Category Archives: Commissioned

The Scale HC3 Advantage

For years I have been pitching ideas about how a performant, highly reliable architecture for the small and medium business market should be approached.  Concepts like maintaining high reliability across all layers of the architecture, reducing the total number of points of failure whenever possible, reducing complexity, leveraging mirroring instead of parity for storage, keeping storage close to compute and so forth.

What I like about the Scale HC3 approach to hyperconvergence is that it addresses all of these items and more.  Scale takes all of the components of a complete platform and reduces them into a singular “package” that is deployed to all nodes in the system making every layer of the system as redundant as all of the other layers.  Even the management console itself is automatically replicated and available from every node in the cluster,  independently!

Compute and storage are integrated into the same package sitting together on each node.  This allows for the lowest latency in communications and the highest reliability both per node, as there is less to fail by way of having fewer moving parts, and also across the entire architecture by making each stable building block replicated so that nothing depends on any single node for computer, storage, management or orchestration.

Instead of the “traditional” approach so often associated with virtualization, where several layers of services are separate yet totally dependent upon each other, the Scale hyperconverged approach flattens these taking us to the opposite end of the spectrum from the infamous “inverted pyramid of doom” to a flat, broad “brick”.  No single points of failure, no highly fragile components.  Keep each component reliable on its own, then make them redundant anyway.

Digging down to the storage layer, the Scale HC3 uses a RAIN approach which can most easily be described as a “network RAID 1e” design.  RAID 1e is an interesting twist on RAID 10 which basically adds a striping mechanism inside the mirrors so that you never have anything but mirroring but the mirrors are not at the disk level but at the block level and spread out among many disks.  RAID 1e is more advanced and more flexible than RAID 10 and brings some minor performance and reliability advantages but, more importantly, allows a RAID 10 approach to be used effectively over a network without having to resort to the fragility of RAID 01.  Scale does not use actual RAID, but true RAIN, that does intelligent block mirroring in the cluster with logic not only about drives, not about nodes on which the drives sit.  We only use RAID as an example for comparison.

The Scale HC3 also adds powerful hybrid spinning disk and solid state drive tiering using a heat map approach to determining which blocks are being used regularly and which are predominantly idle.  This allows the system to intelligently move blocks from slow spinning storage to high performance SSD for performance without needing all storage to sit on SSDs.  This keeps cost down while also allowing for large capacity.  The heat map approach, coupled with a simplistic priority management system, makes this nearly transparent to end users.

Scale also takes the management of a complicated beast like a full virtualization stack and makes it simple and easy for a small business IT department to manage.  SMB IT departments are very busy places being pulled in a lot of simultaneous directions.  It is very important that their solutions, especially the critical ones on which the most depends, are elegant, well supported, effective and easy to manage.  Scale provides a solution that allows for growth, performance and high reliability while maintaining an astonishing degree of ease of use. Even an IT department with all of the resources and experience to manage the most complex and demanding infrastructure can benefit by not wasting time where there is no competitive advantage and, instead, putting their efforts to use where they can most benefit the business.

The Scale HC3 allows small and medium business IT departments to stop worrying about how to design a system to meet their needs, and instead focus on the what’s most important to their business and what provides them a unique, competitive advantage.


[Commissioned for the Scale Blog]

Leveraging Microsoft RDS on Scale HC3

One of the great advantages of a centralized and unified infrastructure platform like the Scale HC3 is the ability to use the platform to provide centralized desktop and end user services along side traditional server services.  The high speed back plane allows server and desktop resources to communicate at high speeds and the centralized management lowers the total cost of ownership often associated with these types of services.  Applications like file serving and latency sensitive communications particularly benefit from the architecture.

For the majority of businesses, the approach that will make the most sense to add remote end user computing services to their environment will be through the use of Microsoft’s own Remote Desktop Servers or RDS, as most environments seek to provide an experience similar to  a traditional Windows desktop.  Microsoft’s RDS is a powerful toolset and relatively easy to implement.  It is a great starting point for offering a range of services.

Because of the high availability nature of the Scale HC3 platform we have good choices as to how to approach planning and provisioning an RDS deployment.  We can choose to keep our deployment small and simple and utilize the Scale HC3 cluster’s built in high availability features to maintain our environment through hardware failure or we can leverage high availability from the RDS environment for this.

In a small deployment, it would be most common to leverage the Scale HC3’s high availability features to maintain the availability of our RDS services.  Doing this allows us to run a single RDS instance with a minimum of licensing, resource and maintenance overhead.  This is the simplest approach and is very effective.  This is an excellent way for the majority of customers to make the best use of the Scale HC3 platform.  Scale can do the heavy lifting in this case and we are able to focus our efforts in places where they are more effective.

In large deployments having a single RDS instance may not be adequate.  At this point it would generally become reasonable to move to a multi-server RDS deployment with workloads load balanced across the instances.  Typically we would want to see no more than a single RDS server deployed per Scale HC3 node to best leverage the available resources.  In this way as many RDS server instances could be deployed as needed to handle the environmental capacity.

In most cases, even with the larger deployment across many nodes, we would use load balancing in front of the RDS farm, but we would still use the Scale HC3 cluster’s built in high availability features to handle hardware failures by moving running workloads from a failed node to an available node.  The load balancer would see the same instance with the same IP address allowing for a nearly transparent recovery that is fully automated and requiring a minimal of effort.

In more extreme cases where the full capacity of the cluster is required it is possible to configure one RDS instance per physical cluster node and to disable high availability features and instead utilize the load balancing functions to shift load to remaining nodes.  This would result in a graceful loss of performance rather than an outage.  Unless resources were extremely constrained already, user allocated CPU and memory would decline but system functionality would remain.  This is not as ideal a full high availability solution but can be a very functional alternative to the cost of hardware for that level of protection.

Microsoft Windows RDS and Scale make an obvious partnership.  Scale focuses on making the platform as easy and robust as possible while Microsoft’s Windows and RDS product provide us the simplest entry point into the shared computing space.  For small and medium businesses and especially for on premises deployments RDS is often the perfect choice for moving to centralized, thin client-based computing and can be a key enabler for workforce mobility, security and even bring your own device options.

[This piece was commissioned for the Scale Blog.]