Software defined storage: compare 7 solutions
In this article, I will briefly describe the software-defined storage (Software-Defined Storage, SDS) and the possibilities of their use, which they provide when building an IT infrastructure. At the end of the article you will find a comparison of seven SDS-solutions. I tested them when my colleagues from Onlanta were working on options for the development of OnCloud.ru cloud infrastructure . I hope that the comparison table will save you a lot of time and effort when choosing a product.
A source
I work as a system engineer for the Onlanta cloud integration group. One of my areas of activity is research (R & D) to study and compare new technologies that could help us improve the quality and reduce the cost of OnCloud.ru cloud services provided by Onlanta. With the results of such a comparison of SDS-solutions you will meet in this article.
In large organizations, data storage systems account for a significant share of the cost of the IT infrastructure (according to expert estimates , up to 25%). This figure can grow significantly. The reasons are the growth of the data volume and the increase in the need for storage system capacities (DSS), including due to laws that oblige this data to be stored. At the same time, companies are actively trying to save IT budgets, which forces them to be constantly looking for the most profitable technological solutions that would reduce these costs without compromising the quality of service. The same applies to the storage and processing of data.
')
Customer requirements for reducing the cost of owning an IT infrastructure are forcing vendors to invest in development and offer new technologies. One of them is software-defined storage systems (Software-Defined Storage, SDS). Companies are starting to think about the implementation of SDS when the procedures for working with data become ineffective and their search takes a lot of time.
A source
The concept of SDS allows you to get such advantages as:
Thanks to SDS technology, you can significantly reduce the cost of storage and their administration. According to Gartner , by 2020, 70–80% of unstructured data will be stored on low-cost systems managed by SDS, and by 2019 70% of existing storage arrays will be available in a fully software version.
Storage management software should provide flexible data storage, as well as:
A source
SDS is defined in the Storage Networking Industry Association (SNIA, Storage Manufacturers and Consumers Association) as a virtualized storage environment with a service management interface , which should include:
I note that SDS needs a standardized management interface, such as the SNIA Storage Management Initiative Specification (SMI-S). It is an integral part of the concept of software-defined data centers (SDDC). This software logic of cloud storage infrastructure and cloud hardware platforms can be an element of traditional data centers. Data storage and processing services can run on servers, storage appliance, or both of these platforms, eliminating traditional boundaries.
Software-Defined Storage is offered by many vendors:
I’ll clarify that the RedHat Storage solution is represented by two products: RedHat Ceph Storage and RedHat Gluster Storage (RH Storage Server). Here they are both implied, but they did not participate in the comparison below, since they are significantly different from the other solutions mentioned.
Ceph is not really a boxed product. Its use without a staff of developers is rather difficult, which made it uninteresting for our company. Therefore, this solution is not in the comparative table.
Conventionally, all SDS solutions can be divided into three categories:
Some manufacturers offer both integrated solutions and software (Huawei, Dell EMC). This allows a flexible approach to the selection of products and the use of inherited "computing" equipment to solve less resource-intensive data storage tasks. Another merit of SDS has been the ability to use disk arrays virtualization in some classic storage systems.
Architectural solutions are built on two principles:
In the first case, fault tolerance is provided by distributed copies of data, but due to redundancy of communication between nodes (nodes), the write speed is reduced. The critical point is the data network, so these solutions are usually implemented on the basis of InfiniBand. Based on this principle, VMware vSAN, HPE StoreVirtual VSA, Dell EMC ScaleIO solutions are built.
In systems without common elements, data is written to one node, and then with a specified frequency is copied to the other to ensure fault tolerance. In this case, the records are not transactional. This approach is the cheapest. Ethernet is most often used as an interconnect. This architecture is convenient in terms of scalability. Its bright representative is CEPH.
Now many companies are developing both software SDS (for example, Atlantis Computing, Maxta, StarWind, DataCore Software, Sanbolic, Nexenta, CloudByte), and releasing complete solutions (Dell EMC, IBM) or specialized devices (Tintri, Nimble, Solidfire).
A source
Of the most well-known on the market, we selected for comparison seven solutions that are most interesting for the tasks of “Onlanta”. It:
In this table we compared their main characteristics.
Click to enlarge the table.
The SDS technology began to develop in the early 2000s, but has not yet been able to replace the classic storage systems for a number of reasons - we are not going to discuss them now. But manufacturers are actively engaged in the development of their products and interest in SDS technology is growing. According to our estimates, in the near future they will become the tool that will allow reducing the cost of IT infrastructure with an increase in the need for increasing storage capacity.
A source
In conclusion, I note that in this article I did not try to offer options for choosing the solution that suits you. Such a solution should be chosen based on the load, SLA, etc. The proposed table compares only the solutions capabilities, and does not compare performance, replication speed, node switching time, etc. This is a comparative analysis of opportunities, and not productive testing.
After a thorough acquaintance with the products of SDS, we came to the conclusion that in their current implementation, they do not fit our tasks very well. For ourselves, we nevertheless chose a classic solution, the implementation of which we are currently engaged in, and about what, perhaps, in the near future we will tell you.
But I hope that the presented comparison results will help you navigate, save time and make it easier for you to choose which solution is right for you.
If someone from the readers finds it possible to share any additional information on the subject under discussion, and perhaps to tell about his choice, it would be very interesting.
A source
I work as a system engineer for the Onlanta cloud integration group. One of my areas of activity is research (R & D) to study and compare new technologies that could help us improve the quality and reduce the cost of OnCloud.ru cloud services provided by Onlanta. With the results of such a comparison of SDS-solutions you will meet in this article.
The trend to reduce the cost of ownership of IT infrastructure
In large organizations, data storage systems account for a significant share of the cost of the IT infrastructure (according to expert estimates , up to 25%). This figure can grow significantly. The reasons are the growth of the data volume and the increase in the need for storage system capacities (DSS), including due to laws that oblige this data to be stored. At the same time, companies are actively trying to save IT budgets, which forces them to be constantly looking for the most profitable technological solutions that would reduce these costs without compromising the quality of service. The same applies to the storage and processing of data.
')
Customer requirements for reducing the cost of owning an IT infrastructure are forcing vendors to invest in development and offer new technologies. One of them is software-defined storage systems (Software-Defined Storage, SDS). Companies are starting to think about the implementation of SDS when the procedures for working with data become ineffective and their search takes a lot of time.
A source
The concept of SDS allows you to get such advantages as:
- abstraction from the lower level (hardware platform),
- scalability
- simplified storage infrastructure
- low cost solutions.
Thanks to SDS technology, you can significantly reduce the cost of storage and their administration. According to Gartner , by 2020, 70–80% of unstructured data will be stored on low-cost systems managed by SDS, and by 2019 70% of existing storage arrays will be available in a fully software version.
When and why need SDS
Storage management software should provide flexible data storage, as well as:
- deduplication
- data replication
- dynamic allocation of capacity
- snapshots of data
- adherence to storage policies.
A source
SDS is defined in the Storage Networking Industry Association (SNIA, Storage Manufacturers and Consumers Association) as a virtualized storage environment with a service management interface , which should include:
- automation — simplified management that reduces the cost of maintaining the storage infrastructure;
- standard interfaces - API for managing, allocating and freeing resources, servicing services and storage devices;
- data access paths virtualization - block, object and file access in accordance with application interfaces;
- scalability — changing storage infrastructure without reducing the required level of availability or performance;
- transparency - monitoring of consumed storage resources, their management and control of their cost.
I note that SDS needs a standardized management interface, such as the SNIA Storage Management Initiative Specification (SMI-S). It is an integral part of the concept of software-defined data centers (SDDC). This software logic of cloud storage infrastructure and cloud hardware platforms can be an element of traditional data centers. Data storage and processing services can run on servers, storage appliance, or both of these platforms, eliminating traditional boundaries.
Compare SDS Solutions
Software-Defined Storage is offered by many vendors:
- Dell EMC (Dell Nexenta, EMC ScaleIO Solutions),
- HPE (StoreVirtual VSA solution),
- IBM (Spectrum Storage Solution),
- NetApp (ONTAP Select solution),
- VMware (vSAN solution),
- Red Hat (Red Hat Storage Solution),
- StoneFly (SCVM, SDUS solutions),
- DataCore (SANsymphony solution),
- SwiftStack
- Pivot3 et al.
I’ll clarify that the RedHat Storage solution is represented by two products: RedHat Ceph Storage and RedHat Gluster Storage (RH Storage Server). Here they are both implied, but they did not participate in the comparison below, since they are significantly different from the other solutions mentioned.
Ceph is not really a boxed product. Its use without a staff of developers is rather difficult, which made it uninteresting for our company. Therefore, this solution is not in the comparative table.
Conventionally, all SDS solutions can be divided into three categories:
- classic (CEPH, Red Hat Storage Server, EMC ScaleIO),
- based on traditional storage systems (NetApp ONTAP Select, HPE StoreVirtual VSA),
- as part of computing systems (VMware vSAN).
Some manufacturers offer both integrated solutions and software (Huawei, Dell EMC). This allows a flexible approach to the selection of products and the use of inherited "computing" equipment to solve less resource-intensive data storage tasks. Another merit of SDS has been the ability to use disk arrays virtualization in some classic storage systems.
Architectural solutions are built on two principles:
- loosely coupled
- distributed (without common elements).
In the first case, fault tolerance is provided by distributed copies of data, but due to redundancy of communication between nodes (nodes), the write speed is reduced. The critical point is the data network, so these solutions are usually implemented on the basis of InfiniBand. Based on this principle, VMware vSAN, HPE StoreVirtual VSA, Dell EMC ScaleIO solutions are built.
In systems without common elements, data is written to one node, and then with a specified frequency is copied to the other to ensure fault tolerance. In this case, the records are not transactional. This approach is the cheapest. Ethernet is most often used as an interconnect. This architecture is convenient in terms of scalability. Its bright representative is CEPH.
Now many companies are developing both software SDS (for example, Atlantis Computing, Maxta, StarWind, DataCore Software, Sanbolic, Nexenta, CloudByte), and releasing complete solutions (Dell EMC, IBM) or specialized devices (Tintri, Nimble, Solidfire).
Of the most well-known on the market, we selected for comparison seven solutions that are most interesting for the tasks of “Onlanta”. It:
- VMware vSAN,
- HPE StoreVirtual VSA,
- NetApp ONTAP Select,
- EMC ScaleIO,
- Huawei Fusion Storage,
- StarWind Virtual SAN,
- Datacore SANsymphony.
In this table we compared their main characteristics.
Click to enlarge the table.
Tool of the future
The SDS technology began to develop in the early 2000s, but has not yet been able to replace the classic storage systems for a number of reasons - we are not going to discuss them now. But manufacturers are actively engaged in the development of their products and interest in SDS technology is growing. According to our estimates, in the near future they will become the tool that will allow reducing the cost of IT infrastructure with an increase in the need for increasing storage capacity.
In conclusion, I note that in this article I did not try to offer options for choosing the solution that suits you. Such a solution should be chosen based on the load, SLA, etc. The proposed table compares only the solutions capabilities, and does not compare performance, replication speed, node switching time, etc. This is a comparative analysis of opportunities, and not productive testing.
After a thorough acquaintance with the products of SDS, we came to the conclusion that in their current implementation, they do not fit our tasks very well. For ourselves, we nevertheless chose a classic solution, the implementation of which we are currently engaged in, and about what, perhaps, in the near future we will tell you.
But I hope that the presented comparison results will help you navigate, save time and make it easier for you to choose which solution is right for you.
If someone from the readers finds it possible to share any additional information on the subject under discussion, and perhaps to tell about his choice, it would be very interesting.
Source: https://habr.com/ru/post/324072/
All Articles