ARM Community: ARM in Servers: How Small Could This Be? - ARM Community

My focus for these blogs is the role ARM’s technology has in large, complex server systems found in today’s tier-1 data centers. As I have mentioned previously, we see the compute requirements of applications such as web hosting, memcached and Hadoop as being particularly well suited to the types of ARM Powered® platforms that are in development. A reader of my last blog entitled “Servers: How Big Could This Be,” challenged me look at the other end and provide a picture of what a really small ARM Powered® server could look like. So to start this blog, I thought I would share some words on the opportunity at that end of the spectrum.

Just under two years ago, Marvell launched an initiative called SheevaPlug. Plugging into a wall socket and providing Ethernet and USB connectivity (subsequent versions would add Wifi support), this was a $99 Linux server. Of course, one of the great things about the ARM ecosystem is choice. Texas Instruments (TI) also offers low cost development “Panda” boards based on Cortex™-A9 technology. It is perfectly possible to construct low cost platforms for devices from other ARM licensees, such as Freescale and NVIDIA.

ARM and other partners have used boards such as these to do native software development. And some platforms like these have made it into modest volume end applications. However, my personal opinion is that the high volume will be found as these devices become integrated into other conventional platforms as opposed to being utilized in a separate standalone appliances. At these power levels, form factor and cost points, devices like these can get integrated into DTVs, digital multimedia adapters/gateways and even some communications infrastructure platforms.

Much has been made about the growth and usage models of smartphones in emerging countries. The well worn story is a fisherman off the coast of Kenya using phone technology to understand which sea port to sail into in order to get the best price for his/her catch. Indeed it extends to trading using the phone since many people do not have a bank account. To support these types of services, there needs to be infrastructure equipment at the other end of the connections that can deal with operating in environments with limited and often unreliable electricity supplies. I see opportunity here, as do several of our partners!

This leads me to the topic of security. At ARM TechCon 2011 about four weeks ago, I was in a meeting with a potential silicon partner. We were discussing security in data centers. He likened data centers as to how polar bears view igloos….crunchy on the outside and chewy on the inside….with the “chewy” being the Eskimo of course! The point is that server security has been focused on ensuring the barriers to viruses, which is similar to the security for the physical entry to a building. But this is just one level of security. Data centers include information and programs from many different companies, so the security theme has quickly turned to ensuring that there is adequate protection/isolation between processor blades, and in some cases nodes. ARM announced earlier in 2011 that it had entered into a relationship with Giesecke und Devrient. The solution uses the combination of TrustZone hardware technology, something that has been around in the ARM architecture since the ARM11 days (I found an interesting write-up from many years ago on this), and software running (and connected to on-chip peripherals) from a secure area of the device to lower the risk of fraud on payments made on phones. ARM is working to define how this technology can be best applied at the other end of the “wire” and which ISV partners we should engage with in this arena.

I mentioned in a previous blog that I had attended Oracle’s JavaOne conference back in October. At that time, one of their keynotes showed a technology demonstrator of their performance-optimized Just-in-time (JIT compiler) running on ARM. Java support has existed for the ARM architecture for a long time. Since the standard deployment of a Java application is on a client device (phone, tablet, etc.), the Java compiler (Oracle refers to it as “C1”) ensures that the start-up time of the application when initiated is short. Therefore, the compiler has a limited amount of time to optimize the application for performance. If an application can take longer to start, the compiler has more time to increase performance. Oracle reported seeing about 20% -40% performance improvement (it is quite application dependent) using the performance-optimizing (C2) compiler when compared to C1. The C2 compiler is at the heart of their Enterprise Edition Java product.

I had anticipated that this technology would be released during 2012. However, Oracle released it earlier this week (and can be downloaded from here). This is particularly important for ARM processor based servers, since many of the compelling applications (Hadoop as one example) include a significant Java component. These optimizations will further extend ARM’s advantage from a performance/watt perspective when compared to alternative approaches.

This is the time of year when many of us reflect on the year gone by, and look forward to the coming twelve months. As the person leading the server initiative at ARM, I am particularly proud of the milestones that the ARM ecosystem has achieved during the past twelve months. The first commercial grade server Linux has been released, a tier-1 OEM has announced a development platform based on Calxeda’s device and Applied Micro has announced the first ARMv8 device to pursue the server market.

I look forward to incredible things in 2012.

Ian Ferguson, VP Segment Marketing, ARM, has spent years fighting from the corner of the underdog. Most of those scars are healing nicely. Ian is particularly passionate about taking ARM technology into new types of applications that do not exist or are at the very formative stages. After driving ARM’s server program for five years, Ian now leads ARM’s vertical marketing organizations, that supports ARM’s partners to grow their business across a wide range of applications with a view to reinvent the way the server function is implemented in networks as opposed to simply replacing incumbent platforms.