ERP’s storage requirements and options in the cloud


Enterprise resource planning (ERP) is the central nervous system of many modern businesses. But, like a nervous system, ERP operates beneath the surface. It runs a myriad of processes to connect the organisation’s operations, processes and departments.

If it works well, users might not even know ERP is there and they just do their jobs. But, by just doing its job, ERP can sometimes come lower down as a priority for enterprise IT.

In this article, we look at the key performance characteristics of ERP and the storage it requires, the input/output (I/O) profile of ERP applications built on databases, the types of storage best-suited to ERP, and the options that have become available in the shift to the cloud.

What does ERP do?

ERP runs business functions, such as finance, supply chain management and even human resources (HR), and brings them together on to a single IT platform.

ERP applications are typically modular, with components that run specific workflows, business tasks or even whole departments.

The difference between ERP and standalone business applications such accounting and HR applications, is that ERP shares a database. ERP also has a common user interface (UI), and digital connections between business functions.

Viewed another way, ERP is a way to present an enterprise database to users in a way that makes sense in their job role, but still allows a single repository for business data and consistent workflows across departments. As an integrated system, ERP should give businesses a better view of their operations than standalone applications.

What are ERP’s I/O requirements?

ERP rests on its databases, so it shares the requirements of any enterprise-grade database system. ERP needs large volumes of storage, with low latency and good read and write speeds, as well as high levels of reliability and data protection.

But, as ERP mostly handles structured data, its storage capacity requirements are not as demanding as applications such as business intelligence, or artificial intelligence (AI) and machine learning. ERP usually operates with terabytes, rather than petabytes, of data.

ERP I/O requirements will depend on the organisation’s needs and the industry it is in, as well as the business functions a particular ERP deployment supports.

ERP for financial services, real-time e-commerce or just-in-time manufacturing will put more demands on I/O than a business that can run batch processing. By the same token, logistics and supply chain management will be more I/O dependent than HR or financial planning.

Businesses may well run multiple instances of an ERP system, so they can tailor hardware – including storage – according to performance requirements.

Many ERP instances are now virtualised, says Philip Dawson, a vice-president at Gartner Research who covers ERP.

“Anything up to a few terabytes can be handled in a VM,” he says. “So, you do storage management, memory management and I/O management as part of the database, and we’ve been doing that for many years.”

Another trend is for ERP to run in-memory, with SAP’s HANA system the best known, but this does not eliminate the need to manage I/O requirements. “You still have to load and unload snapshots, backup and recover. It has become just another tier of I/O,” says Dawson.

What storage technologies does ERP need?

Typically, enterprises have run ERP systems on block storage, whether as direct-attached server storage or in SAN arrays. Enterprises used RAID-based arrays to provide capacity, performance and the first level of data protection.

The move towards solid state storage, especially flash, has boosted performance. Flash is especially useful for ERP and its need for frequent reads and write. ERP systems that handle time-critical transactions are most likely to benefit from flash, whether server-attached or networked via a SAN. Vendors of larger ERP systems have tended not to support network-attached storage, although plenty of firms do use NAS successfully to run ERP.

Improvements in NAS technology have also narrowed the gap between block and file storage.

If the ERP system and its database supports it and performance is acceptable there are advantages in running ERP on a NAS or even object storage. These include reduction in the number of storage architectures a business operates, and better integration with cloud storage.

“ERP does so much that there are times you really want to run it in-memory, and others where it doesn’t make sense to do so,” says Tony Lock, at analyst Freeform Dynamics. “Often, big organisations don’t just run one single instance of ERP, but different instances for different parts of the business or different geographies.”

Increasingly, Lock sees ERP running on hyper-converged infrastructure and in the cloud.

ERP in the cloud?

Cloud computing and cloud storage could have a greater impact on ERP than developments in storage technology.

Businesses can opt to run ERP in public clouds, on-premises or in combination. Large enterprise ERP vendors and smaller challengers offer cloud-based options.

At Gartner, Philip Dawson says around a third of businesses now run ERP in the cloud, a third cannot, and the remaining third are in transition to cloud platforms.

When it comes to storage, the potential pitfalls of a cloud deployment come from latency between compute and storage in the cloud, or delays between the ERP system, users and any sensors or other hardware connected to ERP.

“Cable length” still has to be factored in. This is a particular risk in areas such as in manufacturing or logistics, which rely on real-time or near-real time data flows to keep up with the flow of goods.

For most day-to-day purposes, an all-cloud system is unlikely to affect the end user provided they have sufficient internet bandwidth.

Enterprise software vendors also promote their own end-to-end cloud solutions because they can manage performance and the user experience.

But for firms that build their own cloud ERP, storage performance needs to be factored in, with the higher-performance cloud storage tiers being most suited to ERP.

Lastly, there is another incentive to store ERP data in the cloud and that is easier re-use of data, for business intelligence and analytics, including AI.

When data is in the cloud, adding analytics, business intelligence or AI can be done on demand without the need to reload that data from local storage. ERP analytics and data mining are growth areas in their own right. While they might not yet be enough to justify a cloud migration, they are already a useful side benefit.



Source link