How to Organize Infrastructure for SAP Systems Efficiently?
The fact that SAP users will have to adopt HANA In-memory DBMS by 2027 spurs upgrade of hardware-and-software complexes underlying SAP landscapes. There are two options here: classic upgrade of on-premise architecture or outsourcing, with TAdviser revealing details of each approach.
In-Memory is ideal for advanced ERP systems
ERP systems, including business applications based on SAP, is a corporate standard for many large companies. To be efficient, such systems always need powerful infrastructure capable of providing quick access to critical information. In particular, In-Memory approach steps in as being much faster than traditional architecture where data is stored on disk, while RAM is used solely for caching. In case of In-Memory, it is RAM that acts as primary storage processing most of the data. This minimizes the use of the disk subsystem and improves I/O performance.
How to adapt infrastructure to SAPThe so-called "2027 Factor" which means the necessity for SAP system users to adopt HANA In-memory DBMS by 2027 spurs upgrade of hardware-and-software complexes underlying SAP landscapes. There are two options here: classic upgrade of on-premise architecture or outsourcing.
Upgrade of on-premise architectureThe first approach is for enterprises that want to run business-critical applications in their own data center and infrastructure, and prefer capital expenditure (CAPEX). Capital expenditure on investments generally underpin any business and account for more than half of its assets. This is why many large companies opt for their own equipment upgrade, and therefore vendors release already adapted hardware-and-software complexes.
For instance, Cisco offers computing systems certified by SAP for HANA DBMS, application servers, and container infrastructure used in SAP Data Intelligence solution.
Cisco UCS (Unified Computing System) stands out among server solutions because of its architecture: a server system leverages a unified universal fabric enabling integrated management of server and network infrastructures based on policies and templates, as well as easy replication and migration of configurations.
- Bare-metal servers for Linux and HANA DBMS, having on-premise disk subsystem (usually in a rack) or connecting to external data storage systems (in this case, blade solutions are usually used, but exceptions are possible). Such an approach is demanded by large customers that work with large amounts of data and do not opt for HANA virtualization.
- Servers with virtualization platform. Here, we usually have dual- and four-processor blade servers connected to external disk arrays and used for both application servers and virtualized HANA systems.
- Hyperconverged complex HyperFlex based on several dual-processor rack nodes with internal disks underlying a unified distributed shared storage subsystem. This option is ideal for both application servers and relatively small HANA systems and can be considered as a complete "all-in-one" solution for small installations.
"We can name two technical solutions that often help reduce the total cost of a complex. First, Intel DC Persistent Memory (DCPMM) modules with big capacity (up to 512 GB per module) that are installed into server RAM slots and can sufficiently increase the volume of the HANA database supported by the server (e.g. by up to 18 TB per four-processor server). Another important feature of this solution is a much faster restart of HANA system since most of the data is in DC persistent memory (DCPMM) and if the server reboots, there is no need to upload data from the disk subsystem once again.
Second, Cisco HyperFlex, a hyperconverged complex, is a good option when optimizing infrastructure for small SAP landscapes as it eliminates expenses on traditional data storage systems and SANs and helps avoid hiring and training relevant specialists", explained Evgeny Laguntsov, Solution Architect and Head of Data Center Practice, Cisco Russia.
Infrastructure OutsourcingRunning infrastructure and applications in the cloud, or data center outsourcing are gaining in popularity. Such services are especially demanded by customers with limited resources and without great IT expertise, as well as by those implementing SAP for the first time. Cloud migration allows corporations with large system landscapes to achieve a brand-new level of risk management, since SLA is now the provider’s responsibility . As of March 2021, the Russian cloud market is quite mature, so it is not a problem to find a suitable provider or go to another one any time if you don’t like the current service quality. Despite its seeming complexity, migration from one cloud platform to another can be done rather quickly and with minimum downtime. There are such cases in the market already.
Customers more and more often consider outsourcing SAP infrastructure (equipment leasing + technical support) as a good alternative to a classic on-premise installation. Maintenance by provider actually means a personalized approach and faster response to incidents and emerging tasks; e.g. the provider can quickly arrange the delivery of equipment from its own local partners so that the customer does not have to wait for months.
Changing of SAP system hosting provider: CROC experience
At the beginning of 2020, CROC Cloud Services, a cloud unit of IT company CROC, helped a large pharmaceutical enterprise to migrate its SAP systems from one hosting platform to another in the shortest time. On average, similar migration takes six months at least.
"Building a dedicated SAP-oriented infrastructure for large corporate customers has certain specifics. Firstly, such an infrastructure shall be based on the equipment certified by the vendor. Secondly, RAM per server is important; our project implies more than 2 TB. We design a computing environment in our data center in accordance with the request", noted Maxim Berezin, Business Development Director, CROC Cloud Services.
Migration goes in several stages: conclusion of a contract, preparation of project documentation, negotiations with the current provider, preparation of a target infrastructure, test migration and debugging on a bench, actual migration of systems and, finally, taking up the system maintenance tasks.
"Equipment delivery period is the first thing to pay attention to. Certified hardware for SAP NAHA that complies with the software manufacturer's requirements is delivered within 10-12 weeks on average, but in some periods (we implemented our project at New Year's time) this term may be extended yet for another month.
The first batch of the equipment arrived shortly before the holidays, which allowed us to deploy some of the systems. Since not all of the equipment had arrived, the vendor and distributors arranged for us equipment for temporary use. We received the rest of the target infrastructure only at the final stage", commented Alexader Khaidukov, Project Manager, CROC.
SAP HANA DBMS ran on a target infrastructure featuring unified data storage without system replication, and Pacemaker-based SUSE HAE cluster. Such a solution provides the following advantages: unlimited scaling of disk capacity, ability to change or add extra server capacity into the infrastructure without data migration or copying, as well as lower cost compared to the architecture that stores data on local drives.
Application servers were deployed in VMware virtualization environment. The infrastructure also included a dedicated monitoring system, backup system, and other infrastructure services.
All the production and pre-production SAP load was put on all-flash storage system, while the infrastructure services, as well as testing and development environments, were deployed on a separate hybrid array.
The backup system was created using Veritas Netbackup with altering scripts for copying MCOS configurations.
Zabbix tools monitor the entire equipment, OS and SAP, and, in case of emergency, make it possible to create a request in Jira incident management system and send relevant message in Telegram channel.
"All migration steps were scheduled accurate to the minute, which was a cutover plan listing all the tasks, fulfillment time and responsible employees. All the steps were practiced during the test migration, so, during the actual migration, we just had to follow the plan and coordinate the process", noted Alexader Khaidukov, Project Manager, CROC.
The SAP infrastructure migration by CROC Cloud Services started on October 29, 2019 and ended on February 9, 2020.
CROC designed and implemented an infrastructure for a pharmaceutical company in a dedicated segment of its own outsourcing data center, taking on all the support up to application software level. As of March 2021, itelligence, CROC's partner specializing in SAP system maintenance, is responsible for the business service operability. Thanks to the project, the customer not only got a better service, but also managed to halve its SAP application maintenance costs.