Recently SAP published on the web a list of all storage vendors who have successfully certified their high end storage system for SAP HANA. This gives the impression that especially in-memory data bases (DB) need powerful storage system; but is this really the case?
The term “in-memory” describes the special characteristic of this kind of data base: all data to be processed are stored in RAM. Consequently no additional data needs to be reloaded or refreshed for a transaction. This means the OLTP (OnLine Transaction Procession – random read write of small data blocks) workload does not apply here anymore.
But if this typical DB workload does not exist anymore, what else requires the need for high end storage systems?
The first challenge occurs at the start of the DB, at that point all data needs to be loaded from the storage system into RAM quickly. This workload is called streaming IO, the continuous reading of data. Other typical applications for this workload are playing videos, or backup or restore of data. Almost any given storage system (based on RAID) delivers high data throughput, typically measured in MB/s or GB/s.
Is storage certification really necessary?
The second challenge is the logging of transactions; this means the writing of DB logs. Here, throughput is not a key factor, but response time is (also called latency), which means how fast a transfer can be completed of one single data block. Here too, any given storage system based on RAID delivers low latency, by caching the data in an NVRAM (nonvolatile RAM) first, then writing it to disk shortly afterwards. The resulting response time is way below one milli-second, 200 to 400 micro second (µs) typically.
Both requirements – high throughput during streaming-read and low latency for single block – are not new because of SAP HANA, but exist since the appearance of DBs.
This raises the question why SAP saw a need for an extensive certification package. It is questionable that there is any storage supplier not able to cope with these requirements.
But storage systems need to deliver much more than these two performance indicators for today’s business. Cloud, virtualization, data protection and integrity are just a few buzz words a data center has to deal with, including SAP HANA Tailored Data center Integration (TDI). SAP HANA (TDI) puts no additional requirement compared to other DB or “standard” SAP.
Integration is the key
SAP systems, with or without HANA, need to be backed up, copied, cloned, or mirrored. The key to success is the integration of all components involved. Since SAP introduced their SAP-landscape management tool SAP-LVM (Landscape Virtualization Management) copy, cloning, and post-procession of SAP instances can be easily achieved. With the new version SAP LVM 2.1 the HANA DB is supported as well.
SAP LVM is the interface between the SAP application layer, including data base and operation system, and the HW, especially the storage system layer. This allows the use of storage based functions, like Snapshot or Flashcopy.
Similar for backup, if these storage based functions are used to shorten the backup time window, integration between the application (typically a DB) and the storage system is required. This can be delivered by a backup solution such as IBM Tivoli Storage FlashCopy Manager.
Every in-memory DB needs “disk storage.” This can be a classic HDD or of a silicon type , like SSD. It is not only sufficient for a storage system to pass the SAP HANA TDI certification – which should be easily achievable – but furthermore it needs to fulfill the needs of today’s data centers. Therefore “Enterprise Class Storage,” as used by SAP, has greater usage than just SAP or HANA applications.
Therefore the required certification for SAP HANA is just the decanting of old wine into new bottles.