ProphetStor Federator®SDS 重新定义SDS 2.0
In my previous article, I have introduced Software-Defined Storage 2.0. ProphetStor is leading the evolution of SDS 2.0, redefining and reinventing storage for the cloud and next generation virtualization era.
The abstraction and virtualization of the underlying, and often fragmented storage technology platforms, is the foundation of SDS. However, cloud computing and next generation virtualization require more than just abstraction and storage capacity pooling. They demand flexibility in both capacity and performance provisioning. They demand a rich set of data management features such as snapshots, cloning, replication, migration that are across the entire storage landscape, regardless of platforms, proprietary elements or brands. They demand secure automation and orchestration to match the similar elasticity, multi-tenancy and other tenets of cloud computing and next generation virtualization.
Today, most vendors claim that the separation of the data plane from the control plane as SDS version 2.0. SDS 2.0, in other vendors’ terminology just stops there. Most do not prescribe any form of Quality of Service (QOS) or Service Level Objectives (SLO) required by the present and future cloud computing platforms. These vendors rarely classify the capabilities of the storage platforms it has virtualized, setting the simple precedence that “all virtualized storage are the same”.
ProphetStor recognizes that not all virtualized storage platforms are the same. Not all disk media in the underlying storage platforms are the same. Disks can be 7.2K, 10K, 15K. Solid-state devices can be MLC Flash, SLC Flash. Storage platforms can be of RAID-10, RAID-5, RAID-6 or even JBODs. One hybrid storage array will be able to deliver high IOPS, lower latency compared to an all-SATA storage array. How does a so-called SDS 2.0 vendor discover and understand each and every storage platforms capabilities when it does not have such important features in its playbook? ProphetStor has that ability to do that.
I have flashed this architecture diagram in my previous article. This is the foundation of ProphetStor Federator SDS.
The green rectangle in the diagram above is the Control Plane of Federator SDS. In this Control Plane layer, ProphetStor Federator SDS initiates and communicates with disparate storage platforms with 2 open industry standard protocols developed by SNIA (Storage Networking Industry Association). These 2 protocols are SMI-S (Storage Management Initiative – Specification) and CDMI (Cloud Data Management Interface) and they assist ProphetStor Federator SDS to discover the underlying storage resources, but more importantly, each storage platforms’ capabilities. Both standard protocols are widely adopted by many storage vendors including EMC and NetApp.
For non-SMI-S, non-CDMI or proprietary storage platforms, ProphetStor takes a smart and efficient approach to discover and incorporate these “more difficult” platforms. Since OpenStack has been taking the cloud computing world by storm, almost every storage vendor has a Cinder block driver to provide block storage as OpenStack volumes. Rather than developing unique drivers for each storage platform (extremely resource intensive and time consuming), ProphetStor has developed a Cinder emulator to address this challenge. The Federator SDS Cinder emulator “tricks” storage platforms to believe that it is a Cinder adapter in OpenStack, and therefore allowing unique storage platforms to communicate with Federator SDS with minimal development efforts. You have to admit that this is pretty cool!
I mentioned that most storage virtualization vendors, aka SDS 1.0, have ignored the Intel x86 ODM (original designed manufacturers) storage platforms. ProphetStor recognizes this fast growing market segment, especially in the space of cloud service providers and hyper-scalers such as AWS and Google. ProphetStor has developed its own storage hypervisor, called the FlexVisor (shown as theProphetStor Storage Hypervisor in the Control Plane of the diagram above) to consolidate and aggregate Intel x86 ODM storage platforms. Instead of just a plain-vanilla storage hypervisor, FlexVisor incorporates deep enterprise features for the Intel commodity storage such as thin provisioning, rapid point-in-time copies, writable snapshots and rollback, storage efficiencies such as compression and deduplication, and many more.
One very interesting FlexVisor feature is Elastic Resource Control (ERC). FlexVisor extends solid state devices (SSDs) as a Read/Write cache. The patent pending Traffic Modeling Module (TTM) learns about applications storage IO patterns, and predicts the amount of SSD cache required for an application – dynamically.
Continuing on after the storage resources and capabilities have been discovered, ProphetStor Federator SDS redefines SDS 2.0 with its Data Plane. The Federator SDS Data Plane adopts a 5-step approach to provisioning secure and flexible storage resources to applications up the computing stack. This 5-point approach is listed here:
The browser-based management Federator SDS Dashboard shows the simplified 5-point approach in action.
The understanding and profiling of each supported storage capabilities through SMI-S or CDMI is crucial to the QOS and SLA requirements of the virtualized applications. The Federator SDS allows the authorized user to set a level of performance from provisioned storage resources. For example, one can set a guaranteed IOPS performance level to a particular workload of an application. This means the QOS to the application is maintained, ensuring proper storage performance responsiveness to an application workload.
The differentiation of the Control Plane and the Data Plane is the foundation of Software-Defined Storage. ProphetStor has taken a consistent, but more importantly an industry standards approach to the Control Plane. Using both SMI-S and CDMI, ProphetStor simplifies the discovery of disparate storage devices underneath, removing the complexity of proprietary approaches often used by other SDS vendors. This also ensures that customers are insured for the future, because these standards are widely adopted across the industry. But it is on the Data Plane where the magic happens. ProphetStor applies data management features that enrich the service offerings up north to the compute and application layer. The QOS feature is one.
The QOS feature ensures that the right amount of storage resources, notably performance resources, is delivered according to business applications requirements at the right time. The QOS reduces resource contention conflicts, removing over-consumption of performance resources by heavy processes, and maximizing the utilization of performance resources. This controlled QOS approach leads to greater efficiency and performance responsiveness to business applications.
The other unique feature on the Data Plane is ProphetStor’s Elastic Resource Control (ERC), a Read/Write Cache locked to the underlying solid state devices. Together with the Traffic Modeling Module (TMM), it “learns” the unique application I/O patterns, predicting and eventually fine-tuning the right amount of reading and writing cache resources to the application ProphetStor is serving. This is important and greatly benefits the very dynamic environments of IT-as-a-Service and Cloud Service Providers, where hundred and even thousands of applications and VMs are served per hour.
Concluding this article, ProphetStor Federator SDS is setting an entirely class of its own when it comes to Software-Defined Storage.