard stability as one of the most important indicators, so vendors who produce limited results are misleading their customers.
Stability indicates that response time of 99% services must remain stable under a certain value. If services fluctuate, user experience will suffer. However, most customers have not realized this issue.
SPC, the Most Commonly Used Third-Party Storage Performance Evaluation Platform
Test standards (especially SPC-1 test for random IOPS effectiveness) proposed by the Storage Performance Council (SPC) are a common reference for measuring storage products’ performance in the industry. Mainstream vendors such as EMC, NetApp, Dell, HPE, HDS, Huawei, IBM, and Fujitsu have participated in the SPC tests and used test results to promote their products’ performance and cost effective features. This makes SPC-1 a well-known indicator for measuring the performance of products from various brands.
The test specifications of SPC-1 are strict and fair. Its test program prevents deliberate data value manipulation and its test models are similar to mainstream transaction workloads, meaning SPC-1 is a reliable reference to reflect real-life application environments.
- Service type. For production services, the SPC provides the SPC-1 benchmark test guide. For data analysis services, the SPC provides the SPC-2 benchmark test guide.
- Hybrid service workload. In the SPC-1 test, multiple service workloads are used for concurrency testing to verify response times of storage systems in hybrid service scenarios. Main indicators include different data types (text, binary, and sparse), different block sizes (8 KB to 128 KB), and different read/write ratios, such as sequential reads and writes or random reads and writes.
- Varying service pressure. The SPC increases and decreases dynamic workload behaviors to test storage performance of varying service pressure. Specifically, the SPC starts the test from the peak value before gradually decreasing and then increasing service pressure. Then, it simulates two service peaks as shown in the following figure.
Huawei’s OceanStor 18800F V5 has achieved outstanding results in SPC-1 tests.
- Performance value and latency in different pressure scenarios. The OceanStor 18800F V5 has a maximum performance of 6 million IOPS and a latency of shorter than 1 ms.
- Response time stability. Statistics show that 99.984% I/O responses are within 1 ms.
Huawei OceanStor F V5, Groundbreaking High-Performance All-Flash Storage
Equipped with OceanStor V3 converged storage systems rich store of enterprise-class features and high-availability ratings, Huawei’s OceanStor F V5 all-flash storage takes performance and integration to the next level with architectures specifically designed for all-flash layouts. Huawei’s newest offerings provide organizations with a complete data management solution for their critical services and leverage built-in optimizations to enable all-flash, all-cloud, and all-intelligence strategies. The mid-range and high-end offerings extend Huawei’s flash portfolio, adding to those able to meet even broader range of customer requirements.
Huawei is committed to providing continuously higher performance and reliability for all storage customers satisfying ever-escalating demands from the application environments at organizations of all sizes. Huawei Storage has broken many records in SPC-1
performance testing since 2010. The OceanStor 18800 F V5 is the latest offering to break away from the pack, demonstrating the immense advantages of the intelligence-infused flash storage under scrutiny of industry testing.
The OceanStor F V5 uses innovative self-developed flash end-to-end architecture (including multi-core CPU optimization), and leverages adaptive cache and self-developed SSD algorithms to achieve in-depth integration of system software and hardware. Its superior performance provided by all-flash storage helps deliver superb user experience, helping core business easily embrace the era of all-flash storage.
This storage system uses next-generation self-developed SSD controller chips with high performance to work with new-generation efficient patent algorithms, providing users with reliable and high-performance SSD drives. Equipped with self-developed dedicated controllers optimized for all-flash storage, industry-leading system architecture, and multi-controller processing architecture with load balancing, the OceanStor F V5 provides powerful all-flash processing capabilities and fully meets algorithm processing requirements of all-flash systems. Moreover, advanced processor resource allocation algorithms, LDPC error correction algorithms, and RAID 2.0+ technology achieve longevity of SSDs and fast I/O response, unleashing the potential of all-flash storage.
(Contributed by Wang Tao, Storage Product Line)
The post Key Approaches to Selecting High-Performance All-Flash Storage appeared first on Huawei Enterprise Blog.
Source: Huawei Enterprise Blog
—