Seek cuts SAP core over to cloud

By on
Seek cuts SAP core over to cloud
Seek's Cameron Amling.

As it moves away from running its own data centres.

Seek has moved its core SAP environment into the cloud and switched out the underlying database as part of a continued effort to reduce its data centre footprint.

Enterprise applications technical manager Cameron Amling told the recent AWS Summit in Sydney that the company went live with SAP on AWS in mid-November last year.

Until then, Seek had run its SAP system - which is predominately CRM, together with only the finance module of ECC - out of its own data centre environment.

“We had two data centres,” Amling said.

“[In the primary data centre, there were] seven application servers, Microsoft clustering for our central services, and then we had SQL Server always-on across four database servers [for the SAP environment].

“In the second data centre, we had to mimic that exact same footprint, albeit idly, in case there was an event.”

Amling said Seek had a “strategic direction” to move away from running its own data centres and to move as much of its infrastructure as possible up into AWS.

However, running SAP on a SQL Server database backend in the cloud presented unpalatable licensing options.

“When you run SQL Server up in the cloud there are certain restrictions under the licence agreement you have to have under SAP,” Amling said.

“We didn’t want to run on those restrictions so we thought if we’re going to run on the cloud we need to change our database.

“Being on SAP the logical progression there was to move to HANA but that in itself is a ‘catch 22’.

“If we were going to go to HANA, we need some pretty powerful compute, and that compute’s a lot easier to get in the cloud.”

Amling went to SAP to sort out the HANA licensing.

At the same time, though, he took the opportunity “to review what else we had from SAP as well as what we were going to need from SAP in the future”.

“We rejigged our entire SAP landscape and shrunk our on-premises SAP footprint, which actually made it a lot easier to move because there was less of it,” he said.

Seek was also able to take advantage of an earlier project that had also been aimed at keeping the size of the SAP environment contained.

“Fortunately we had already done an extensive archiving strategy early in 2017,” Amling said.

“We were running out of space in our data centre and since our goal was to get out of the data centre, we were forced to actually have a look at what we were using from a data perspective [to avoid having to] go out and buy more storage for that centre.

“So we were already pretty good in terms of having done our housekeeping and cleaned up our systems.”

However, Amling had only a seven-month window in which the migration of SAP into AWS could be completed.

Having rejigged its SAP footprint, Amling said that “to take advantage of what I’d agreed with SAP we had to do it within the next licence window which started on December 31 [2017]”.

“We signed the agreements in late June/early July, which gave me a very short timeframe to get the whole landscape moved up and onto the cloud, and then to HANA,” he said.

SAP is considered by Seek to be “mission-critical” and integrates with a range of systems including Ascender, Workday, Anaplan, Redshift, Zendesk, Tableau, Reval, Concur, Xero and Salesforce.

In addition, Amling noted that “basically every job ad that goes onto the Seek website ends up in SAP”, putting pressure on the project team to get the migration right.

Seek spent three months from July to September last year planning the move, using a mix of its own staff as well as consultants from AWS and DXC Oxygen.

Having sized the environment it wanted to move, Seek set about standing up appropriate infrastructure instances for the project.

“The really important part was getting our landscape design right,” Amling said.

“We were already running a Highly Available (HA) solution in our data centre, so we needed to have a HA solution up in the cloud.

“AWS do produce a white paper and we used that as a starting point, but we got solutions architects in to help us design the HA architecture.

“[In AWS] we’ve got the same amount of compute from four application servers,” Amling said, noting that Seek had decided to run four Windows m4.xlarge instances with 4vCPUs and 16GB RAM.

“We only have one central services. On the database side we run HANA real-time replication and that’s sitting on top of SUSE Linux and we chose SUSE because that has the SAP high availability agent to handle the failovers,” he said.

Seek started its migration by cutting its dev environment over at the end of September last year.

“We did a couple of sandpit migrations just to try it out, and then did our dev [environment],” Amling said.

“For our dev migration we did to a full database export, copied the files across to S3 and did an import from there.

“We did the same approach for staging, and the reason we did that was so it didn’t go through our Direct Connect. Our Direct Connect funnels a lot of our website traffic and we didn’t want to impede that in any way.”

The team did “another couple of trial migrations after the staging migration” before standing up the full production environment in AWS.

While this “did go across the Direct Connect”, the company limited the impact of that in part by performing the migration well outside regular hours.

With the environment up in AWS, the team then tried to optimise the amount of compute it was consuming.

“We can see quite detailed what our SAP spend is,” Amling said.

“We went live in mid-November on our production environment, and by Christmas we’d already reduced our application servers by one full instance size and that’s purely because we had greater visibility on our utilisation and we could drop that down.

“Because we only operate in A/NZ, we also turn off our dev and staging environments overnight and that saves us a lot of money as well.”

Amling said the team is working to further improve the HA architecture of the AWS solution, in particular targeting the “one central services cluster” which is currently a single point of failure.

“We’re working with AWS and SAP on a HA solution for that central services,” he said.

“We’re going to continue working on optimising our resources by monitoring them and seeing if we can downsize the instance classes.”

Other work is focused on increasing the performance of SAP on HANA.

Got a news tip for our journalists? Share it with us anonymously here.
Copyright © . All rights reserved.

Most Read Articles

You must be a registered member of iTnews to post a comment.
| Register

Log In

Username / Email:
  |  Forgot your password?