Until recently, managing your storage and infrastructure was often a lot like this. Inefficient, messy and downright painful. These are some of the examples of what was considered normal for way too long:
- Products with designed-in obsolescence
- Effectively forceful tech refresh cycles
- Black box solutions (i.e we know better than you)
- Controller bound bottlenecks. If you ran out of performance before that next refresh you were forced to buy a more powerful model
- Disruptive data migrations
- Architectures that don’t work for your business requirements
- Proprietary hardware technologies – duplication of different hardware standards
- Locked-in vertical stacks
Fortunately there has been an overall improvement in the industry as people have been voting with their wallets to escape these traps – by moving some workloads to public cloud, adoption of HCI / software-defined solutions and so on.
For those in the know though – there has been one product out there that has been avoiding all these pitfalls since day 1 for at least six years now —- PowerFlex (regardless of how many name changes it’s had!). This is proper enterprise-grade software defined storage.
Let’s drill into those pain points and see what things look like from a PowerFlex perspective:
- Products with designed-in obsolescence
- Effectively forceful tech refresh cycles
Most storage products in the market today go through a design cycle which eventually forces you into an upgrade of all the hardware and software involved, i.e. a forklift upgrade.
PowerFlex on the other hand being software-defined storage, allows you much more freedom in when and how you will perform your next hardware upgrade. Maybe you’re fine running servers that are 6-7 years old because they are meeting your requirements, and when you do decide to upgrade – the PowerFlex software itself is a perpetual license so you do not need to purchase it again. This greatly helps your TCO / ROI over the long term.
- Black box solutions (i.e we know better than you)
“You can have any color you like, as long as it’s black.” – This is what you get with traditional storage products. Components that have been pre-chosen for you, that only come in S, M, L sizes.
While there is a much merit in engineered HW+SW solutions, the restrictive nature of the market today still leaves a lot to be desired.
With PowerFlex however, you can choose from an almost limitless amount of configuration choices. Do you want SATA, SAS, NVMe or Optane based devices? No problem. Do you need 10/25/40/100GbE connectivity? Go ahead, choose what’s right for you. How about the latest and greatest Intel “R” CPU’s? We’ve got you covered. The Flexibily in PowerFlex isn’t just a name, it’s a mandate. The best part is that this flexibility is delivered to you in a way where the testing and QA on all the HW components has been done for you in advance, giving you all the benefits of an engineered system, but without all the onerous rigidity.
- Controller bound bottlenecks. If you ran out of performance before that next refresh you were forced to buy a more powerful model
Too many people underestimate what performance means and what it brings to an organisation when it comes to block storage. It means efficiency in getting data to a CPU for processing. Keeping those I/O wait states as low as possible, getting the most efficiency out of your CPU/RAM to deliver results to your organisation. It also means extremely high-availability, with the PowerFlex mesh-mirroring architecture this results in rebuild speeds like you have never seen before in your life. This means no more complaints about slow storage, about degraded storage. And so on. Suffice to say, performance always matters when it comes to block storage.
With every PowerFlex node being a controller, performance bottlenecks due to this are a thing of the past. Imagine 512 nodes working in parallel and then compare this with traditional enterprise storage arrays of only 8-16 controllers.
Further to this performance conversation, although PowerFlex can reach incredibly high IOPS (~200K+ IOPS per node), for many workloads what is more important is the consistent and predictable latency that it provides. This is ideal for OLTP like workloads and results in more transactions per second and an overall much better user experience in terms of ‘snappiness’. Much like when the world moved from HDD’s to SSD’s, the performance gains are soon taken advantage of and become the new norm. Try taking it away from a user once they have had a taste of performance!
- Disruptive data migrations
How in this day and age any kind of disruptive operation can be tolerated is beyond me. Even ten years ago we were in a 24/7/365 kind of world and downtime is simply not an option – yet many legacy storage arrays still have terrible data migration strategies today that take immense planning and coordination. The hidden costs are simply crazy.
PowerFlex has always been a non-disruptive upgrade process starting from the earliest 1.32.x release all the way up to the current 3.5 release. Apart from some very basic co-ordination, this can be done 100% online without affecting the applications running on top.
- Architectures that don’t work for your business requirements
Does a traditional architecture work best for you today? i.e. Compute and Storage separate – or does an HCI architecture work better for you now in terms of space savings / power savings? Both architectures have their merits and there is no right or wrong answer when it comes to these choices – as a favourite saying in the IT industry “well… it depends”.
Another great aspect of PowerFlex is that you get to choose – and you are never stuck with your choice as your evolving requirements change. Many organisations are starting with HCI but they then find that they only need to add some Compute-only nodes, or some Storage-only nodes. Very few things on this planet grow in neat little predictable ways – which is why PowerFlex allows you to add only what you need. This can help avoid a tremendous amount of waste and effort around trying to move your workloads around to get the best fit. It can also help you dramatically save cost on application licensing and hypervisor licensing too.
- Proprietary hardware technologies – duplication of different hardware standards
Do you really need Fibre Channel these days? To put it bluntly – no. So why have different technology standards in your DC when everything can run on standard x86 and Ethernet?
PowerFlex is IP based, with proven native IP multi-pathing. If you prefer to keep your storage network air-gapped, then PowerFlex will let you do this without any worries whatsoever. Would you prefer to save some cost and run on your existing ToR switches? No problem either.
- Locked-in vertical stacks
Remember Storage-arrays where the whole point of centralising storage was to eliminate wasted local storage on local nodes? And to aggregate performance? Ironically, many other HCI solutions today end up unintentionally creating silos over again – Storage tied to nodes with a locked-in architecture that effectively leave you at the mercy of that vendors future direction. Change of direction? Moving your workloads to another hypervisor or container platform? Too bad – your storage investment is now worth $0.
PowerFlex is a horizontal architecture that gives you all of the benefits of traditional centralised SAN storage – but also with the freedom of a software defined solution. Need to allocate capacity to a Windows host today? Do it. OpenStack tomorrow? No sweat. Google Anthos, Rancher, OpenShift — go dog go.
- Summary
This is a technology that has been close to my heart ever since I first came across it five years ago. From my days as an end-user, I wished for a product like this to exist all the time. I had tons of stranded storage doing nothing in x86 nodes, whilst my storage-array was unable to cope with the ever increasing workload. From my days as an account pre-sales, I wished for a product that would allow my customers to scale almost limitlessly and do their own data migrations non-disruptively. I have better things to be doing than to be going through the miniature of how many LUNs and what sizes they are that need to be migrated.
This mature technology is ready for your most demanding workloads and mission critical systems. If you’re in APJ and you would like a deeper overview, please let me or your Dell Technologies account team know. We’d be more than happy to show you a better way.
