HP Launches Moonshot: Little Servers for Big Applications

HP's Moonshot offers high-density, low-power servers that are ideal for hyperscale data centers, but don't expect them to unseat virtualized servers and blades in the enterprise.

Howard Marks

April 9, 2013

3 Min Read
Network Computing logo

HP has announced the first in a series of Moonshot servers that promise to bring new levels of compute density and power efficiency to hyperscale data centers. Moonshot takes the blade server concept to the next level by hosting 45 low-power servers in a single chassis.

While much of the early hype around Moonshot talked about ARM-based servers, the first Moonshot devices to come to market will be the Moonshot 1500 chassis and an Intel Atom S1260-based server cartridge that supports 8GB of memory, a SATA drive, and dual 1Gbps Ethernet ports. As with some high-density disk systems, server cartridges load into the 4.3u chassis from the top, which boosts density.

Future server cartridges are expected with AMD and ARM processors from Calxeda and Applied Micro, with additional special-purpose cartridges with digital signal processing from Texas Instruments or even custom FPGAs down the road. Even more than in today's blade chassis, management functions are handled at the chassis, rather than blade, level. As an example, Moonshot servers don't have conventional VGA chips or interfaces.

In many ways Moonshot looks a lot like a commercial version of the "Group Hug" board that was introduced at the Open Compute Summit. Group Hug provides a common backplane using x8 PCIe slots to link up to 24 system on chip (SoC)-based daughter cards. The big advantage of the SoC based systems is that while each server delivers 5-10% of the horsepower a Xeon or Opteron server, they do it at just 1-2% of the power.

These servers are a natural fit for hyperscale applications such as Web 2.0 businesses, video on demand and high-performance computing applications like weather prediction or Monte Carlo simulations. The key thing these applications have in common is that they have a high degree of parallelism and are already scaled out across tens or hundreds of Xeons. If they can be scaled out across thousands of little servers, the same task can be run at much lower power. In addition, when applications are built out that massively, the failure of a single node is much less significant than when a smaller number of larger servers are used.

[ Join us at Interop Las Vegas for access to 125+ IT sessions and 300+ exhibiting companies. Register today! ]

Some of my analyst friends call the little server model the future of computing, and predict Moonshot will put a hurt on more conventional blade servers. In the long run, we may all replace our SQL servers and Oracle RACs with massive scale-out applications, but the corporate data center isn't going that way for a long time. I can run 40 or more virtual machines, each as big as a Moonshot's Atom processor and 8GB of memory, on a single blade today. If I can virtualize it, why buy little servers to run the same workload? Didn't I get VMware to consolidate in the first place?

So what if we use the little servers as virtualization hosts? That doesn't make any sense to me. Each Moonshot server has just a 2Ghz Atom processor and 8GB of memory. That would mean that each server could support just 2 to 4 small virtual servers before running out of memory. Add in that most hypervisors and their associated management tools are priced by the processor socket, and the economics of virtualization on little servers just don't make sense.

Little servers, including Moonshot, are exciting for the fastest growing parts of the server market: hyperscale data centers and those cloud environments where they can shift from provisioning VMs to provisioning little servers. This would save the cost, in dollars and CPU cycles, of the hypervisor. Remember, however that while these are the fastest growing parts of the server market, they're not the biggest. That's still the corporate data center, where virtualization makes more sense.

About the Author(s)

Howard Marks

Network Computing Blogger

Howard Marks</strong>&nbsp;is founder and chief scientist at Deepstorage LLC, a storage consultancy and independent test lab based in Santa Fe, N.M. and concentrating on storage and data center networking. In more than 25 years of consulting, Marks has designed and implemented storage systems, networks, management systems and Internet strategies at organizations including American Express, J.P. Morgan, Borden Foods, U.S. Tobacco, BBDO Worldwide, Foxwoods Resort Casino and the State University of New York at Purchase. The testing at DeepStorage Labs is informed by that real world experience.</p><p>He has been a frequent contributor to <em>Network Computing</em>&nbsp;and&nbsp;<em>InformationWeek</em>&nbsp;since 1999 and a speaker at industry conferences including Comnet, PC Expo, Interop and Microsoft's TechEd since 1990. He is the author of&nbsp;<em>Networking Windows</em>&nbsp;and co-author of&nbsp;<em>Windows NT Unleashed</em>&nbsp;(Sams).</p><p>He is co-host, with Ray Lucchesi of the monthly Greybeards on Storage podcast where the voices of experience discuss the latest issues in the storage world with industry leaders.&nbsp; You can find the podcast at: http://www.deepstorage.net/NEW/GBoS

Stay informed! Sign up to get expert advice and insight delivered direct to your inbox

You May Also Like

More Insights