Developed by the elastic computing block storage team of Alibaba Cloud, Apsara Block Storage is a distributed block storage product tailored for leading enterprises in the market. It is an elastic and reliable cloud storage service platform that provides high performance and supports large-scale horizontal scale-out. Apsara Block Storage can be connected to various types of computing platforms, including Alibaba Cloud ECS, ZStack, OpenStack, Kubernetes, and physical servers, so it can satisfy customers’ demand for agile access to massive storage resources in the cloud era. It is applicable to cloud storage business scenarios for offline data centers of government agencies, financial institutions, operators, and large manufacturing enterprises. Apsara Block Storage aims to compete with distributed storage products of mainstream players in the traditional enterprise storage market, such as EMC and Huawei. Alibaba Cloud will officially release Apsara Block Storage in the near future, marking Alibaba Cloud’s foray into the global enterprise storage market. Alibaba Cloud will become one of the few service providers in the world that are capable of providing enterprise-oriented professional storage services for both public and private cloud markets.
Why Does Alibaba Cloud Provide Block Storage Services Independently?
Customer Pain Points and Industry Trends
With a wider and deeper application of cloud computing, big data, mobile Internet, and artificial intelligence (AI) in all walks of life, digital transformation has become an important driving force for customers’ business innovation and agile development. Traditional centralized and siloed IT system architecture can no longer meet the increasingly complex business development requirements, especially when it comes to flexible use of storage resource.
- Performance and capacity cannot grow in a synchronous and linear manner. The traditional centralized controller architecture has encountered increasing performance bottlenecks.
- Capacity expansion and configuration adjustment are complex and time-consuming, which cannot meet customers’ demand for fast business growth.
- Capacity planning does not match the actual business requirements. Over-investing has become rampant, and resource utilization remains low.
- Resources among multiple application systems are difficult to be pooled and streamlined, leading to poor data sharing and flow.
- The problem of data silos has become increasingly prominent. Offline remote backup and disaster recovery require a huge investment and long construction cycle.
In this case, a resource- and platform-based cloud storage architecture with flexible expansion capability has become the best choice for massive data storage in customers’ modern data centers.
On a global scale, distributed storage products have become the first choice for customers to build cloud virtualization platforms, because of their inherent advantages in architecture and compatibility with the hardware of commercial x86 servers. According to a global survey on distributed Server SAN products carried out by Wikibon a few years ago, the enterprise distributed Server SAN products and hyperscale Server SAN products are projected to see their compound annual growth rates (CAGRs) of above 44% and 22%, respectively, nibbling away at the market share of traditional centralized storage products, as shown in the following figure.
Enterprise Server SAN products have undoubtedly become “star” products in the global external storage market, with the annual market value reaching tens of billions of dollars in the next few years.
Connecting Multiple Types of Computing Resources
With the deep understanding and extensive application of cloud computing and big data, more and more customers have asked for better and more advanced services from our Apsara Stack products. The following are some examples:
- Unnamed government customer: Alibaba Cloud block storage products are very stable. We hope that they can be not only used for the ECS computing resources of the Apsara enterprise edition, but also directly connected to our existing physical servers to replace the traditional storage array products that will soon be out of warranty.
- Unnamed state-owned enterprise customer: We have set up a cloud computing R&D department with the capability to develop an OpenStack-based computing platform. We hope to introduce Alibaba Cloud block storage products to establish a win-win architecture that integrates our self-developed computing platform with Alibaba Cloud block storage.
- Unnamed Internet customer: We have already built several large data centers, and we have years of resource investment and accumulation in terms of OpenStack-based cloud computing. We hope to connect Alibaba Cloud block storage products to the OpenStack platform and physical machines, to gradually replace the existing storage resources such as storage arrays and local server disks.
The elastic computing block storage team of Alibaba Cloud is facing challenges such as how it can quickly meet the requirements of these customers for connecting to more types of computing resources and how it can satisfy enterprise customers’ agile and variable data storage requirements in the cloud environment. Undoubtedly, the product and technical teams must follow the industry trends and satisfy customers’ real needs.
Highlights of Apsara Block Storage
Developed based on Alibaba Cloud elastic computing block storage services, Apsara Block Storage is an all-new product developed for the enterprise market and a cloud solution for traditional IT infrastructure. Through communication with our customers and partners, we are fully aware of our next huge industry opportunity following the public cloud. That is, we need to think about how to deliver the block storage technology and service capabilities we have been developing on the public cloud for years to our customers in the form of independent services. Apsara Block Storage introduces the following features:
Software-defined storage is designed to simply and quickly meet the scale-out and scale-up needs of applications for on-demand and automated storage by abstracting and pooling storage resources in a data center. Provided as software services, such as user interfaces (UIs) or service application programming interfaces (APIs), software-defined storage will not increase users’ costs. At the same time, software-defined storage can also be customized. Through service APIs, storage resources can be orchestrated and seamlessly integrated with the existing IT system.
Resource Abstraction and Pooling
Apsara Block Storage abstracts a variety of entities from physical storage resources, such as volumes or logical unit numbers (LUNs), snapshots, images, and backups. It also maintains the mapping of these abstract entities. For example, a volume can have multiple snapshots, and each snapshot can be converted into an image. Another important abstraction lies in the unified and pool-based management of storage resources. By shielding the underlying physical distribution of storage resources and storage media types, Apsara Block Storage schedules storage resources distributed in various data centers and equipment rooms in a uniform manner to present an externally unified resource pool. It can also schedule applications on demand, including dynamic scheduling. For example, it can migrate a disk online from one data center to another.
Programmable Service APIs
Besides the abstracting and pooling of hardware storage resources, software-defined storage also features user-defined and elastic abstract resources. For example, Apsara Block Storage allows an application to specify the service level agreement (SLA) of data security when it creates a disk. Then, a configuration with high redundancy can be used in application scenarios with high-level data security. Moreover, these capabilities can be deployed anytime and anywhere. For example, Apsara Block Storage can expand the capacity of a disk online and even convert the type of a disk online. When visits to an application grow, the original serial advanced technology attachment (SATA) cloud disk cannot meet performance requirements. In this case, Apsara Block Storage can convert the SATA cloud disk into a solid state drive (SSD) cloud disk or an enhanced SSD (ESSD) cloud disk, without stopping the application. Apsara Block Storage orchestrates and defines APIs to ensure that these elastic events follow the orchestrated logic and automatically proceed for the application.
Distributed Server SAN
Traditional enterprise storage is based on traditional proprietary storage devices, and the storage networks are also based on, among others, the fiber channel (FC). With the advent of flash memory, 25 Gbit/s and 40 Gbit/s Ethernet networks, and remote direct memory access (RDMA) network technologies in recent years, traditional enterprise storage seems to lag behind hardware development and lacks technical advantages. At the same time, cloud computing and big data have achieved great success in public cloud scenarios. All cloud computing platforms can run on general x86 servers, whereby the reliability and performance of their architecture have been fully verified. Thanks to the development and extensive use of technologies such as x86 servers, high-speed networks, flash memory, and software-defined storage, Server SAN has been developed. Server SAN not only provides high-performance I/O capabilities, but also implements scale-out, error correction (EC), and online data deduplication based on distributed technologies, thereby further reducing users’ costs.
Data Backup and Disaster Recovery
Data security is the first priority for storage. Both real-time data protection and post-event data recovery are essential. Apsara Block Storage supports the full-link end-to-end CRC capability and protects users’ data through second-level snapshots, continuous data protection, remote data backup, and data recovery.
Flexible Deployment Architecture
Apsara Block Storage provides a flexible deployment architecture for different application scenarios. In converged deployment mode, all computing and storage software components are deployed together. In separate deployment mode, computing software components are deployed on computing nodes only, whereas storage software components are deployed on storage nodes only.
Hyper-converged deployment mode
In hyper-converged deployment mode, all ABS components are deployed on the same machine. In this manner, the same cluster can provide computing services, such as creating virtual machines (VMs), as well as storage services, because VM data is stored directly on the machine where the cluster resides.
Separate deployment mode
In separate deployment mode, computing software components and storage software components are deployed on computing nodes and storage nodes, respectively. The advantage of separate deployment is that the computing resources and storage resources are completely decoupled and do not affect each other, lowering maintenance costs. We recommend the separate deployment mode for clusters with a certain scale.
Apsara Block Storage is an open platform. We believe that Apsara Disk+ can be connected to all platforms where block storage is required. Based on software-defined capabilities, Apsara Block Storage supports mainstream cloud platforms such as OpenStack, ZStack, and Kubernetes, and delivers the technical capabilities that Alibaba Cloud block storage services have been developing on the public cloud for years to the traditional offline IT systems. In compliance with the Internet Small Computer Systems Interface (iSCSI) standard block interface protocol, Apsara Block Storage builds a channel between traditional IT systems and cloud systems, allowing traditional IT systems to be easily connected to the cloud without any reconstruction.
As shown in the following figure, Cinder is a storage subsystem in the OpenStack system. Alibaba Cloud has developed Aliyun EBS Plug-ins for Cinder to seamlessly integrate Alibaba Cloud EBS block storage with OpenStack. In this way, users only need to focus on the scheduling and control of computing, leaving Alibaba Cloud EBS to carry the block storage services and provide the upper layer with abundant functions, such as disks, snapshots, images, maintenance, and monitoring.