Scale Computing: when you don't have time for gimmicks
June 26, 2017
Scale Computing: when you don't have time for gimmicks
Score 10 out of 10
Vetted Review
Overall Satisfaction with Scale Computing HC3
Scale Computing is used by us for both primary production workloads and mission critical testlab infrastructure workloads. We have a cluster at a co-location facility whose purpose is to host websites, a virtualized file server, some e-mail servers, a few databases and some VDI instances. These are the workloads upon which our small business depends.
One of the services we offer our customers is Testlab as a Service. We do everything from reviews for technology magazines to full tear-down beta testing. Accomplishing anything remotely like efficiency and profitability requires that a great deal of the testlab's digital infrastructure be rock solid. This includes domain controllers, e-mail servers, databases, monitoring software and much, much more.
We also use a Scale Computing SNAC node (https://virtualizationreview.com/articles/2016/11/30/scale-computing-single-node-appliance-review.as...) to back up our workloads.
One of the services we offer our customers is Testlab as a Service. We do everything from reviews for technology magazines to full tear-down beta testing. Accomplishing anything remotely like efficiency and profitability requires that a great deal of the testlab's digital infrastructure be rock solid. This includes domain controllers, e-mail servers, databases, monitoring software and much, much more.
We also use a Scale Computing SNAC node (https://virtualizationreview.com/articles/2016/11/30/scale-computing-single-node-appliance-review.as...) to back up our workloads.
- Scale is reliable. Drives can fail. Nodes can fail. All sorts of weird things can happen, but data doesn't get lost.
- Scale support is fantastic. I call them "Datacenter Butlers" because they take a lot of the miserable scut work off of plate. I used to constantly put out fires just to keep things running smoothly. Now, whenever there's a problem, Scale support likely knows about it before I do, and will fix most of it for me.
- Speed. Scale isn't the fastest on the market, but they also cost a fraction of the highest end gear out there. It doesn't make a lot of sense to compare, for example, a hybrid solution using SAS/SATA to EMC's top-end custom all-flash DSSD. That said, dollar for dollar, I would say Scale manages to be the fastest solution amongst their peers.
- Scale pay attention to the little things. They have a robust QA program and an extensive beta testing program. They get industry experts to bang on their latest and greatest before it hits the streets and it shows. Scale is the opposite of the current tech trend to push out alpha-level software as generally available and then patch things according to the pyramid of screaming.
- Scale needs a new UI to be able to handle more than 500 VMs in a single cluster.
- Scale needs a multi-cluster management UI.
- Scale needs a full private cloud offering with a self-service portal, template and recipe marketplace and fully integrated configuration management.
First off, I'd like to object to the statement that "Hyper-converged Infrastructure products like Scale Computing HC3 are designed to make it easy to run applications in a highly scalable manner" in the prompting question. They're not. They're designed to make running workloads easier. Name me one HCI provider out there that makes it easy to scale workloads. Just one. You won't be able to because that's not the job of HCI. That's the job of the management and orchestration layer that turns virtualization - whether HCI or not - into something most of us would call a "cloud".
I think this is an important piece of context for discussion. HCI in general - and Scale's offering very specifically - is based emphatically on ease of use. If what you're looking for are edge cases like high frequency trading or other extreme use case niches, look elsewhere. That's not what HCI is about.
Those workloads are the equivalent of commuting to work on a Saturn V. Think of Scale Computing more like the Honda Civic of IT. It is powerful enough to deal with the overwhelming majority of use cases. It will meet the needs of the mass market. It will run all your day-to-day mundane infrastructure workloads from the SMB to the enterprise and fit very nicely in ROBO scenarios.
What you're not going to get from a Scale HCI cluster - at least not until an all-flash NVMe version running on top of 25GbE or 100GbE networking comes out - is something that will handle huge databases with ultra low latency requirements and trillions of transactions per second. Nor is Scale going to be the super cheap near-zero-margin $/TB bulk storage solution required for large-scale data warehousing. If you're building a massive radio telescope, maybe look elsewhere.
But for the rest of us, for those running a few thousand people in an exchange environment, or a modest use case database for a point of sales application, for some VDI or image rendering or a hundred other workloads I've run on Scale, it just works. That's what hyperconvergence is: It Just Works(TM) in a turnkey appliance for 90% of workloads you could ever want.
I think this is an important piece of context for discussion. HCI in general - and Scale's offering very specifically - is based emphatically on ease of use. If what you're looking for are edge cases like high frequency trading or other extreme use case niches, look elsewhere. That's not what HCI is about.
Those workloads are the equivalent of commuting to work on a Saturn V. Think of Scale Computing more like the Honda Civic of IT. It is powerful enough to deal with the overwhelming majority of use cases. It will meet the needs of the mass market. It will run all your day-to-day mundane infrastructure workloads from the SMB to the enterprise and fit very nicely in ROBO scenarios.
What you're not going to get from a Scale HCI cluster - at least not until an all-flash NVMe version running on top of 25GbE or 100GbE networking comes out - is something that will handle huge databases with ultra low latency requirements and trillions of transactions per second. Nor is Scale going to be the super cheap near-zero-margin $/TB bulk storage solution required for large-scale data warehousing. If you're building a massive radio telescope, maybe look elsewhere.
But for the rest of us, for those running a few thousand people in an exchange environment, or a modest use case database for a point of sales application, for some VDI or image rendering or a hundred other workloads I've run on Scale, it just works. That's what hyperconvergence is: It Just Works(TM) in a turnkey appliance for 90% of workloads you could ever want.
- Scale frees me up from looking after my infrastructure so that I can engage in revenue generating activities. Pretty hard to argue with that as a net result. I cost $100k/year. I used to have to put at least 40% of my time into keeping the lights on. So right off the bat Scale frees up nearly $40k/year for us, and that's before I factor in things like not having to buy shared storage, or other headline savings.
- The only negative impact Scale has for us is that unlike previous solutions, we aren't provided with spares of consumables (read: hard drives) to keep on the shelf for when something dies. The time between drive failure and the new drive arriving in the post is a bit nerve wracking.
A question like "What other products like Scale Computing HC3 have you used or evaluated?" can be a bit leading. What do we consider "like" Scale? Nutanix? They're HCI, but they play in a completely different arena. VMware's VSAN? SimpliVity? Yottabyte? DO we go a little orthogonal and take aim at other storage vendors, from behemoths like EMC to next-gen startups like ioFABRIC? I've played with virtually everything on the market, used a significant chunk of it in production, and there are use cases where each of them will pummel their competitors into submission.
Scale Computing has a niche: they make virtualization infrastructure for situations where it has to be dead simple to use and utterly reliable. To get there, they sacrifice a lot of functionality, and they're quite a bit behind the curve when it comes to the latest trends in computing. They do not, for example, have a generally available containerization solution, integrated configuration management, a self-service portal, marketplace, etc.
Of course, it's hard to argue many of their competitors do. Yottabyte has all the pieces, and has put them together in the right order. So does Hypergrid. VMware has all the pieces, but there are 15 different groups arguing over what the end result should be and the puzzle doesn't get assembled. Nutanix is slowly building something future proof, but they cost so much that's like comparing a Toyota Corolla to a Lamborghini.
I could talk about other competitors like Maxta who are maybe more in a similar price and feature range as Scale, but even then there are differences. Scale is a bare metal solution with most of their technology in the storage and automated remediation layers. Maxta has bet the farm on analytics.
Each of the dozens of companies vying for a place in the hyperconverged infrastructure market - to say nothing of storage as a whole - has their own shtick. Scale met our needs when we were at a point in our company's evolution where we simply couldn't afford to worry about our IT. That was the most important thing, and that drove us to use Scale.
If we were more focused on regulatory compliance we might seek out a vendor with better analytics and reporting. If we needed cross-platform integration we might look to vendors that managed multiple hypervisors. And if we were exposing our infrastructure for consumption by customers we'd look for a vendor offering hybrid cloud solutions. We weren't really needing any of that, and I, for one, am glad Scale was around to offer a more no-frills solution for the busy small business owner.
Scale Computing has a niche: they make virtualization infrastructure for situations where it has to be dead simple to use and utterly reliable. To get there, they sacrifice a lot of functionality, and they're quite a bit behind the curve when it comes to the latest trends in computing. They do not, for example, have a generally available containerization solution, integrated configuration management, a self-service portal, marketplace, etc.
Of course, it's hard to argue many of their competitors do. Yottabyte has all the pieces, and has put them together in the right order. So does Hypergrid. VMware has all the pieces, but there are 15 different groups arguing over what the end result should be and the puzzle doesn't get assembled. Nutanix is slowly building something future proof, but they cost so much that's like comparing a Toyota Corolla to a Lamborghini.
I could talk about other competitors like Maxta who are maybe more in a similar price and feature range as Scale, but even then there are differences. Scale is a bare metal solution with most of their technology in the storage and automated remediation layers. Maxta has bet the farm on analytics.
Each of the dozens of companies vying for a place in the hyperconverged infrastructure market - to say nothing of storage as a whole - has their own shtick. Scale met our needs when we were at a point in our company's evolution where we simply couldn't afford to worry about our IT. That was the most important thing, and that drove us to use Scale.
If we were more focused on regulatory compliance we might seek out a vendor with better analytics and reporting. If we needed cross-platform integration we might look to vendors that managed multiple hypervisors. And if we were exposing our infrastructure for consumption by customers we'd look for a vendor offering hybrid cloud solutions. We weren't really needing any of that, and I, for one, am glad Scale was around to offer a more no-frills solution for the busy small business owner.