IT Innovation

Destination: SPARC

Organizations are following Oracle’s SPARC roadmap to improved performance.

By David A. Kelly

March/April 2012

In 2010 Oracle committed to delivering new SPARC chips and server hardware every 12 to 18 months. Over a five-year period, this SPARC roadmap called for improvement of 4 times the cores, 32 times the threads, 16 times the memory capacity, 40 times the database transactions, and 10 times the Java operations per second. So far, on-the-ground results exceed those commitments.

Instead of what was promised on the SPARC and Oracle Solaris roadmap, Oracle is delivering more. The driving time to major SPARC milestones has gotten shorter, and with the recent release of the SPARC T4 chip, Oracle is providing better-than-promised performance.

“With the SPARC T4 chip, we’ve actually delivered a faster chip in terms of raw gigahertz than we promised when we initially announced the roadmap,” says Chris Armes, vice president of engineering at Oracle. “And that’s the type of improvement we’ll continue to deliver against the SPARC and Oracle Solaris roadmap through 2015.”
Unisource Energy Corporation (UNS)
Location: Tucson, Arizona
Industry: Utilities
Employees: 2,000
Oracle Products: Oracle E-Business Suite, Oracle Database, PeopleSoft applications, Oracle Utilities Customer Care and Billing solutions, Oracle Fusion Middleware, Oracle Hyperion solutions, Oracle WebLogic Server, Oracle Solaris 10, SPARC Enterprise M5000 and Sun SPARC Enterprise T5120 servers, Sun ZFS Storage Appliance

Integrated Platform from Bottom to Top

Advancements in core chip technologies deliver the best results if the operating system, infrastructure, and applications running on top of the chips can leverage those chip advances. To that end, Oracle has invested heavily in integrating hardware and software; in many respects, that integration starts with Oracle Solaris.

In November 2011, Oracle announced the Oracle Solaris 11 operating system, with its new-and-improved integration with the rest of the Oracle technology stack.

“We optimized Oracle Solaris to run on SPARC, and we optimized Oracle Applications and Oracle Database to run on Oracle Solaris,” says Armes. “We’ve introduced a whole set of new technologies into Oracle Solaris 11 that the rest of the Oracle stack can leverage directly.”

“Future Oracle Solaris updates in 2013, 2014, and 2015 will track with SPARC innovation,” says Armes. “They’ll provide higher availability, increased memory, improved virtualization, enhanced system management, greater I/O capacity, and improved scalability.”

That’s good news for companies such as Tucson Electric Power Company (TEP), in Tucson, Arizona, that are counting on the future of SPARC and Oracle Solaris to run highly available business applications in virtual environments.

The First Cloud OS

Oracle Solaris 11 is not only keeping pace with advancing hardware capabilities; it’s also designed to deliver infrastructure as a service. The first cloud OS, Oracle Solaris 11 delivers built-in server, storage, and network virtualization to meet the security, performance, and scalability requirements of mission-critical deployments.

“The concept of the cloud is built into Oracle Solaris 11,” says Chris Armes, vice president of engineering at Oracle. “Virtualization, for example, is built into the networking functionality, so that the networking stack is completely virtualized. Virtualization was added at the design stage, not as an afterthought.”

For Oracle Solaris 11 users, the cloud features deliver the ability to support many users in a single, secure multitenant platform. “When you’re sharing the same platform, you need an operating system that has security at its heart, with the ability to scale and provide a virtualized infrastructure for cloud environments,” says Armes. “That’s what Oracle Solaris 11 does.”

Power Zones

When you’re in the business of delivering a critical utility, downtime isn’t an option. That’s why TEP decided long ago that SPARC-based servers and Oracle Solaris should power its business.

TEP has been using SPARC-based systems since 1998, when the organization first decided to deploy Oracle E-Business Suite. The organization chose a SPARC-based solution to run Oracle E-Business Suite because of Oracle’s strong support for SPARC and Solaris.

A subsidiary of UniSource Energy Corporation, TEP serves more than 400,000 electric customers in southern Arizona. The organization has nearly 1,400 employees and focuses on providing power generated from clean, sustainable sources.

Providing power to that many customers requires a range of mission-critical systems that need to be always available, including its customer care and billing system and Oracle E-Business Suite, which the company uses for financials, project planning, accounts payable, accounts receivable, and more. In addition, TEP relies on Oracle’s PeopleSoft applications for its HR systems and has many Oracle databases for custom applications.

One of the driving factors for TEP in designing its server architecture was the fact that the company has limited space in its data center. So, instead of each application or system having its own server, the organization used Oracle Solaris Zones to consolidate several existing SPARC-based servers onto Oracle’s SPARC Enterprise M5000 servers. First introduced in Oracle Solaris 10, Oracle Solaris Zones are built-in, lightweight virtual machines that securely isolate workloads on the same system.

“We went from using more than 20 servers and a bunch of different operating systems to having everything running on Oracle Solaris 10 inside Oracle Solaris Zones,” says Scott Myers, senior systems administrator, IT Operations Group, at TEP.

Myers likes the zone approach because it doesn’t come with a performance hit. In fact, Myers notes that even if an application is going to take up the full capacity of a SPARC Enterprise M5000 server, he still puts it into Oracle Solaris Zones to make it portable. It makes upgrades easier, provides greater flexibility, and reduces potential downtime.

“We create Oracle Solaris Zones that keep systems isolated,” says Myers. “Oracle Solaris Zones allow us to consolidate more operations on a single server, cut costs, simplify maintenance, save space, and reduce power consumption.”

With Oracle Solaris Zones, TEP is protected against risks such as system outages. “If we have a system outage or need to upgrade the hardware, we can migrate a zone to another host with just a couple minutes of downtime,” says Myers. “It provides us with added flexibility and greater portability.”

Management is also simpler. “Consolidating our systems has really cut down on the management overhead compared to having everything on a separate physical server, because we can do global zone patches with Oracle Solaris,” says Myers. “When we patch a server, the patch rolls up to everything on the server, so we have fewer systems to patch and manage. It’s definitely easier.”

Oracle Solaris Zones capabilities have also enabled Myers and his IT colleagues to address new business requirements faster and more efficiently.

“We keep excess capacity in all servers, so if they do have to house more zones or we need to bring up new applications or move applications, we have the capacity to do it,” says Myers. “It has tremendously increased our flexibility in being able to bring applications to market.”

Following SPARC

An important part of TEP’s commitment to Oracle Solaris and SPARC-based servers has been Oracle’s own commitment to SPARC, including the future SPARC roadmap and continued upgrade path. TEP is currently beta-testing the new SPARC T4-based servers and is impressed with the results.

“The SPARC T4 is really an incredible chip in performance, expandability, and capabilities,” says Myers. “We’ve been testing it with Oracle E-Business Suite, running a nightly batch process. So far, our T4-based test system, a SPARC T4-2 server, is outperforming our [SPARC Enterprise] M5000 nicely. We’re pretty impressed with it.”

Another advantage that Myers sees in the SPARC T4 chip is its improved single-thread performance capabilities.

“Our T3-based systems worked great for Web servers, but the T4 chip has improved single-threaded performance so we’re running databases on T4-based systems, and they’re performing every bit as well, if not better, than our current M5000 systems,” says Myers.

Eventually, Myers sees T4-based systems taking over the solutions running on the company’s SPARC Enterprise M5000 servers. “It’s a more energy-efficient chip, and it can handle more processes,” says Myers. “It’s a revolutionary design with on-chip cryptographic acceleration and built-in 10 gigabit Ethernet, which improves performance, throughput, and I/O.”

The company is also looking forward to Oracle Solaris 11, with all its new features. “There’s so much good stuff baked into Oracle Solaris 11,” says Myers. “We’re really excited about moving to it.”

“One thing that we’ve really appreciated about SPARC and Oracle Solaris is that over the course of all the upgrades and server switchovers we’ve done since 1998, we’ve only needed about an hour of planned downtime for each upgrade,” says Myers. “We’ve never had an issue with the application binary compatibility, thanks to the Oracle Solaris Guarantee Program. We’ve never had a single issue migrating from Solaris 2.6 through Oracle Solaris 10 with different SPARC chipsets. Everything has just gone flawlessly.”

“Oracle’s binary compatibility guarantee is a huge money saver for us or anyone that has to upgrade applications or systems,” Myers adds. “It’s just huge to be able to upgrade without worrying about your application, so we can do upgrades in about an hour of planned downtime.”

According to Myers, Oracle Solaris 11 streamlines the patching process as well as adding fast reboot. “Oracle Solaris 11 allows us to do a fast reboot for new patches, and even if there’s an issue it can fall back in a couple of minutes to the unpatched or previous version,” he says.

Another important factor for Myers is the reliability that his SPARC-based systems deliver. “We maintain five-nines reliability, and that’s without clustering,” says Myers. “We’re doing it with Oracle’s Sun hardware and Oracle Solaris and our applications.”

From TEP’s perspective, Oracle’s ability to deliver on its SPARC and Solaris roadmap is paying big dividends.

“I think Oracle’s doing very well staying on their SPARC roadmap,” says Myers. “In fact, I’d say [Oracle is] ahead of my expectations.”

SPARC Engineered System

Designed as a universal platform for data center consolidation, Oracle’s SPARC SuperCluster T4-4 is built on the new SPARC T4-4 server, which uses the new SPARC T4 processor. The SPARC SuperCluster T4-4 is a highly scalable system that can be used for the consolidation of multiple databases and applications, and for deploying multiple tiers of applications, database, and middleware.

“The SPARC SuperCluster is aimed at large, data-intensive and performance-intensive applications, spread across many clusters,” says Chris Armes, vice president of engineering at Oracle. “The SPARC SuperCluster makes it possible to integrate them all into one rack.”

The integrated components of the SPARC SuperCluster T4-4 include

  • Oracle’s SPARC T4-4 servers

  • Oracle Exadata Storage Servers

  • Oracle’s Sun ZFS Storage Appliances

  • QDR InfiniBand networking technology

  • the Oracle Solaris operating system

  • Oracle Solaris Cluster software

  • Oracle’s Exalogic Elastic Cloud Software

SPARC SuperCluster T4-4 combines capabilities of Oracle Exadata and Oracle Exalogic systems, and it supports both database and application/cloud/middleware workloads, so organizations can deploy an entire multitier application on one system.

The result is a consolidated increase in performance and a smaller footprint.

“The SPARC SuperCluster provides an opportunity for massive consolidation and massive increase in performance delivered and supported by one vendor,” says Armes. “You wheel it in, and it’s plug-and-play. Migrate your existing applications and data across, and you’re all set to go.”

Oracle on Oracle

A company that stands behind the software or hardware products it sells is one thing, but a company that runs its mission-critical enterprise resource planning (ERP) business on its own applications shows real commitment. According to Chris Armes, vice president of engineering at Oracle, Oracle’s business is committed to Oracle technology.

“We run our ERP business on a combination of SPARC, Oracle Solaris, Oracle databases, and Oracle Applications,” says Armes. “Oracle very much believes in running Oracle on Oracle’s technology.”

Nothing serves as a better example of this belief than what happened after Oracle closed its acquisition of Sun, when the organization needed to integrate Sun’s IT systems and almost 30,000 former Sun employees as quickly as possible into the main Oracle IT systems. Oracle needed to rapidly scale its data center to support the vastly increased demands for its core business applications, as well as integrate IT resources from the acquisition.

At the core of the challenge was finding a way to scale Oracle’s global single-instance database and the middle-tier infrastructure that supports Oracle’s financial and human resources applications as well as its management, supply chain, marketing, sales, and partner portals. The combined solutions need to have 24/7, 99.99 percent availability.

“We needed to take Oracle’s existing infrastructure and add 150 percent capacity to it to support the additional 30,000 employees,” says Armes. “And we didn’t have a lot of time.”

To meet the challenge, Oracle migrated its existing SPARC infrastructure from running on four Sun servers to running on two SPARC Enterprise M-Series servers running Oracle Solaris 10—all in two months.

The results were significant. Oracle was able to realize more than US$50,000 in yearly savings on facilities expenses alone, due to the reduction in space, power, and cooling costs. At the same time, overall performance increased two times for mission-critical processes, and disaster recovery time was cut by 95 percent.

“It was a huge migration, but we did it all in less than eight weeks with only 16 people,” says Armes. “And the reason we could do it in eight weeks is because the whole stack, from the CPUs up to the applications, is engineered to work together.”

Next Steps

READ more about
 the SPARC roadmap
 SPARC SuperCluster T4-4
 Oracle Solaris 11
 Oracle Solaris Guarantee Program
 Oracle Solaris Zones

 DOWNLOAD Oracle Solaris 11


Photography by Clay Banks, Unsplash