Mythics Blog

SPARC is NOT DEAD: Innovation through Integration, Efficiency and Ease of Management

Posted on December 7, 2015 by Erik Benner

Tags: SPARC, Fujitsu, m10

Rebroadcast from the Oracle Systems Technology Enablement for Partners (STEP) Community Blog, by Torrey Martin (Dec 3, 2015)

Blog: https://blogs.oracle.com/step/entry/sparc_is_not_dead_an

Guest Blogger:
Torrey Martin
Fujitsu M10 Product Specialist
Fujitsu-Oracle Center of Excellence  

I had the opportunity to speak with Erik Benner at Oracle OpenWorld 2015 in San Francisco about current enterprise computing and data center trends.  Erik is an Enterprise Architect and Oracle ACE with Mythics, an Oracle and Fujitsu partner and system integrator. He is also highly experienced, having worked with various system environments including Solaris for nearly two decades.

Erik engages with public and private sector customers who have Big Data, data center consolidation or enterprise cloud transformation projects.

What are the big trends in enterprise computing today?

In the public sector, Mythics is seeing consolidation of older technology onto new platforms. There are customers who have been on SPARC platforms and want to remain on SPARC. They have modest workload growth, their hardware is aging, and they want a newer platform.

With commercial customers, we are seeing migration toward SPARC from other vendors. This is due to the licensing advantage of SPARC versus other vendors. License stacking allows them to reduce the number of software licenses required. They also realize significant gains in computational performance and can consolidate databases for greater efficiency.

What do you find are the most demanding customer challenges?

One challenge we have with customers is in dealing with the “one size fits all” mindset. Customers are often under the assumption that Intel solutions are always better. But, this assumption is based on technology shifts from more than 10 years ago.

There are alternatives. A common myth is that “SPARC is dead,” but in reality there is more innovation happening on SPARC than people realize. Because Oracle and Fujitsu are the Executive Level Members of SPARC International, Inc., they can do deeper integration, all the way up the stack to the database tier.

The myth “Solaris is very difficult to administer” might have been true at the end of the last century, but much progress has been made, especially with Solaris 11. Now, it is very simple to apply and revert patches. You can’t do this on Linux.

Another myth is that SPARC is expensive. But one must look at the workload it can handle. A Fujitsu M10-4 SPARC server can replace large numbers of Intel blades, perhaps as many as 30 blades. If you look at costs, the Fujitsu M10-4 server wins hands down on capital costs. And then when you add on the OS, the hypervisor and support software, and build the entire system, the operational costs are much less. It is around 1/5 the cost to run SPARC in the datacenter.

You have worked with the Fujitsu M10 business server. What makes the M10 stand out as a customer solution?

Why M10?” customers ask. It’s simple. The Fujitsu M10 business server runs workloads efficiently; buy the right tool for the job. Buy SPARC to run a database, for example. It is definitely where Java runs best - on Solaris. Run file and print on Windows and Intel. So the message is use the right tool.

Suppose you require in-memory computing – why not just run on Intel? As I mentioned before, it all depends on the application. When you run on SPARC you have significant advantages:

  1. There is increased scalability for memory and processing. Intel x86-based systems don’t scale to over 30 Terabytes of shared memory or to 1000 cores like the M10-4S servers.
  2. Performance when operating within memory - it’s running in silicon, which means all calculation happens at the speed of silicon. So now it goes from really, really fast to warp-speed fast. Why not go with the fastest solution? Customers can implement a solution with SPARC at a lower price than with Intel because of hardware acceleration features such as decompression, SIMD and encryption.

We had a customer that evaluated different configurations. With commodity hardware, the solution was expensive, at around $3 million. But with a configuration using Fujitsu M10-4 servers, they could run the same workload faster. And by using license stacking and consolidation with logical domains, the configuration cost shrunk to $1.5-1.6 million. With Fujitsu M10 servers, the customer was using the right technologies.

Anything else you would like to add?

Yes, a big takeaway from my part in the Fujitsu Executive Solution Session is that hardware matters. Pick your hardware to run your specific workload. Don’t buy one size fits all. One size does not fit all! While it may work, it may not be optimal. Making the right hardware choice really matters.

Erik Benner, Enterprise Architect, Mythics Inc.

Comments

  • ! No comments yet

Leave a Comment