Linux Workloads on x86 Versus LinuxONE
The LinuxONE system provides all of the benefits of cloud—faster accessibility, greater scalability, high availability and more.
By Suresh Pathak03/01/2019
Today’s IT leaders need to do more with less. With consistent business growth, IT operations are getting more complex while hardware, software, labor and energy are getting more expensive. Businesses demand the right technology at the best value, and IT leaders are looking for ways to reduce cost, increase efficiency and simplify IT operations.
Although commodity servers are inexpensive, they can increase complexity and are often underutilized. The LinuxONE* system is different. It provides all of the benefits of cloud—faster accessibility, greater scalability and high availability—at a lower cost with heightened security, governance and control of your data on-premises.
Here’s what makes the LinuxONE system different from commodity servers:
- Seamless horizontal and vertical growth in one server
- Workloads that might normally run on tens to hundreds of commodity servers can all run on one LinuxONE server
- “Share everything” architecture that offers flexibility for different workloads
- Consistent performance at higher (about 85 percent) average utilization
- Fewer resources needed to run the workloads
- Capability to encrypt 100 percent of your data at rest and in flight
- Lower total cost of ownership (TCO) with reduction in software, labor and data center costs
Workload Consolidation on LinuxONE
Three conditions maximize use of LinuxONE for Linux* x86 workloads from a technical and financial perspective:
- Client software that’s priced per core (including most commercial software such as WebSphere* Application Server, Oracle, Db2*, WebLogic, etc.)
- Easy migration. Software products supported on multiple architectures on Linux minimize the effort to move to LinuxONE.
- Underutilized x86 servers. Distribution of production and non-production workloads across multiple servers inadvertently drives waste and cost.
To prove the benefits of consolidation and to instil confidence in the sizing methodology used, the IBM IT Economics Consulting and Research team selected the configuration from a few clients of their current distributed servers with four different workloads and replicated these environments in a lab. Based on the performance data collected from these servers, we projected the sizing and then ran these workloads on LinuxONE by assigning the projected resources and compared the results. We used various workloads including representative client applications, competitive commercial databases, open-source databases, IBM WebSphere Application Server and Db2. The sizing projections from x86 to LinuxONE were performed based on x86 processor type, CPU utilization, number of workloads being run, overheads of virtualization and type of workload.
When we ran the competitive commercial database on x86 servers we found that if the workload is light on I/O, the SMT2 functionality on x86 helps increase the workload throughput. We ran the workload on x86 bare metal servers and then consolidated workload from 10 x86 servers to LinuxONE.
The actual transfers per second (TPS) delivered on LinuxONE matched or exceeded the projected TPS based on x86 performance data. Depending on the utilization of x86 servers, the core consolidation ratio varied. If the workload was more I/O intensive, the consolidation ratio of x86 to LinuxONE increased. We also observed that when projected sizing is done based on peak utilization, average utilization and number of workloads being run on x86, the actual results are very close to the projected sizing.
We also ran the open-source database workload on x86 with Node.js and MongoDB. We created separate VMs for Node.js and MongoDB with ratio of 2-to-1. Each x86 server had 30 VMs. When we consolidated workload from seven x86 servers (120 VMs) to LinuxONE, we found that the TPS delivered on LinuxONE exceeded the projected sizing based on x86 server’s performance data.
An Effective Sizing Methodology
The IBM IT Economics team’s methodology for sizing the LinuxONE capacity for x86 server consolidation based on performance data of x86 servers works very well for varied workloads. Consolidation onto LinuxONE helps reduce the number of cores significantly which, in turn, reduces the software, labor and data center costs for running these workloads.
Suresh Pathak is an IBM IT economics consultant and leads the IBM IT Economics ream for Asia Pacific.
Sponsored ContentAchieve Compliance Without Impacting Productivity
Post a Comment
Note: Comments are moderated and will not appear until approvedcomments powered by Disqus