A Petabyte customer faced a significant challenge to swiftly reduce operational costs. One focal aspect of their optimization target was the IT expenditure, the largest cost of which was in hosting their datacenter. Their target was to reduce infrastructure costs by at least 45%.
For the customer, Petabyte had run an SAP/ECC FMS (Fashion Management) in a HANA database since 2016, with over 100 users, 1 TB of data and 27 Linux servers. All this was hosted in AWS. “If we couldn’t reach their optimization target,” said Douglas Bernardini, Services Owner at Petabyte, “we may have been forced to consider discontinuing our current infrastructure and replacing it with an extremely basic ERP simply in order to reduce our costs. This wasn’t something our developer teams were keen on doing!”
What Douglas needed to find was a solution that would significantly reduce costs quickly and without hassle.
Douglas’ first solution was to schedule the relevant instances to be reducing during non-working hours. This, however, only achieved an overall cost reduction of 15%, far below the target of 45%. Douglas quickly realized that spot instances were the only way to achieve the target savings. But the risks of instance terminations would seriously compromise the continuity needed for SAP applications.
“We had to maintain integrity and guarantee application quality, availability, and performance,” said Douglas. “It would not be acceptable to jeopardize performance for the final users or interrupt business processes in unplanned SAP stops.”
Why Spot on AWS
During Petabyte’s search for AWS spot instance experts they decided to engage with Spot, an Amazon Advanced Tech Partner with RealCloud, Spot’s representative in Latin America.
“After our review process, we chose Spot as they best understood our technical needs, along with RealCloud for having wide-ranging and extensive experience with AWS Spot Instances” – Douglas Bernardini
The team planned to move over their workloads in three phases:
- Dev and QA instances
- Non-ERP solutions (PI, Solman, Web Dispatcher, Portal, BW)
- SAP/ECC production, applications and the HANA database
A collection of technical adjustments inside SAP were made:
- Applying 17 SAP nodes from support to adjust SAP/ECC for AWS spot instances
- Orchestrating stop/starts and scale up/down procedures in stateful Linux instances to adapt to demand needs
- Upgrading their HANA database to version 2.0
- Split SAP application instances to enhance compute capacity for peak workloads
- Using AWS automation to erase unnecessary disks (EBS) and to delete old AMIs.
- Creating a completely new back up process
These procedures took only 21 days to implement, with all changes being executed in parallel with the running SAP production, preserving the continuity of the business processes. This meant that IP addresses, disks, ephemeral data and connections with external legacies could all be preserved during the process.
Petabyte are already bringing their customers significant benefits from AWS that wouldn’t have been possible without Spot. “Saving money and running our applications with the highest possible discount is amazing,” said Douglas. “Plus, we’re able to provide SAP support with complete confidence that the environment we’re using is highly reliable and secure.”
Petabyte’s customer enjoyed an infrastructure cost reduction by over 65%, with the company estimating that the dual effect of the improved scaling and the use of spot instances will save them hundreds of thousands of dollars over the next five years.
Petabyte is a cloud compute consultancy firm, providing infrastructure intelligence for enterprise applications. Their team of IT architects focus on digital transformation of large and medium-sized companies, and specialize in cost optimization and performance improvement.
IT infrastructure projects are Petabyte’s core-business. These include: ongoing cloud management; datacenter and application migration; operations integration enhancements; security and compliance improvement; databases migration and management; big data and machine learning servers; infrastructure for artificial intelligence (AI) Mobile devices.