Oracle Open World Recap Part IV.5.1

Exalogic Larry Ellison continued his key note from Sunday by reintroducing Exalogic a “high performance server with hardware and middleware specifically designed for running public or private cloud systems. “ We spent a lot of time optimizing Oracle software to run on the Exalogic box,” Larry said. He referred to Exalogic as “one big honkin’ cloud,” and called the system “the fastest computer for running Java applications software” and said “it could be used for application consolidation or for running both public and private cloud systems.” The internal components of the system include:
The Oracle Exalogic Elastic Cloud server combines 64-bit x86 hardware, a total of 30 compute servers with 360 cores, with Oracle middleware such as the Weblogic server, Oracle Coherence data grid software, JRockit Java runtime software and Oracle VM virtualization software. The system uses infiniBand technology (capable of handling 40 gigabits per second) to link its internal components, has 2.8 TB of DRAM, 4TB of read cache and 960 GB of solid-state disk storage. Oracle will offer Linux and the Solaris operating systems with Exalogic
. What I found interesting is that despite proclaiming Exalogic as the solution for cloud computing, Larry took exception to the term-“Cloud!” Larry believes that the term ‘cloud computing’ is a widely abused phrase that’s ill-defined and applied to a broad range of information technology. Larry said that “people use the term ‘cloud computing’ to mean many different things and I find it confusing—I think a lot of people find it confusing. Is it rebranding, rebirth or true innovation? The term is sometimes applied to specific applications and other times to platforms.” Larry then went on to define what Oracle means by cloud computing:  “it’s a platform that means hardware and software, that’s right, a box in software. It’s a computer. It really is a computer. A cloud by the way folks is a lot of computers on the network. It’s a lot of boxes and a lot of software.  Now we think the software has to have certain characteristics. We think it has to be virtualized. It has to be elastic. Our view and Amazon’s view [of the cloud] are pretty much the same. So we kind of took what Amazon had, added an Oracle flavor to it, and put the hardware and the software so our customers can build private [or public] clouds out of this box…..And what we’re doing is we think the box  should be efficient. It should be fast. It should be reliable. And if you engineer the hardware and the software to work together in the box, you’re going to get a much better experience.” As a result, by integrating the hardware & software to work together Oracle saves customers time and money. The presentation continued to build on this theme of flexibility. Exalogic offers customers the option of building private or public clouds. Exalogic consists of 30 compute servers, 360 cores, all interconnected via InfiniBand with an integrated software appliance that stores the software and application files. “The interesting thing about this is [that] when you patch, you patch the microcode, the virtual machine, the OS, your applications, you download one file to that storage appliance and you patch everything together. As result, Exalogic is easy to install, it’s easier to patch, it’s safer to patch and it’s much more reliable.” Again it’s an example of Oracle engineering software and hardware to work together. They do the integration and deliver customers a complete box saving customers money and delivering customers much better performance than they could get otherwise. For example, Exalogic takes advantage of Weblogic server for memory coherence. There are 30 separate servers on the machine and the coherence software creates the impression that 1 large cache of memory exists across all 30 servers. In addition, to maintaining memory across the servers, Weblogic also supports JRockit and HotSpot JVMs, Oracle VM, Linux and Solaris. The Exalogic server is faster than using conventional machines and has on demand elastic capacity which makes it a perfect platform for cloud computing. Elastic capacity means that capacity expands with demand, so as resources are consumed, “you simply add more virtual machines to handle that additional capacity.” Exalogic does this by design; whenever applications/reports etc., demand additional resources more virtual machines are added automatically. In regards to speed Exalogic is the cloud computing platform of choice for Extreme Performance.
  • 12 xs improvement in Internet Applications running Java
  • Over 1 million HTTP requests/per sec
o   All of Face books web traffic on 2 full racks o   All of China’s rail ticketing  system on 1 rack
  • Over 1.8 million messages/sec
o   4.5x improvement over conventional hardware and conventional software “So this machine delivers enormous performance gains and cost savings because you need 20% as much hardware and software to do the same job. “ Exalogic provides huge cost savings, huge performance improvement and consumes less energy than traditional platforms. Exalogic is much faster than traditional machines, fault tolerant, scalable, and secure.  Oracle uses virtualization to store customer data, unlike other cloud architectures, like Salesforce.com, which use multi-tenancy.  Oracle argues that multi-tenancy is “a terrible idea” because every customer has all of their data in the same database, one database.  Oracle suggests that multi-tenancy is a 15 year old idea with dated technology.  Oracle espouses that the correct way to do this is with virtualization. “Virtualization gives you the ability to support lots and lots of customers in a secure fashion. Your data is separated on your own virtual machine. There’s fault isolation. There’s data protection.” And there’s scalability and security. Customers can run any application on Exalogic. It is the “right platform for delivering the benefits of hardware and software engineered to work together.” The system is fully fault tolerant, there is no single point of failure in either the Exadata or Exalogic boxes; it is also fault tolerant in the software and hardware. According to Larry “they just keep running if there are failures. They tolerate failures, not just disk failures but CPU failures, memory failures, networking failures, complete server failures, if power supplies go out, this [platform] keeps running. There are no single points of failure.” As a result, “it is an absolutely fault tolerant system built into the box.” The system starts at a ¼ rack and builds up to a full rack of Exalogic boxes and it scales up to 8 full racks. Larry then mentioned an interesting concept that many customers may find distasteful—All the Exadata and Exalogic boxes are identical. According to Larry “[m]ost of [Oracle’s] customers have unique hardware/software configurations [and] no one’s tested that configuration other than you. They’re all different. And we‘ve never tested that exact configuration. That’s a problem.” As a result, software vendors cannot guarantee that your exact configuration is going to work because the pieces have never worked together in that exact way—your way—and the components have never been tested together until you test them. Under this new approach Oracle will “pick a single configuration and test it for millions of hours and make sure it works before you get it. That's the idea of hardware and software engineered to work together. [Oracle] will do all of this testing, all of this optimization, all this performance tuning before you get it. The integration is done. [Oracle] knows the pieces work together and work together well. [Oracle] then delivers that same exact configuration to hundreds and thousands of [their] customers. Your configuration is not unique anymore. Your configuration is identical to thousands of other configurations all over the world.” All Exalogic Machines have these standard features and configurations:
  • All Exalogic Machines are built and configured the same
o   Delivered tested, configured and ready-to-run
  • All Exalogic Patches and Upgrades are Tested Together
o   Firmware, VM, OS, Java VM, Middleware
  • All Exalogic Machines “Phone Home”
o   Remote telemetry alerts Oracle of problems
  • All Exalogic Machines have unified monitoring & management
o   From Applications to Disk As a result, of having a configuration that is already tuned and optimized for performance, when a bug is discovered the path can be developed for every single customer since everyone has the same exact configuration. “[Oracle] can then come up with a patch, thoroughly test the patch and patch everybody's configuration safely to minimize bug rediscovery” for every single Exalogic/Exadata customer all at the same time.” Larry continued by saying “so hardware and software engineered to work together has huge benefits for our customers. When you install the system, it's thoroughly tested. When you patch the system, you patch everything together: firmware, VM, operating system, middleware, applications. All those changes are tested and we know they work together, millions of hours of testing rather than what you get now which is, on your configuration, no testing at all unless you do it yourself.” The Single Patch process for the Exalogic System includes:
  • All customers run the exact same configuration
  • All software components can be patched together
  • All  patches are built, packaged, and tested together
  • Enterprise manager automates patch and upgrade procedures
Finally Exalogic is much less expensive than the competition. It is a “much lower cost solution than a configuration delivering similar performance” from IBM (Power 795). IBM’s machines aren’t fault tolerant; Oracle is. IBM scales vertically up instead of horizontally out. IBM isn’t parallel; Oracle is. Oracle has “no single points of failure; [IBM is] loaded with single points of failure.” In conclusion, Exalogic is much faster and is 4xs lower cost than IBM’s Best – so with Exalogic you have to spend less to get more!

Exalogic

IBM

40% more CPU cores Old SMP Vertical scale-up system
Horizontal scale-out architecture No Fault Tolerance
InfiniBand Network Fabric
Fully Fault Tolerant