Way back in October 2016, not long after IBM announced its “LinuxONE” brand, we wrote a report detailing the differences between LinuxONE system architecture and x86 systems architectures (see “The Major Difference Between IBM’s LinuxONE and x86 Linux Servers” here.) Our key points in that report were that:
- LinuxONE used a different microprocessor (z), allowing LinuxONE to stack multiple workloads and run more virtual machines (up to 20 per core);
- Intel x86 servers had less cache, but threw a lot of processing power at executing workloads.
To me, IBM’s newly announced LinuxONE Emperor 4 is déjà vu all over again. This new system is now based on IBM’s Telum processor, an 8-core, high-speed processor (clocked at 5.2 GHz) that offers extended on-chip cache (to increase thread performance). Now, add artificial intelligence (AI) to this mix, and you get a chip that can drive rapid response times augmented with on-chip, real-time AI inferencing.
When comparing an x86 Linux server to LinuxONE, IT executives need to consider the differences in the core engine that will drive their server. So whatever x86 microprocessor is considered should be matched up against IBM’s Telum, which offers:
- Up to 200 LinuxONE cores across 32 IBM Telum processors;
- A redesigned 32MB private level-2 cache. The level-2 caches interact to form a 256MB virtual Level-3 and 2GB Level-4 cache
- 7nm technology, offering 5.2GHz performance;
- Up to 40TB memory;
- On-chip encryption; and,
- On-chip accelerators for AI and data compression.
This final point—on-chip accelerators for AI and data compression—is worth a close look as it serves to highlight a significant difference between x86-based servers and LinuxONE. Consider a recent LinuxONE customer scenario where a large U.S. bank could not score all of its transactions in real-time, even when using an off-platform scoring engine. As a result of this processing challenge, 80% of the bank’s transactions went unscored. Now add AI and inference processing to the picture.
Using IBM’s LinuxONE, the bank was able to score 100% of its transactions as transaction response time dropped from 80ms to 1ms. The number of transactions processed went from 1800 TPS to 20,000 TPS, and with on-chip AI, the bank was able to reduce $20 million in exposure risk per year. This example highlights that IBM’s microprocessor architecture is different than x86 architecture and can thus process specific workloads more efficiently.
But wait. There’s more…
The new Emperor 4 highly scalable system design offers:
- One, two, three or four 19-inch frames;
- Up to 200 cores overall, and up to 125 LinuxONE cores in a single 19-inch frame; and
- Up to 16 internal NVMe carrier cards for flash storage.
This design can enable up to 20 billion secure transactions per day – meaning that a fully configured LinuxONE Emperor 4 can do the work of almost 2000 typical x86 cores! Not only is the Telum microprocessor a significant differentiator, but the system’s advanced architecture design also adds new efficiencies as compared to x86 servers.
Telum also features significant innovation in security, with transparent quantum-safe encryption of main memory. Also, LinuxONE’s Secure Execution mode has been designed to provide increased performance and usability for Hyper Protect Virtual Servers and trusted execution environments. This advanced security makes IBM’s Telum ideal for processing sensitive data in Hybrid Cloud architectures. IBM’s “Confidential Computing” (protects data in use), pervasive encryption (an option to encrypt data at rest and in flight), and integrated compliance automation also serve to differentiate LinuxONE from x86-based server environments.
With the Telum chipset, IBM has also made further improvements in availability with key innovations, including a redesigned 8-channel memory interface capable of tolerating complete channel or DIMM failures and designed to transparently recover data without impacting response time.
From a resiliency perspective, as part of its LinuxONE Emperor 4 announcement, IBM is emphasizing that with the new climate reality (hotter heatwaves, rolling power outages and various natural disasters), Emperor 4 has been designed to allow users to shift capacity on demand between Emperor 4 systems at different datacenter locations – thus allowing administrators to avoid impending power outages. Further resiliency features include that Emperor 4 has been tested and certified to operate at 40⁰C (104⁰F) – and it will power down at 50⁰C (122⁰F) to protect the system and its data.
Not to be overlooked as part of its LinuxONE Emperor 4 announcement is IBM’s emphasis on “sustainability.” By consolidating x86 server workloads onto an Emperor 4, IBM statistics show that energy consumption can be reduced by 75% and datacenter floor space by 50%, thus helping to reduce carbon footprints and physical space requirements.
The way Intel x86 processors and Telum processors handle workloads is distinctly different. IT executives need to consider the characteristics of their workload and then make their server choice. Further, the qualities of service (security, availability, resiliency, etc.) are hugely different. Buyers should also consider QoS before making a Linux server choice.