Not logged in » Login
Nov 09 2015

Critical Workload Processing Revolutionized: FUJITSU Server PRIMEQUEST 2000, Second Generation


The IT revolution of the past decades has massively altered the ways in which individuals and organizations work. At the same time, it has sparked a multitude of ongoing debates about issues such as data growth, performance bottlenecks or optimal compute and networking architectures. Equally important, but somewhat less popular is the topic of critical, database-reliant workloads, so-called database applications: in a world that's increasingly IT-dependent, their number is also constantly on the rise. Unfortunately, the number of adequate server systems that can handle such jobs without wrecking budget plans is fairly limited. Fujitsu's second-generation PRIMEQUEST 2000 systems were built to end that scarcity.

Processing critical workloads and supporting essential business functions has always been a main task for IT systems since they first entered corporations in the 1950s and 1960s. However, at the time acquiring a piece of hardware – or even just renting slices on a time-sharing service – amounted to major investments, and so the number of workflows that received such 'preferential treatment' was kept to a minimum. During the 1970s, as computing became more commonplace, more and more companies also began using standardized databases and database applications to perform recurring business processes and functions, e.g. financial transactions. The landscape changed drastically once again when personal computers arrived in the early 1980s: now even small organizations were able to 'computerize' jobs that were previously carried out by hand, such as accounting, debt collection or HR and payroll management. Consequently, the number of critical workloads linked to database applications substantially increased, only to be outclassed by the Internet revolution of the mid-1990s that brought the power of such applications to every PC used at home. Another 20 years later, we are living in a hyper-connected world and working with devices that could bring this model to near-perfection – if it weren't for the numerous cases in which these workloads and applications don't run on servers that are truly up to the task. Analysts have pointed out that oftentimes it seems as though customers could only choose between buying adequate systems that come at a prohibitive price or affordable ones that hardly meet obligatory requirements regarding performance, availability and reliability. Fujitsu's second-generation PRIMEQUEST 2000 series closes that gap.

Eliminating Bottlenecks
When optimizing a system to handle critical workloads, one key aspect that you focus on is performance. More specifically, you'll try to identify and eliminate any possible bottleneck that prevents a server from reaching mandatory data transfer rates or I/O speeds, especially when working with database applications from SAP or Oracle. In the case of PRIMEQUESTs, which typically come with a management server and extra storage attached, our hardware developers designed a new, tiered storage configuration that uses RAM and plenty of PCIe slots for super-fast eSSDs within the system but also externally for processing, caching, and database operations. Fig. 1 below shows how this setup works in a mission-critical FUJITSU Server PRIMEQUEST 2800E2 system.


Fig. 1

Fig. 1: Storage Architectire of PRIMEQUEST 2800E2


As you can see, the system features a four-layer storage infrastructure comprised of both internal and external memory and storage subsystems. Layer 1 is the in-memory database, used by e.g. SAP HANA, MS SQL Server or Oracle DB, which resides in the server's main memory, in this case 192 DDR4 DIMMs that make for a total of 12 TB RAM. Layer 2, also called Database Tier 1, is distributed across 16 PCIe SSDs with 1.6 TB capacity each; the total internal storage space thus amounts to 25.6 TB. Building this layer from PCIe SSDs provides a key advantage, since users can now access information stored in this DB tier almost (if not quite) as fast as if it were kept in memory. The third layer or Hot Cache for frequently accessed data resides on an external subsystem that directly connects to the PRIMEQUEST via the PCIe bus and is equipped with 44 SSDs and a total capacity of 88 TB. Finally, the fourth layer or Database Tier 2 is kept on yet another storage subsystem whose capacity and performance vary depending on whether it builds on SSDs or HDDs; theoretically, the capacity is infinite.

Along with these four layers, we have two more categories of components that are absolutely crucial for processing critical workloads at an adequate speed: the so-called I/O units and I/O boxes – essentially extra sets of PCIe slots that serve to both broaden the storage capacity of a PRIMEQUEST server and keep up its high processing speeds. The main difference between the two is that I/O units are integrated into the system, whereas I/O boxes are attachment devices that can be stacked onto a server. With regard to the above-mentioned layers, I/O boxes connect the Hot Cache to the CPUs via the I/O unit. Likewise, if an application must access less frequently used data in DB Tier 2, it can use direct links between the I/O units and the matching storage subsystem. As a result, data from any storage layer remains available at all times, and there's no single point of failure that could harm application performance or cause system crashes. Moreover, redundancy of all main parts, the presence of enterprise-level RAS features such as advanced memory protection, MCA health checks, PCIe live error recovery, intra-socket memory mirroring or double-device data correction, advanced self-healing capabilities like Dynamic Reconfiguration (on-the-fly, real-time balancing of system resources) and a Reserved System Board (to automatically replace defective system boards without delay) ensure that downtimes and manual maintenance efforts are kept to a minimum. In short, PRIMEQUEST 2x00E2 mission-critical systems provide the same level of resiliency and fault tolerance as a mainframe at lesser costs.

Practical Benefits
As noted before, the PRIMEQUEST 2x00E2 series was optimized to run database applications. Oracle users in particular will benefit from the setup described above, for various reasons:

  • The FUJITSU Server PRIMEQUEST 2000 series second generation E-models build on third-generation Intel® Xeon® E7 v3 product family processors1, which have helped us set several world records over the course of 2015, e.g. in the SAP SD 2-Tier and VMmark V2 benchmarks, thus beating our own results from the previous year. Running Oracle, PRIMEQUEST E2 models offer three times the OLTP performance per core of their predecessors, and about 30% more DB performance than systems in the market.
  • Thanks to our specific approach to an optimized database setup consisting of DDR4 memory and extensions – namely, the use of I/O units and boxes as 'memory extensions' that substantially improve CPU utilization – Oracle users will be able to drastically cut license and support costs: the relevant Oracle Core Factor for PRIMEQUEST 2x00E2 systems is at 0.5.
  • With their specific design and storage architecture, PRIMEQUEST enterprise systems of the second generation are the only x86-64 server systems on the market that have high availability (HA) built in. For Oracle users, this means they can now add functionalities such as archiving at their own discretion and link them to the adequate storage layer, without first having to implement an HA cluster. At the same time, it is possible to add more storage capacity upon request whenever necessary or, better still, to keep an on-demand reserve.

FUJITSU Server PRIMEQUEST 2000 series second generation enterprise grade systems were designed to bring business- and mission-critical compute capabilities to x86-64 platforms and Windows or Linux environments. To achieve this, our developers put special emphasis on building a multi-layered storage infrastructure that not only supports record-breaking application performance, but also provides HA and maximum flexibility out of the box, without the need for extra hardware investment. IT departments planning to speed up critical workload processing and/or consolidate the related infrastructure will hardly find a better deal.

[1] Intel, the Intel logo, Xeon, and Xeon Inside are trademarks or registered trademarks of Intel Corporation in the U.S. and/or other countries.


Comments on this article

No comments yet.

Please Login to leave a comment.


Please login

Please log in with your Fujitsu Partner Account.


» Forgot password

Register now

If you do not have a Fujitsu Partner Account, please register for a new account.

» Register now