Arm value Geometry Calxeda Server Evaluation report: last Chapter

Source: Internet
Author: User
Keywords Server ARM server

ARM based servers adhere to very low power consumption and excellent per watt performance ratio, it's also possible to put an incredibly large number of servers on a rack--and now you've implemented up to 1000 arm servers in one rack (48 server nodes in a 2U chassis), All of these nodes consume less than 5KW of power (or a four-core arm node consumes about 5W).

But when a new technology is advertised, it's best to remain skeptical. Because people are happy to know something new, the media always advertise and talk about new trends, but in the end, the system administrator has to keep their IT services working and persuade their bosses to invest in new technologies.

There have been hundreds of comments and more on the conflict between arm and x86 server, but there has been no testing in the real benchmark environment, which is what we're going to explore today. We loaded some heavyweight loads on the Boston viridis cluster system where we ran 24 sites--and other applications--and measured the flow rate, response time, and power consumption. We compared it with Low-power Xeon (Xeon) to reveal the current commercial arm server versus the best Intel Xeon products, including the performance-power ratio and performance-price ratio; no matter how much your measurements are, we have reliable data.

At first glance, the relatively low per-core performance of the ARM CPU seems to be a poor partner for the server. There is no doubt that the dominant CPU in the server market is Intel's Xeon processor. The success of the Xeon series is largely due to its superior single-threaded (or per-kernel) performance at medium power consumption (70-95w). By combining this superb single-threaded performance with the corresponding kernel number, you can achieve good performance for almost any application. Economies of scale and corresponding price scales are also important, but if response times are shorter and electricity bills are less, the server market is willing to pay a little more.

One piece of data that proves that single-threaded performance is still important is the evolution of the T-series that Oracle (or what you would call Sun). Sun T3 has 16 cores and 128 threads, but T4 has only 8 cores with 8 threads, and its chief executive Larry Ellison more than once sold its single-threaded performance has been greatly improved to achieve a speed of up to 5 times times. Do we really need another server that is built on a slower, but more energy-efficient kernel cluster? Isn't history telling us that a few bulls are better than a bunch of chicks?

History has also shown that memory on each server is also important, and many high-performance computer (HPC) and virtualization applications are subject to the number of random memory (RAM). Today's cortex-a9-generation ARM processor uses a 32-bit address bus and cannot support more than 4GB of memory.

However, interest in arm based servers is growing, and it is no longer on the propaganda level. Yes, the arm-based processor still lacks processing power, and the Xeon processor's memory controller can handle a large number of dual-column direct-memory modules (DIMMs), but when it comes to cost and power, the ARM processor scores a lot higher.

The ARM based processor has also made great strides in performance. The test data showed that a 1.2GHz dual-core arm cortex-a9 (Samsung Exynos 1.2GHz), released in 2011, was 10 times times faster than the general-purpose arm 11 compression released in 2008. According to Anand's iphone test, Sunspider's performance has also increased by 20 times-fold (as soon as this boost is almost certainly partially benefiting from browser and software optimizations). The latest arm CORTEX-A15 is undoubtedly more powerful, providing about 50% performance improvements. A57 will increase support for 64-bit and 20–30% performance improvements. In short, single-threaded performance is being quickly upgraded and can be connected to a larger amount of RAM. ARM cortex-a9 is subject to 4GB, but CORTEX-A15 can handle 16GB, and A57 can handle much more.

It seems that arm products in the server segment of the market to start a large number of shipments is only a matter of time, then how long will it take? The best way to look at this is to explore the most sophisticated arm processors that have been shipped commercially today: The Boston viridis server based on Calxeda. See what this server can handle now? What potential can it play? What are the disadvantages? Let's find out together.

1234 Next Read full-text paging navigation 1. Preface 2. This is a cluster of 3. More in-depth observation of server nodes 4. Exchange Matrix (Author: Wang: Wang)
Related Article

Contact Us

The content source of this page is from Internet, which doesn't represent Alibaba Cloud's opinion; products and services mentioned on that page don't have any relationship with Alibaba Cloud. If the content of the page makes you feel confusing, please write us an email, we will handle the problem within 5 days after receiving your email.

If you find any instances of plagiarism from the community, please send an email to: info-contact@alibabacloud.com and provide relevant evidence. A staff member will contact you within 5 working days.

A Free Trial That Lets You Build Big!

Start building with 50+ products and up to 12 months usage for Elastic Compute Service

  • Sales Support

    1 on 1 presale consultation

  • After-Sales Support

    24/7 Technical Support 6 Free Tickets per Quarter Faster Response

  • Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.