SAP’s Data Tiering Options for Data Accessibility, Performance and Cost Management
The continuous and explosive growth of business data is a well-established challenge for today’s organizations.
As data grows, more storage and processing power is needed to scale up and out to meet the demands and complexities of enterprise information management systems. Although data grows unbounded, IT budgets are limited. Therefore, business owners must come up with a data management strategy that both satisfies user requirements for data accessibility and performance, and observes cost constraints.
Fortunately, the value of data tends to change over time. Aged data becomes less interesting and is less frequently accessed, and the largest percentage of active reporting and analytics takes place against the most current operational data available. Managing and storing older data in a cost-effective manner is key for organizations seeking to decouple their data growth from expensive hardware growth.
SAP offers users several technology options, based on cost, scale, and data volume, for managing and accessing each temperature tier of data.
Hot and Cold Data Tiering Options
Hot data is stored in-memory within SAP HANA. Non-volatile memory—known as SAP HANA persistent memory (PMEM)—has a lower TCO and is larger than the classic DRAM (dynamic random-access memory), can be added to SAP HANA to expand overall memory capacity and increase the volume of hot data that can be stored in-memory. Data in PMEM maintains its state after system shutdown, which translates to improved database startup time since data does not need to be loaded into memory from disk.
Cold data is primarily stored on the lowest cost storage tiers but remains accessible to SAP HANA via SQL on request. The cold data tier for native SAP HANA applications and SAP Business Warehouse (SAP BW) applications is managed by SAP Data Hub or the SAP HANA Spark Controller, for data storage in Data Hub or Hadoop, respectively. Customers using SAP BW also have the BW NLS (near-line storage) option to store cold data in an external repository such as SAP IQ.
Warm Data Tiering Options
Let’s turn to the technology options for warm data.
The first option, Paged Attributes, is an alternative that is only applicable to SAP S/4HANA and SAP Business Suite powered by SAP HANA. Paged Attributes allows table columns to be loaded page by page from the HANA persistence layer into memory as needed, without requiring the entire table to be loaded into memory at the same time. Paged Attributes are preferably unloaded from HANA memory to disk based on a least recently used algorithm. Find out more about Paged Attributes.
SAP HANA dynamic tiering is an optional add-on to the SAP HANA database for managing less frequently accessed warm data. It extends SAP HANA memory by providing the ability to create, query, and load data into disk-based, columnar database tables—called extended tables and multistore tables. While easy to implement and manage, along with good scalability to 100TB of compressed data, SAP HANA dynamic tiering supports only SAP HANA native applications. The SAP HANA dynamic tiering server may be deployed on commodity hardware, making it a lower cost option for managing warm data, although with slower performance and reduced functional parity compared to HANA in-memory nodes.
An Extension Node is a dedicated HANA in-memory node for warm data processing that allows the SAP HANA database system to scale out at a lower cost and is appropriate for both SAP BW and native SAP HANA applications. It is, at its core, an additional SAP HANA node with relaxed memory and CPU requirements but with the same functionality as a pure SAP HANA node.
An Extension Node keeps most of its data primarily in the disk-backed persistence layer and loads data into memory only when needed. When a query is executed to access warm data on an extension node, that data is pulled from disk into memory and operates like a normal SAP HANA worker node for processing. If memory is scarce, the Extension Node unloads other data from memory to disk based on a least recently used algorithm. Because the data can reside, for the most part, on disk, an Extension Node can manage up to four times the volume of data as a regular SAP HANA worker node. With a requirement for certified SAP HANA hardware, Extension Nodes are the highest cost but best performing option for warm data management.
Data Tiering: A Trusted Approach for the Intelligent Enterprise
Today’s challenges of balancing and managing rapidly growing data volumes—as businesses continue to accumulate information required to control their daily and strategic operations—are made manageable with the data tiering options from SAP. Organizations can implement a data lifecycle management strategy, that moves data to the storage and processing tier with the cost and performance characteristics best suited for that data. Data tiering options from SAP provide a trusted approach for intelligent enterprises to manage data growth effectively and economically.
Stay tuned as we provide a deeper-dive into SAP HANA data tiering options over the coming months.
Thanks for the informative article. Is there any update you can share on plans to enable dynamic tiering support for solutions such as S4/HANA?
Hi, Brendan: S4/HANA will be staying with paged attributes, and not adopting dynamic tiering. Dynamic tiering is not performant for some of the OLTP type operations that S4 does. Dynamic tiering is intended for analytic type applications, such as a HANA data mart. - Courtney
I am in the process of assembling an Archiving solution and we have an existing landscape of SAP ECC 6.0, S/4 HANA 1610, and Ariba. ECC will get sunset in about 18 months, we are struggling with what functionality to support an archiving approach, a number of resources want to align on Dynamic Tiering and are hoping it will be available in a 1809, any response would help.
S/4 will stick with paged attributes for now for warm data. They will not be using dynamic tiering, because they have OLTP type workloads, as well as analytics type workloads, and DT is not designed for OLTP activity. There is a new capability coming out with SP04 called Native Storage Extension. It is an evolution of paged attributes, and is more scalable. S4 may pick up NSE in the future, but will need to validate it first. - Courtney
Can we consider PMEM in the persistence layer for warm data or is it part of the hot in-memory?
If persistence memory considered as part of the Hot layer can we deploy HANA Dynamic Tiering solution for warm data same time?
Excellently written, clear and concise - you've made something complicated easy to understand. Thank you.
The one thing about data-tiering that I don't quite understand yet is why go to such lengths to literally turn HANA into a traditional DBMS like Oracle or DB2, who natively deal with memory and disk persistence without such complexity.