The path to fully virtualized infrastructure started with VMware virtualizing datacenter infrastructure in 1998. Up until then, at least on the x86 platform, if you wanted to run your workload you had to buy a separate server for it; this was the core reason internet startups in that age needed so much capital, you had to buy separate servers for each workload and you needed them to be in a centralized datacenter. Server sprawl was a real and huge issue. Shortly after that, at Cambridge University, the Xen project launched; Xen ended up being the virtualization platform of choice for multiple cloud providers with the budget to customize it. VMware managed to give enterprises a path to running their workloads in a more efficient manner and the result of reducing a computer to a set of virtual files led to a new wave of innovation; it was easy to back them up, it was easy to have replication, it was easier to ensure that they were stable and insulate the user from the effects of unreliable hardware. This made VMware a billion dollar company; this is also what unleashed a wave of innovation that we’ve seen in the cloud space since. VMware represents choice. VMware has doubled down on it’s Software Defined Data Center strategy that has seen it also virtualize networking, storage and compute. I have been using VMware infrastructure from around 2004/5 when I was introduced to workstation and shortly there after ESX.
Networking has always been a huge revenue earner; companies in the space have made a fortune. One would be wrong to think that the source of their revenue was the hardware; it has many times been from the software. Many OEMs required specialized training, they had their own software and operating systems; as hardware veered towards becoming a commodity, and software the secret sauce. This lock-in was very problematic for companies as, once you went down a path with one vendor, you were stuck and you couldn’t really choose another vendor (you would many times have to retrain your staff, etc). Naturally, this chaotic environment made it fit for disruption. There was a huge problem even within VMware around networking; you could have a virtual machine sitting in a physical host, you could move that virtual machine to a physical host whose network you couldn’t access. So, naturally, the next thing that VMware did was to abstract networking, in 2012, with their purchase of Nicira. For a very long time, anyone in the virtualization space had a huge headache: how do you deal with networking? VMware virtualized everything from layer three upwards, ensuring that companies could now have a single pane of glass for their operations. This was NSX. This was a game changer in networking as it removed the vendor lock-in once and for all. At Node Africa we currently use three different networking vendors in tandem, simply because it was what we could afford at the time. Given that all the intelligence is handled by NSX all we needed to do was use some basic managed switches and away we went. That’s saved us money and made it significantly easier to run our operations
Storage has had the same model as networking and compute – the OS was a way of selling storage hardware at a premium. Similar to the networking scenario, software is also the magic has allowed OEMs to sell the storage at a premium. At the end of the day, storage providers were writing their software atop of regular *NIX systems and selling them at a premium. This presented a huge opportunity for the storage vendors but a lot of pain for the customers, simply because the vendors locked you into a proprietary platform; you weren’t allowed to scale up your storage array with commodity hardware if you needed to. One had to buy a new storage array from the current vendor. This presented several sizing challenges, e.g.
- One would either under-provision one’s storage array (raw capacity) for a low upfront purchase cost, this had one major issue – should your storage requirements grow faster than anticipated (which oft happened) there was no direct upgrade path and you had to buy a new and very expensive storage array
- One could over-provision one’s storage array and end up with a device that is never fully utilized and, by the time they were fully utilized, the storage arrays would be end of life
- Sizing for I/O requirements was particularly problematic; many companies ended up choosing to over-provision for this; there was no way to right size for what you needed at the time.
- Sizing for networking was another huge problem as you had to many times get expensive fiber gear that was only useful for the array
We’ve spoken to too many customers whose hardware is either end of life, or the operating system for the storage array is no longer supported, or they’ve filled up their array and the only solution that’s being proposed is another expensive monolithic storage array.
These are the problems that VMware’s vSAN is looking to solve. vSAN is a distributed scale-out storage platform and it is designed to offer a storage pool within a VMware cluster. We use vSAN at Node Africa; we have DL 380 servers that double up as both compute and storage. vSAN takes the same approach VMware has taken previously with networking and compute infrastructure: decoupling the storage OS from any one vendor and allowing you as a company to make the right choices regarding your hardware.
With software defined storage, you are no longer locked into a particular hardware platform, you can simply buy compatible commodity server hardware off any x86 vendor and roll out your storage. You have the ability to mix and match your requirements; you can optimize your server for disk speed (say SSD) or capacity, etc. Your storage array scales infinitely, so you could start with a very small low capacity group of servers and scale it out to over 1 petabyte. That’s flexibility that’s unparalleled. You can choose to run affordable networking that can be upgraded over time, which may not be necessary as when you increase the number of pods as the total throughput of the array increases. You can have up to 64 pods.
With vSAN, you get higher performance at a significantly lower cost, with one additional benefit; your storage array also doubles up as your compute platform; you don’t need to buy additional servers – you can completely get rid of your storage budget and invest only in servers that come with their own networking, compute and storage, pool the storage that you get from the servers and use that for your distributed storage. You can connect your physical servers to this storage pool (you have the flexibility to create whichever targets you want) and you can use them for most of your day to day operations
Hence, the biggest benefits of vSAN are:
- It’s cheaper to build and run your array. Storage typically was responsible for roughly 40-60% of datacenter infrastructure costs. This figure can be reduced significantly with right sized vSAN clusters
- It’s future proofed; many cloud providers can manage your workloads on vSAN (AWS, IBM, we support it at Node Africa).
- The cost of entry is very low and if the experiment fails and you conclude that vSan doesn’t work for your organization, you still have regular compute servers, the risk is very low.
- Compatibility with the current VMware environment
- Erasure coding helps protect against data loss and increases storage efficiency
- Object storage is a more efficient way of handling files (though for some workload types block storage is better).
- You can upgrade your networking infrastructure, disks, and servers as you go along; your storage array is finally designed for exactly what you need – no more no less. It’s fully customizable and bespoke; you won’t be constrained by a manufacturers idea on what your options are at that stage.
Scenarios where it’s easy to introduce vSAN:
- You are a young company and you want to have a very small footprint within your DC or in a managed cloud provider
- You have a mid-cycle hardware refresh, this is an easy way to have vSAN introduced to your DC as you can simply buy regular compute servers that are vSAN compatible and have more and increased adoption
- There are budgetary constraints and your hardware is end of life. In many such scenarios, vSAN works out to be cheaper than any comparable storage array in the market
This is the future of storage, one with unparalleled flexibility. The risk of vendor lock-in is very low as you can easily port your data out of the platform but, more importantly, it saves you money while giving you higher performance. VMware has also over time demonstrated its commitment to open platforms with support for the Open Virtualization Format, support for public cloud providers, etc. We are working hand in hand with them as our partners for this reason; they have a shared commitment to helping the customer succeed in the longer run. We’ve deployed this seamlessly for a leading investment bank in Kenya. This technology works.
If you would like to see how software defined infrastructure can help your business or need an audit of your current infrastructure, do reach out to us at [email protected]