Hitachi VSP One Block portfolio now offers a choice between high-performance TLC NVMe media and cost-optimized QLC NVMe media, presenting a new architectural decision point for HANA storage architects. Learn when to pick QLC SSDs for cost‑effective capacity or TLC SSDs for low‑latency SAP HANA workloads on Hitachi VSP One Block storage.
Executive Summary
As SAP HANA landscapes grow in size and complexity, storage architects face increasing pressure to balance performance, endurance, and cost efficiency. With the introduction of QLC (Quad-Level Cell) flash options alongside traditional TLC (Triple-Level Cell) media in the Hitachi VSP One Block portfolio, organizations now have a meaningful opportunity to optimize HANA deployments for both capacity and total cost of ownership - without compromising reliability or certification.
Hitachi VSP One Block brings enterprise resiliency, advanced data reduction, and SAP HANA certification to both media types, enabling deployments that map SAP HANA’s I/O characteristics directly to the optimal drive technology. Lab testing and field data show that QLC SSDs can sustain demanding HANA data-volume workloads when deployed with adequate drive count and supported by the platform’s caching and wear-management capabilities.
For SAP HANA architects, the result is a clear decision framework:
- Use TLC SSDs where latency and write endurance drive the SLA.
- Use QLC SSDs where capacity at a lower cost per GB, efficiency, and consolidation drive the business case.
With this combined approach, organizations can modernize their SAP infrastructure, reduce TCO, and design scalable systems without compromising performance or stability.
Introduction
TLC SSDs have long been the trusted, high-endurance workhorse. However, QLC SSDs promise radical cost reduction and massive capacity density. But can a technology defined by lower Program/Erase cycles truly handle the sub-millisecond latency and write amplification of a mission-critical HANA system?
With Hitachi VSP One Block storage, your SAP landscape is in good hands, using TLC drives or QLC flash as QLC Hits the Storage Sweet Spot: Flash Performance That Won’t Break the Bank.
In this blog post, we will provide help and guidance for your decision on choosing the right drive type for your SAP HANA workload and where recommended capitalize on QLC's economics without jeopardizing HANA certification or performance stability. In short, we try to address cost, capacity, ROI for business and latency, endurance, HANA and storage metrics for engineering.
Hitachi VSP One Block Storage
VSP One Block storage offers worry-free, enterprise-class block storage with certified sustainability for mission-critical applications. With VSP One Block, you have the flexibility to choose the optimal solution to meet the specific demands of your workload.
Select between TLC and QLC drives depending on performance needs and overall cost considerations. VSP One Block with TLC flash is ideal for scenarios requiring superior performance and greater endurance, while QLC is advantageous for workloads that prioritize high capacity and cost-effective storage where data is infrequently modified.
- COMPACT: All NVMe storage platforms with the ability to support up to 3.96 PBe in just 2U minimizing rack space and reducing power and cooling costs.
- SIMPLE: At its core, embedded management with ClearSight, Dynamic Drive Protection always on compression and deduplication (ADR).
- SUSTAINABLE: Sustainability dashboard reports energy usage & CO2 impact, always on compression’ with Advanced Data Reduction (ADR), bezels now contain 40% recycled materials, and Intelligent power consumption with automatic ECO mode during low utilization periods.
QLC and Sustainability
While in the meantime both QLC and TLC drives give roughly the same usable space, QLC drives can be considered more sustainable due to their efficiency of manufacturing and material usage.
QLC drives store 4 bits per where TLC drives only store 3 bits per cell. Thus, the denser QLC drives require 25% less NAND chips per drive of the same size. Because of this, the energy consumed during manufacturing is also lower than for TLC drives, since the energy required per NAND chip, TLC or QLC, is considered constant.
VSP One Block TLC/QLC Supportability
For supported TLC and QLC drive types, see the VSP One B20 Hardware Guide in the Hitachi Vantara Documentation Portal.
All VSP One Block storage systems are certified as SAP HANA Enterprise Storage and ready to be used in your HANA Tailored Datacenter Integration (TDI) landscape. For details on the certification, see the SAP HANA Hardware Directory.
Our reference architecture paper SAP HANA Tailored Data Center Integration on Hitachi VSP One Block 20 provides a great starting point with in-depth information on setup and configuration of your storage for your consolidated SAP HANA landscape.
VSP One Block Architecture: Enabling QLC/TLC Choices
With the availability of QLC drives for VSP One Block Storage, we are Elevating VSP One to a New Level with Samsung QLC.
QLC as a Capacity Flash Tier
QLC flash is not designed to replace TLC in latency-critical tiers. Instead, it establishes a new enterprise flash tier that replaces legacy HDD-based and hybrid capacity layers while retaining flash characteristics such as consistent read latency, high parallelism, and NVMe connectivity. In VSP One Block, QLC enables consolidation of large, predominantly read-centric datasets without the operational complexity of multi-tier architectures.
Key Differences
Before deciding on the right drive type, QLC or TLC SSDs, what are the key differences?
The following table will give you a good idea of the basics before we go into more detail later.
|
Feature
|
QLC (Quad-Level Cell)
|
TLC (Triple-Level Cell)
|
|
Bits per cell
|
4 bits
|
3 bits
|
|
Capacity (density)
|
Higher - more data per physical area → greater capacity
|
Lower than QLC
|
|
Cost per GB
|
Lower / most economical
|
Higher than QLC
|
|
Read performance
|
Good - especially for read-centric workloads
|
Good - generally better balanced than QLC
|
|
Write performance
|
Weaker - slower and more limited under sustained writes
|
Stronger - better for write-heavy workloads
|
|
Endurance / lifespan
|
Lower (fewer write cycles)
|
Higher (more write cycles)
|
|
Best suited for
|
Archival, analytics, backups, cold data, large content libraries
|
Mixed workloads, heavier transactional usage, primary flash storage at better latency
|
From a general storage design point of view, here's a short overview of how to choose:
- Choose QLC SSDs if: Your primary need is to store a large amount of data at a lower cost, and the workload is predominantly read-based with infrequent writes.
- Choose TLC SSDs if: Your application requires high-speed, consistent performance, especially for frequent data writes, and endurance is a key factor.
Additional Context on Use Cases:
- QLC Drives: Best suited for read-intensive workloads, data center archival, high-performance block storage for AI (data preparation, collection, ETL), media, hyperscale, HPC, and cloud storage.
- TLC Drives: Optimized for performance and endurance, making them ideal for larger, mixed workloads.
- QLC Advantages: Cost-effective and higher density, but better for read-heavy or archival workloads where endurance and peak performance are less critical.
Looking at the above, it is important to emphasize that QLC SSDs are not intended to replace TLC SSDs, but to replace legacy capacity tiers (HDD, hybrid, object-backed block) while retaining flash characteristics. In your storage environment, the question is not TLC SSDs or QLC SSDs but where to make the best use of both drive types.
SAP HANA
Understanding HANA IO is crucial for selecting the right storage layout and drive type. The SAP HANA storage requirements Whitepaper provides details about the storage requirements for your SAP HANA system. The I/O Patterns section shows a detailed table for IO happening in the different scenarios of HANA operations.
HANA Volumes
The operating system and the SAP HANA share volume for binaries, trace and config files have very low storage requirements for both performance and size. Thus, we will focus on HANA data and log volumes for a more detailed look.
The Log Volume Requirement:
- Need: Must satisfy extremely low write latency (often sub-millisecond) and high sequential write throughput. Endurance (DWPD) is a major factor due to constant, transactional writing.
- Media Recommendation: TLC SSDs. It offers guaranteed high endurance and consistent low latency required to prevent transaction backlogs and ensure system stability.
The Data Volume Requirement:
- Need: Large capacity, primarily read-intensive access (once loaded), with periodic large sequential writes (savepoints/startup/recovery). Endurance is a secondary concern.
- Media Recommendation: QLC SSDs. This is the ideal tier for QLC drives to achieve maximum cost/capacity optimization while leveraging the VSP One Block controller's read-optimization features.
Without further consideration, the above already sounds like a good guideline. For better understanding, we will have a look at the requirements of both data and log volume, to better understand the needs and base the decision on data, business needs and SLAs rather than gut feeling.
What kind of IO can we expect on the HANA volumes?
HANA Data Volume:
- Async random write to persistent storage
- write transactions, save points, snapshots and delta merges
- contains table data, undo log information, modeling data and more
- Data is read during data backup, HANA DB or HANA node restart and table loads
HANA Log Volume:
- redo log information
- synchronous write to persistent storage when a transaction is committed
- Overwrite after backup operation
See a detailed table in the HANA storage requirements whitepaper linked above.
HANA Workload
During normal operations, there are mainly write operations for the HANA data and log volume while read operations only happen during database restart or failover, column table reload (data volume only) and when performing a log- or full data-backup (both volume types respectively).
See Analyzing Disk and I/O Issues in the SAP learning journeys for a good description of the different types of IO of your HANA database which will provide more details on the above.
With SAP's strategy for SAP HANA as a multi-model database that can handle multiple workload types including OLAP, OLTP and mixed workload scenarios in a single database, it is very difficult to map certain application scenarios to specific IO workload patterns and requirements.
In addition to that, consolidating multiple HANA nodes onto a single VSP One Block storage system in a TDI environment will cause an additional mix in storage requirements.
HANA Monitoring Views to get better understanding of IO pattern
See Monitoring HANA System Resources for details and help on HANA monitoring views that will help understand the IO statistics of your HANA database.
HANA Workload vs. Media Endurance
However, it is still possible to get a good overview of the requirements of the full SAP HANA landscape.
When moving existing HANA environments to a Hitachi VSP One Block storage, historical data from the HANA systems as well as the existing storage unit might be available. Actual size requirements, compression ratio and data growth expectations as well as hardware information down to the drives' "media life remaining" or "percent lifetime used".
This kind of data will help you to make a well-founded decision on the right drive type when considering TLC and QLC drives.
HANA Application Scenario
From a higher-level perspective, we can give some recommendations on choosing the right drive type for your SAP systems and landscapes running on VSP One Block storage.
Choose TLC drives for:
- ERP
- CRM
- very large servers
- everything with low latency requirements, especially for log volume
Choose QLC drives for:
- Consolidation of smaller systems
- OLAP workload with scheduled data loads and reporting
- QA/DEV landscapes with higher capacity demand and lower performance requirements
- Secondary VSP One Block storage that operates as backup or replication target
Media Life Remaining
Hitachi Vantara constant lab testing over multiple weeks can be summarized as a rule of thumb that lower drive counts lead to a higher wear rate.
Lab testing
After 28 days of constant heavy testing with a total of 7.8PB of reads and 6.2PB of writes using a VSP One B26 with 24 x 30TB QLC drives the used-endurance indicator of the QLC drives showed 3%. A rough estimation with constant workload on these 24 QLC drives suggests that it would take approximately 877 days (or 2.4 years) before the drives reach the 95% used warning threshold. With increasing wear rate, there will be more retries, increased thermals, slowed performance and adding risk to drive failure. On VSP One Block storages when the write endurance rate reaches 99%, Dynamic Sparing begins to run, and data is saved to an available spare drive.
Real Life Workload
On real workload outside of a test lab environment, such a constant workload is not always expected and heavily depending on the application type and landscape setup of your SAP HANA environment. Getting an overview of the total reads and writes from HANA perspective will require the accumulation of the data from the IO monitoring views mentioned earlier.
However, on storage level, “Endurance Used” can be reported using Hitachi Remote Ops (HRO) when these statistics are enabled and regularly collected from the existing storage system.
Performance Considerations
Using general workload tests, it can be seen that both, QLC and TLC drives show great peak performance so that other parts of the infrastructure will become the limiting factor, depending on the test cases. With larger dataset configurations, peak performance on QLC drives can drop and higher response times can be seen, which is expected due to wider access range.
Some workload tests show increased response times for QLC drives, which will increase for larger data set sizes, when compared with TLC drives.
Comparing smaller data sets with large data set configuration on QLC drives, lower throughput and higher response times can be observed.
With a small drive count in your storage pool, the performance differences between TLC and QLC drives might be the most obvious, increasing the number of drives will also increase the impact of other factors like caching.
Why Data Size affects Flash Performance
Storage cache misses, including longer metadata cache lookups & misses result in higher processing time. Looking in more details for reads and writes, this can be summarized as follows:
SSD Reads:
- Data Fragmentation: Larger datasets are scattered across NAND, resulting in increased access times.
- Reduced Cache Hit Rate: Large datasets exceed cache capacity, resulting in slower direct reads.
- Error Correction Overhead: More reads increase the likelihood of errors, necessitating additional correction steps.
- Background Task Interference: Internal SSD operations can interrupt or delay reads.
- Mapping Table Latency: Larger datasets result in larger mapping tables, which slow down address lookups.
- Thermal Throttling: Sustained reads also generate heat, which can reduce performance.
SSD Writes:
- Write Amplification: Larger datasets trigger more internal data movement, resulting in increased overhead.
- Garbage Collection: More blocks in use means frequent cleanup, slowing down writes.
- Cache Saturation: SLC caches fill up quickly with large datasets, forcing slower direct NAND writes.
- Thermal Throttling: Sustained writes generate heat, leading to throttled speeds.
- Lower Endurance: QLC NAND wears out more quickly, requiring additional error correction and retries.
Conclusion: Cost-Optimization with Confidence
In short, TLC drives are better optimized for performance and endurance, making them more suitable for larger, mixed workloads. QLC drives, while cost-effective and higher in density, are best used for read-heavy or archival workloads where performance and endurance are less critical.
The performance drop of QLC drives can be considered as a tradeoff to achieve many times larger data sets and is a suitable option for capacity intensive solutions, where performance may not be the top priority.
VSP One B20 QLC will be a great fit for larger landscape consolidation setups, especially with virtualized HANA systems where IO performance with focus on low latency is not the main concern. Using a larger storage pool with the high capacity QLC drives will provide a reliable basis with good performance while keeping an eye on the budget.
For large server size S/4 HANA systems, VSP One Block storage with TLC drives provide the required low latency and increased throughput for high pressure workload of your mission critical systems. With the focus on high performance, TLC drives are the best choice with your VSP One Block storage.
#HitachiIntegratedSystems
#VSPOneBlock