New Heights in Power and Performance

Multi-core processors provide a significant improvement for the IT infrastructure. It will provide more choice for server management including consolidation, virtualization, space, density, power, and cooling. One of the most significant business benefits is that the multi-core processors allow for workloads to be moved off of more expensive and larger platforms to multi-core-rich (and more cost-effective) servers. For the CIO this will help lower cost of computing and importantly position the future technology platform for a more open computing environment, supporting the rewiring of the value chain in financial services.

Multi-core processors, which consist of multiple processing execution units or “cores” on one chip, promise to boost computing power, allowing servers, workstations, and PCs to perform more functions simultaneously. By the end of 2006, major chip vendors (AMD and Intel) expect more than 80 percent of their server products to be shipping with multi-core technology. Because most server-side software is licensed “per processor,” it has caused confusion among some software vendors regarding whether to charge their customers “per processor” or “per core.”

Microsoft has been driving thought leadership in this area by charging the same amount per processor, regardless of how many cores are in the processor.

Michael K. Wons Technology Strategy Director, Woldwide Financial Services Group

Microsoft was the first database vendor to make this announcement, in October of 2004, and continues to be the only vendor to date that has taken this position. This strategy is based on the belief that multi-core processors are a natural extension of Moore’s Law (that the number of transistors on a chip doubles about every one to two years), and that the benefits should be passed on directly to customers.

For example, if you are using SQL Server Enterprise Edition on a four-processor server with dual-core processors, using all eight cores (two cores x four processors), you will require only four processor licenses. That can mean substantial savings for customers. In addition to passing on to customers the cost-savings based on the multi-core licensing policy, SQL Server also supports feature-inclusive packaging where customers do not pay separately for enterprise capabilities such as manageability, high availability, and high-end business intelligence capabilities.

Multi-core processors help further enable server consolidation. Today, poor server utilization is continuing to drive organizations to consolidate their server farms. Firms use virtualization technology to collapse multiple, overlapping operating system and application instances running on multiple, distributed servers onto fewer, larger systems. Using new dual and multi-core processor systems, administrators can more easily deploy OS instances to specific cores, effectively doubling capacity utilization. With additional virtualization technology, firms can save money on systems management and eliminate the need for costly and time-consuming ports of older applications to new operating platforms.

---

AMD:	Raising the Bar on Performance Per Watt

AMD began designing the first x86 multi-core processors in the late 1990s and was the first company to announce plans to deploy multiple x86 64-bit processors on a single die. In April 2005 AMD introduced the world’s first 64-bit, x86 multi-core processors for servers and workstations at its second-anniversary AMD Opteron™ processor celebration. The company has been continuing to drive technology leadership in multi-core computing through R&D and customer-centric investments. Pat Aughavin, senior business development executive, AMD Financial Services Sector discusses the company’s efforts and their significance for the financial industry.

Pat Aughavin Senior Business Development Executive AMD Financial Services Sector

Performance per watt is at the heart of this industry; what level of performance per watt can companies now expect from AMD?

Companies can expect AMD to continue to deliver leading overall performance-per-watt. When purchasing servers, decision-makers look for three things: performance, price/performance and performance-per-watt – AMD64 is the only x86 architecture that can deliver a balance of all three on 1-, 2- and 4-processor server platforms. Specifically, the latest evolution of the AMD Opteron processor with Direct Connect Architecture demonstrates the company’s continued performance-per-watt leadership through high-performing, low-power DDR2 memory and consistent 95W standard thermal design power with low-power options.

How can AMD’s performance per watt translate into profitability for the financial services sector?

The AMD Opteron processor can help reduce Total Cost of Ownership (TCO). When Quad-Core AMD Opteron™ processors are available, customers can plan on a non-disruptive upgrade from Socket F (1207)-based Dual-Core AMD Opteron™ processors to Quad-Core AMD Opteron™ processors within the same socket and thermal infrastructure.  Only AMD has stated plans to offer dual-core x86 server processors that are planned to be upgradeable to quad-core within the same thermal envelope. This is the same way AMD addressed the upgrade from single-core to Dual-Core AMD Opteron™ processors in 2005. This type of predictability is what datacenter managers look for, particularly as they try and budget for growth and what they can expect to save on power and cooling.

In addition to the ability to help reduce operating expenses, AMD Opteron™ processor-based server and client machines provide powerful computing technology that can help increase individual productivity and provide the performance necessary to accelerate ‘time-to-insight.’

It seems for many years financial firms have been coping with vast increases in the amounts of data they manage. What type of data management challenges have clients approached you about?

Market data volumes are exploding in equities and options due to algorithmic trading.  Futures, options and foreign exchange volumes are also growing rapidly. In addition to this, many companies are trying to analyze their own customer data more effectively in order to improve service and efficiency. New offerings are key to growth and competitive advantage in this industry. Global profitability analyses by customer/product/geography are driving additional data management requirements.

Financial firms need high-performance, reliable, computing technology that utilizes the advantages of 64-bit and multi-core technology and helps turn insight into action and make better decisions – faster.

One of the ongoing pain points in this industry is compliance. A newer wave of emerging regulations includes Reg NMS in the US and MiFID in Europe. What challenges will the regulations pose from a processing perspective?  

Reg NMS is generating a lot of quote, order and cancel/replace messages as US equity trading becomes more electronic. In Europe, the Markets in Financial Instruments Directive (MiFID) should lead to higher volumes of market and trade data as well. Every institution is trying to reduce latency in their processing and AMD’s products are leading the industry in this effort.   
Where does AMD now stand in terms of the types of clients it now serves?

Today, more than 90 percent of the top 100 and more than 55 percent of the top 500 of the Forbes Global 2000 companies or their subsidiaries are using AMD64 technology.  

How has that success expanded AMD’s global reach?

AMD’s success in the US has translated to success globally as well. For example, one of AMD’s key customers in Europe is the London Stock Exchange. China continues to be a strong area of growth for us with platforms being offered by local OEMs such as Huawei and Dawning. Recently, Sun Microsystems servers based on AMD Opteron processors were chosen for the largest supercomputer in Japan at the Tokyo Institute of Technology.

Virtualization seems to be an emerging trend in server configuration. How can virtualization help companies make the most of their resources? What kind of virtualization solutions does AMD offer?

x86 virtualization enables data centers to achieve high levels of efficiency and utilization by allowing the division of a computer into several virtual machines. Hardware-enabled AMD Virtualization™ technology is designed to improve all aspects of x86-based virtualization.

AMD Virtualization technology includes added instructions to help reduce the complexity of virtualization software and new architectural configurations to improve the efficiency of switching between the hypervisor and the guest operating systems (OSs).

To meet customer and ecosystem demand, AMD has developed enhancements to the AMD64 micro-architecture in order to enable more efficient virtualization in the 64-bit computing market. Additionally, feature enhancements that are planned for future AMD processor products are intended to further enable the availability of 64-bit virtualization software, and to provide virtualization vendors with the infrastructure needed to offer mainframe-quality virtualization to the x86 market.

AMD’s integrated memory controller is designed to provide efficient isolation of virtual machine memory, to improve security while creating less burden on the hyper visor software. x86 processors that do not integrate the memory controller, like Intel Xeon, can require more memory management be done by software.

AMD’s Direct Connect Architecture delivers, through optimum efficiency of memory and I/O subsystems, the ability to provide an increasing number of virtual machines and/or load in a virtual machine to be handled. AMD Virtualization (inclusive of processor and I/O virtualization technology) enables x86 virtualization to become more streamlined and efficient.

Are there security concerns related to virtualization? How can they be addressed?

The advanced memory handling capabilities offered by AMD Virtualization allow increased isolation of virtual machines and can exclude device access, both of which increase the overall security of virtualization. AMD’s upcoming I/O virtualization technology will increase security by enabling devices to be directly mapped to virtual machines and AMD’s Enhanced Virus Protection (EVP) technology, run with a supporting OS, is designed to act as a preventative measure against certain malicious viruses, worms, and Trojan horses.*

HyperTransport™ technology was invented at AMD with other industry partners, and AMD is a founding member of the HyperTransport Technology Consortium. What business challenges is the company working to address through this technology?

AMD continues to focus on the overall platform, not just the CPU. Customers want overall system balance from a throughput perspective, which they can get from AMD’s uniquely enabled Direct Connect Architecture. HyperTransport technology and an integrated memory controller are key technologies that allow AMD to deliver this type of balanced architecture.

For more information go to http://enterprise.amd.com.

* Enhanced Virus Protection (EVP) is only enabled by certain operating systems including the current versions of Microsoft® Windows®, Linux®, Solaris and BSD Unix. After properly installing the appropriate operating system release, users must enable the protection of their applications and associated files from buffer overrun attacks. Consult your OS documentation for information on enabling EVP. Contact your application software vendor for information regarding use of the application in conjunction with EVP. AMD strongly recommends that users continue to use third party anti-virus software as part of their security strategy.

---

Intel:

Driving an Evolutionary Shift in Computing

Intel: Driving an Evolutionary Shift in Computing

Intel is working to ensure that its multi-core processors become key building blocks in improving the speed, manageability and security of real-time business transactions. Intel’s multi-core architecture could eventually feature dozens or even hundreds of processor cores on a single die. The company is on the fast track to deliver multi-core processors in high volume across all its platform families. Eric Doyle, global industry manager for financial services, discusses the technology’s potential for the financial industry.

Eric Doyle Global Industry Manager Financial Services

Why is the level of performance of multi-core processing superior to what can be achieved with a single core? 

With a single core processor the clock speed dictates the major portion of the processing performance. Multi-core enables multiple compute engines in the same power envelope; providing increased platform performance without the restrictions created by power/thermal issues.  Applications like pricing, algo-trading and even traditional ad-hoc data mart style functions will see immediate benefit.

How aggressive is Intel on multi-core processing in terms of product availability? 

Very aggressive.  By the end of 2006, a majority of Intel server, desktop and mobile processors will be dual or multi-core enabled.  But it’s also important to recognize all the work Intel does prior to launching a product line like Intel® Core™ 2 Duo. We follow a very rigorous process of qualifying, seeding, testing and validating our products within a large ecosystem of customers, ISVs, OEMs and others to ensure our products are production ready at launch dates. 

What are some of the usage models in the financial industry that could be affected by the performance improvements and other benefits of Intel’s multi-core systems? 

As you can imagine, having  multi compute engines greatly increases the available compute power and resources on a system.  This additional compute power gets us closer to real-time computing even with the huge increases in market data. In addition to increased overall system performance, the dual core capabilities on the desktop enable the running of separate operating systems on separate cores so traders may run daily Windows applications on one core and leverage the resources of other cores for advanced, highly demanding trading applications to execute trade strategies. Multi-core enables even more resources to be pooled for a variety of groups or individuals (portfolio managers, quants for back testing and algorithmic trading, trade pricing and analytics, etc.). Dual core solutions on the desktop allow for greatly improved VoIP and video capabilities without degradation to your office productivity applications. 

There has been much talk about the increasing electricity costs of running a data center. How can multi-core processing cut those or other costs? 

We continually face increased need for processing power in the data centers while facing limitations presented by limited access to power and cooling in the data center.  Often our customers are thinking in terms of performance per watt.  This is where Intel is taking a leadership position through the introduction of our multi-core products. Products like the Dual-Core Intel® Xeon® processor 5100 series for servers offer increased processing performance through the use of dual and multi-core processors while reducing the power required to operate these solutions. Couple this with the Core 2 Duo solutions on the desktop, and you get increased performance with improved performance per watt characteristics. So the power savings obtained on the desktop can be used to increase your data center capacity.

What are some of the ways you envision multi-core technology might be put to use down the road, in terms of virtual technology? 

I believe one of the more interesting opportunities of dual and multi-core solutions can be found on the trading floor. Imagine the available computational power existing on a trading floor populated with hundreds or thousands dual core enabled desktops. When the trading day is done, what if you could re-provision that computational power on the desktop into a powerful grid computing infrastructure where you can run your clearing and settling solutions, or your risk management solutions. Using virtualization to take advantage of the available computing power across the enterprise will certainly drive new usage models.

What commitment has Intel made toward multi-core in terms of training and building a community around it? 

As mentioned earlier, dual core hardware by itself is interesting and when combined with software designed and optimized for dual and multi-core, you have a high-performance solution delivering capabilities far beyond a single-core based product.  Intel maintains a close relationship with dozens of Financial Services ISVs, providing seed units and software development expertise to these ISVs. In addition, those products and services Intel offers to ISVs are also offered to the In-house developer community. We collect best known methods for developing code from our own software Application Engineers, and offer this information to ISVs and In-house developers via a vast array of face-to-face expertise and Web-enable materials to enable the developer community to tap into Intel’s capabilities to drive MC usage and adoption.  Intel is committed to ensuring the entire technology echo system is able to take advantage of new technology when it’s launched.