Data centers have rapidly embraced cloud computing in its various forms. And what’s not to like? Cloud computing is more efficient and leads to higher resource utilization. It is also the quickest way for users to get the compute resource they need, when they need it.
For many of the customers we talk with, a private cloud (with a hybrid option) is probably the most attractive way to go, especially for those customers with large amounts of data being shared between applications. In this model, the customer’s data center is set up as a private cloud, but there are also options to use public clouds for overflow compute power, or for projects that are of short duration, like application testing.
Powering your private cloud
We also believe, not surprisingly, that our Fujitsu SPARC M12-2S coupled with Solaris is the perfect foundation for private clouds. When you run many applications on each of your hardware platforms, the underlying servers need to be rock solid, provide great performance, and gracefully and inexpensively scale to handle even the largest aggregated workloads. We believe that the Fujitsu SPARC M12-2S delivers the goods when it comes to these requirements.
When it comes to core performance, the Fujitsu SPARC M12-2S outperforms its predecessor Fujitsu M10 system by a factor of 2.5x and provides exceptional performance, reliability and scalability when compared to x86 systems.
One important note about Fujitsu SPARC M12-2S server performance is how balanced it is. Most processors today tend to perform better on integer workloads vs. floating point operations. The SPARC64 XII processor that fuels the Fujitsu M12-2S server offers industry-leading performance on both integer and floating point operations, meaning that it can easily handle transactional-oriented integer workloads but also highly compute intensive demands from, for example, analytics and numerically complex applications.
Scaling to meet the demands of your business
Scalability is a strong point of the Fujitsu SPARC M12-2S server. It scales linearly from one to 32 CPUs (two to 384 cores) and can host up to 32 TB of memory – all within a single operating system instance, which makes it one of the largest systems available today.
For private clouds, scalability is very important because it dictates just how many systems you have to purchase and support to meet your needs. Having a scalable single image of the operating system is more effective from a utilization standpoint – it allows many applications to share the underlying hardware much more efficiently compared to the standard x86 virtualization model that requires each application to have its own instance of the operating system.
As important as top-end scalability is how you get there. The Fujitsu SPARC M12-2S has an innovative and unique approach to scalability that allows you to acquire just enough hardware to meet your needs today, but then add capacity in small, inexpensive increments to handle growth.
The system is modular, comprised of up to 16 single or dual processor system chassis. You can start with a single or dual processor chassis system, and add an additional chassis without interrupting current processing, as needed, as a chassis can be added without rebooting the whole system. With proper system configurations, an existing chassis can be removed from a multi-chassis physical partition (PPAR), upgraded from a single processor configuration to a dual processor version and then be reincorporated into the running system.
Flexible private cloud for now and into the future
At the individual server level, Fujitsu offers Core Activation, which allows you to purchase only the number of cores necessary today, and then add more in the future.
For example, you can purchase a Fujitsu M12-2S with only two cores activated. This saves money and matches up compute capacity of the new system with your current compute needs.
When more workloads are added to the private cloud, you can then dynamically add more cores to the system without interrupting any of the other workloads on the system. This is a key point, as other capacity on demand mechanisms require rebooting the entire system. This could be a major inconvenience when operating a private cloud that is hosting a large number of applications.
You only pay for the number of cores you’re using as you use them, which saves budget dollars while giving you the ability to grow your computing power very quickly and in a non-interruptive way.
Solaris 10 and 11 offer Solaris Zones, a mechanism that allows customers to run multiple workloads on a single Solaris operating system instance and, at the same time, ensure that each application gets the compute resource it needs, when it needs it – automatically, without operator intervention. Zones keep each application isolated from the others, making sure that one application can’t affect operations of the other applications on the system.
A Solaris Zone is set up for each of the applications on your private cloud. With Solaris Zones, CPU resources are pooled together and each zone is allocated a share of the pooled processors. Here’s a simple example: With Solaris Zones, Application #1 could be allocated 30% of the processor pool, while Application #2 might need 50%, and the remaining 20% would go to Application #3.
The system would automatically give the applications their rightful share of processing activity based on how you allocated the pool in your initial zone set up. But in this manual configuration, if Application #2 doesn’t use its full 50% share of the processing power, the processor won’t automatically allow another workload to use the idle capacity.
There is a solution to this problem, a solution that is much more efficient and still makes sure that the business goals of the organization are met. To use it, you simply enable the Zones Fair Share Scheduler. The scheduler will take into account all of the business goals of the configuration (App #2 needs 50% of the CPU capacity, App #1 needs 30%, etc.) and maintains the allocations. However, when one of the applications demands more processing power, and if the other zones aren’t using their full allotment, then the Fair Share Scheduler will give the requesting zone more CPU power as long as it is available.
There are many options when it comes to configuring zones. In addition to CPU capacity, physical and virtual memory can be managed and capped, for example. Zones can easily be set up on the command line or through the Solaris management GUI.
The Fujitsu SPARC M12-2S combined with the power of Solaris Zones is a great foundation for any private cloud. With the cloud usage model, hardware utilization will be much higher (particularly when using the Fair Share Scheduler), thus saving precious budget dollars. Customers can also add capacity only when it’s needed – eliminating the days of buying an over-configured system with ‘head room’ – which also saves significant money. A private cloud based on the Fujitsu SPARC M12-2S server might be just the ticket for your data center. Please visit our homepage for more information.