By Cobus Smit – 5 March 2020
We have a few major software-defined storage vendors, and this is “according to me” the core of Hyper-converged. Without SDS, you will just have another X86 system. In my opinion, this started the whole HCI game.
SDS is a way in which software, rather than hardware, defines storage characteristics like performance, availability, and resiliency. Such programs are independent of the physical storage devices, which allows eliminating their limitations. The main goal of SDS is to provide you with a centralized management and efficient utilization of storage resources and the ability to meet specific data requirements. You can look at this as vendors build their own software to manage the disks inside their servers…
Multi-billion dollar companies were started on some of these storage features. A lot of these vendors, to name a few (NetApp, HPE etc..)do not have an SDS solution when it comes to their hyper converged solution and they depend on another company like VMware that use vSAN.
To be a player in Hyper converge today and specifically SDS you have to integrate the below as a minimum:
- Continuous data protection
If your HCI company or vendor does not have these then don’t even bother in the HCI space. Therefore, smaller and well known HCI vendors out there uses some version and software on top of their SDS hardware. We will talk about some players they depend on below.
Some vendors (that help other hardware vendors):
- an Open Source, massively scalable storage solution for modern workloads like cloud infrastructure, data analytics, media repositories, and backup and restore systems. a Project backed by many IT corporations such as SUSE, Canonical, Fujitsu, Intel, and Red hat. Ceph is unique in its kind because is a distributed storage management tool that exposes object, block, and file systems.
Storage Spaces Direct
- Storage Spaces Direct uses industry-standard servers with local-attached drives to create highly available, scalable software-defined storage. Its converged or hyper-converged architecture simplifies procurement and deployment, while features such as caching, storage tiers, and erasure coding, together with the latest hardware innovations such as RDMA networking and NVMe drives, deliver good efficiency and performance.
Distributed Storage Fabric (DSF)
- This is at the core and birth of the Nutanix platform and expands upon the Nutanix Distributed Filesystem (NDFS). NDFS has now evolved from a distributed system pooling storage resources into a much larger and capable storage platform.
- vSAN is part of the VMware ESXi kernel and runs on industry-standard x86 servers from original equipment manufacturers (OEMs), including Cisco, Dell EMC, Hewlett Packard Enterprise, Fujitsu, Lenovo, and Supermicro, etc… Because vSAN is a native component, it does not require additional software, and the user can enable it with a few clicks.
VxFlex OS – formerly Scaleio
Dell EMC VxFlex OS
- (formerly named Dell EMC ScaleIO) is a software-only solution that uses existing servers local disks and LAN to create a virtual SAN that has all the benefits of external storage, but at a fraction of the cost and complexity. Using the driver, Block Storage hosts can connect to a VxFlex OS Storage cluster.
Almost all of your HCI OEM hardware vendors will run one of these as their SDS solution.
One sales cliche’s that everybody advocates is that it avoids vendor lock-in
It does allow you to load the software on most x86 hardware, but most vendors will lock you in just in a different way. But if you implement your own open-source software-defined storage like CEPH. You will require skills and if you use one like Nutanix, vSAN or VxFlex they still kind of lock you in. Vendor lock-in is just another way smaller players try to sell a solution where they are dependent on the above players so now you have double lock-in 🙂
Some pros for SDS
- It can be flexible with and easily deployed with vendors like Nutanix, vSAN and Flex OS. The smaller players like the HCI vendors in my menu have awesome software that integrates with the above SDS solutions and makes it easy to implement and manage.
- It integrates with your infrastructures and you can “kind of” easily scale up with compute and memory, just add nodes to your environment.
- There has always been a discussion and blogs regarding poor performance on SDS but this is not relevant anymore as caching and CPU usage for storage is at a much better place than in 2013.
- A good feature of software-defined storage solutions is that most vendors particularly design it to allow the automation of workflow through the APIs that are openly programmable. It is useful for the execution of the integrations of the applications. And it is also necessary as they depend on the SDS of other companies.
- Software-defined storage solutions infrastructure abstracts the resources in the infrastructure from the hardware living under them.
- You can use your existing hardware – SOMETIMES. This is another selling point to double-check.
- In the long term, it will in most circumstances save you money compared to a traditional SAN or external storage device.
Some cons for SDS
- It is actually more complex to manage because they spread out it on all your servers. The servers guys are not used to managing storage, so it could be a burden on them.
- Software-only solution – what BS as you will always require some kind of hardware. This is just another selling point to make you think software is the future. While it is true, you must just look at it in perspective of your solution.
- You always had a SAN/Backup operations team, and now the vendors tell you you will need less staff as you need not manage the storage silos anymore. This is also not true as SDS will use more staff to manage properly, and today’s SANs are not like the 2001 silos anymore. It is easier to manage.
- You will always need a bigger cluster, more licenses with multiple nodes if you require high throughput and IOPS.
My conclusion is that you need to be informed of what your business requirements are when choosing a cloud solution. Remember that the vendors out there have professional writers that do not know the solution the way they should. They use these people and some charismatic executives to sell those solutions, which is a multi-million marketing game. They also just add some software to run their plain old Intel servers… This will gradually phase out as open-source like CEPH get more widespread and known. The time will come where open source projects will get easier to implement and HCI vendors will gradually fade away, my opinion only 🙂