Alternatives to VMware: Scale Computing

This is one of a series of posts looking at alternatives to VMware following the completion of the acquisition of the company by Broadcom. In this post, we review Scale Computing, Inc., with a focus on edge computing.

Background

Scale Computing was founded in 2008 (before many HCI vendors) by Ehren Maedge, Jason Collier, Jeff Ready, Mike Olson and Scott Loughmiller. To date, the company has raised around $200 million, with the latest funding round of $55 million in July 2022. Olson left Scale Computing in 2011, Maedge left in 2012, and Collier departed in 2019. However, Jeff Ready remains as CEO, while Loughmiller holds the position of Chief Product Officer.

At the heart of Scale Computing’s solutions are a series of hardware appliances that run KVM virtualisation and bespoke management software (which we will detail in a moment) to deliver hyper-converged infrastructure (HCI). The initial products were focused on the SMB market. However, in recent years Scale Computing has developed solutions for the edge market and scaled up with all-flash nodes.

SC//HyperCore

Scale Computing Platform (SC//Platform) HCI solutions are a tightly integrated combination of multi-node clustered servers and software. The software component, branded SC//HyperCore, is a customised Linux installation running KVM and a series of additional bespoke software components. These include SCRIBE, HEAT and AIME.

SCRIBE (Scale Computing Reliable Independent Block Engine) is a scale-out storage layer that spans each of the nodes in a SC//HyperCore cluster. In contrast to the design of Nutanix’s NTFS, for example, the storage functionality runs in the operating system hosting KVM and not within the virtual machines on the host. This design is similar to how VMware implements vSAN. SCRIBE provides the resiliency and availability management for storage across the cluster, automating rebuilds and re-protecting data as clusters are expanded and contracted.

SCRIBE is complemented by HEAT (SC//HyperCore Enhanced Automated Tiering), a feature that ensures data in hybrid configurations (mixed HDDs and SSDs) makes optimal use of the more expensive SSD storage. Theoretically, HEAT could also be used to manage all-flash configurations with a mix of fast and slow (or expensive and cheaper) SSDs.

The entire SC//HyperCore environment is managed by the AIME state engine (Autonomous Infrastructure Management Engine). AIME monitors the hardware and software, ensuring that a cluster operates at the desired state of the customer. AIME performs root cause analysis on problems and automatically repairs them where possible, raising alerts where issues can’t be resolved in software.

Hardware

Scale Computing now offers five hardware variations, four of which are rack-mounted systems and the fifth a small form-factor design. The HE100 series is based on Intel NUC and designed for edge computing requirements such as branch offices or retail environments. HE500 series are 1U servers with up to 128GB of DRAM per server. The high-end HC5000 series provides up to 2TB of DRAM per node, NVMe storage and 25Gb networking.

More on the hardware configurations can be found online (here).

Comparing Scale Computing to VMware, we will look at the seven aspects highlighted in our introduction blog post (available here).

Hypervisor

Scale Computing HyperCore implements KVM virtualisation. KVM is a mature and efficient hypervisor built into the Linux kernel and has over 15 years of development behind it. KVM uses hardware-based virtualisation through the use of instruction extensions added to the x86 architecture (Intel VT-x and AMD-V) in the mid-2000s.

KVM supports physical device driver emulation (full virtualisation) and virtual device drivers (VirtIO) for storage and networking (also known as paravirtualisation). VirtIO drivers for both storage and networking have been built into the Linux kernel since release 2.6.27, therefore, Linux guests are supported automatically. Windows guests may need to slipstream drivers at first boot (SC//HyperCore provides a CD-ROM device at boot time for this purpose).

Platform Support

SC//Platform is sold as a tightly integrated hardware and software package. As a result, Scale Computing performs all the validation to ensure successive versions of SC//HyperCore continue to support customer hardware. For large enterprise customers, an integrated package may not be the most desirable configuration. However, we need to look at Scale Computing’s origins, which was to target SMB customers. In this part of the market, simplicity and ease of use/deployment are more critical success factors. This makes integrated solutions more attractive for long-term deployment and support.

Ecosystem

Scale Computing HyperCore provides a local and integrated GUI for systems management. There is also a REST API integrated into the platform that enables automation and scripting of any command available via the GUI. We’ve looked at the capabilities and ease of use of the API, including a small piece of code to query a running system and extract a list of virtual instances (see this blog post). It is easy to build some simple automation processes, but for large-scale environments, a more comprehensive fleet management solution is needed.

Fortunately, in May 2022, Scale Computing announced a fleet management SaaS platform (SC//Fleet Manager) for customers to manage large-scale deployments. This feature is great for large and dispersed environments that need a visual overview of hardware and virtual instance status. We believe Fleet Manager evolved from the requirement to support a large retail client running thousands of small SC//Platform clusters. In February 2023, Scale Computing introduced Zero-Touch Provisioning, a feature that enables customers to rapidly install and configure remote hardware at the edge without needing to dispatch an engineer to manage the process.

Commercials

As a hybrid solution of hardware and software, configurations of SC//Platform are typically sold as a 3-year hardware purchase, including all software and support, priced per node. In 2018, Scale Computing extended the licensing model to enable MSPs to sell SC//Platform on a price-per-node, monthly Opex subscription.

Scale Computing doesn’t license the SC//HyperCore software separately, other than as a disaster recovery option in Google Cloud Platform known as Cloud Unity. So, for some customers, the lack of hardware flexibility may be a problem. However, as Scale Computing continues to grow, then the option to license the software may become a possible option for attracting some customers to switch from competing platforms. We believe this type of offering is subject to the company being able to put adequate support capabilities in place.

Deployment Model

We’ve touched on this already, highlighting the edge-focused HE150 series and HE500 which is also suitable for edge locations. Currently, SC//Platform is an on-premises solution only. Cloud Unity exists to provide disaster recovery support only. We don’t have visibility of whether this direction may change in the future, although we can see no technical reason why SC//HyperCore wouldn’t run on the public cloud, for example.

Features

The depth of features is possibly the only area where Scale Computing is at a disadvantage compared to VMware or Nutanix. SC//Platform does not offer complex software-defined networking, for example. There is no application management capability or self-service features that might be expected for larger customers.

We see the lack of these features as both an issue and an opportunity. Large enterprises will be reluctant to consider a solution that isn’t as fully complete as VMware. However, at the same time, the API offers the possibility for third-party vendors to build on the platform and front-end some of the missing capabilities.

Migration and Mobility

SC//HyperCore supports the migration of virtual instances between clusters. As the underlying virtualisation solution is KVM, virtual instances can be exported and restored from a local SMB server. We’ve tried importing non-Scale Computing virtual machine images, which is possible but requires some work editing configuration files, which is not an ideal scenario.

Scale Computing offers a solution called Scale Computing Move, which is essentially Double-Take software under the covers. This option is offered by Scale Computing and partners. We can see more opportunities here for further migration tools to be developed that work with the SC//HyperCore platform to move instances between virtualisation solutions.

Obviously, as already mentioned, there is currently no support for SC//HyperCore on public clouds, so mobility is limited to cross-cluster and import/export options.

The Architect’s View®

We’ve been following Scale Computing for the last ten years and running an SC//Platform test cluster in some form since 2015. The technology is reliable and robust, with an intuitive interface and solid API. The challenge for any potential customer looking to migrate from VMware is the lack of advanced capabilities, such as self-service or zero-trust networking. This makes SC//Platform more suited for the SMB and midrange market than complex enterprise customers.

However, as we watch the evolution of VMware under Broadcom, it’s clear that the top 20% of existing VMware customers will be the focus for Broadcom, while mid-range and SMB customers that don’t need or want the advanced features are much more likely to consider alternative solutions. As a result, we believe that Scale Computing is strongly positioned to gain customers that simply want a reliable HCI platform for running virtual instances, particularly those with edge requirements. For this part of the market, Scale Computing will be hard to beat.