Densify is an analytical service company whose mission is to help organisations move their server workloads to the public cloud and to optimise workloads once they are there. While the name is new, the company has been around (previously known as Cirba) since 1999 — that is, back to times when cloud was a twinkle in the eye of ASPs. Densify operates a subscription model, charged on a per-managed-virtual-instance basis.
In terms of the offering, Densify has data gathering and analytics tool known as Cloe — “the Cloud-Learning Optimization Engine”. Cloe outputs information about specific workloads, their current performance, and how they might be better sized, scaled and priced in a cloud environment. In other words, Cloe automatically balances application demand with cloud supply. Cloe’s outputs are interpreted by Densify experts — called “densification advisors” — who work with customers to set priorities and propose alternative targets for the workloads concerned.
Given that I’ve long taken the stance that no wholesale “move to cloud” exists, I was keen to catch up with Ayman Gabarin, EMEA VP and find out more about the Densify approach. Note that I am taking Densify’s approach at face value — that is, I am thinking about the challenge it seeks to address, whether it is good at what it does or not (feel free to check out Densify customer stories).
1. Why isn’t it as simple as “moving to the cloud?”
The answer is that there is no one size fits all. When instances are defined they are often over-specified, hindsight can find them too big or too small for the workloads. This is hard to get right — for example, AWS has over three million combinations of infrastructure and services so it is no surprise that customers err on the side of caution.
If they have a large virtual machine and they want to move it into the cloud environment, they will probably select an M3 instance, but can then find they are only using it for two hours a day, but still pay for capacity they are not using. The result is lots of potential dead wood.
The challenge is exacerbated by use of reserved instances, i.e. instances that are locked down over a year or more. These offer guarantees that prices will not go up, but at the same time, customers can find they are locked into costly, over-provisioned infrastructure.
2. So what does Densify bring to the party?
We offer technical, business and governance analysis. Densify’s heritage is in understanding server-based physical and virtual environments — we brought in analytics to help reduce server sprawl. Indeed, we still offer on-premise VMware services but our focus has moved to the growing challenge of cloud environments.
Cloe is our machine learning engine for optimising hybrid environments. We can plug Cloe into workloads running on-premise or in AWS/Azure/Google environments and then establish how the workloads might be better provisioned and run. Customers can see up to 80% cost savings on their public cloud bill based on our services, so ROI is almost immediate.
3. 80% — that’s quite a figure. How does it break down?
We tend to focus on workloads that are prime candidates for moving to the public cloud, and workloads that have been moved to the cloud but are proving more expensive than expected. We work at a number of levels:
– at the application level, we can deliver 15-20% of savings with basic analysis of the application and its needs.
– at the machine level, an additional 30% can be generated through resizing instances by policy.
– at the architectural level, we can deliver additional savings from database resizing, optimisation of reserved instances and containers
4. Cloe’s about machine learning — why are people needed?
Customers have a business to run – as they move to the public cloud, they tend to move to smaller teams with less expertise. The public cloud is so complex, it’s not simply a question of pushing a button and migrating from one environment to another.
What we offer is a kind of concierge service. Our experts can interpret the results generated by Cloe, but as importantly, can help our customers make the complex decisions they need to make around strategy, prioritisation and forward planning. It’s a bespoke service as every customer has unique challenges and needs.
5. Why a subscription and not a one-off payment?
We have always offered our services on a subscription basis. Environments and needs continue to change, so customers are OK to carry on paying a subscription as long as our services are useful to them.
Densify enables your applications to become self-aware of their resource requirements and to dynamically match their needs to optimal cloud supply. As this changes over time, so do our recommendations to ensure that they keep costs in check and maximise efficiency.
My take: Densify is a sign of the times
Whatever happened to the notion of elasticity? One of the most significant attractions of the cloud model was that organisations would only pay for what they needed, potentially down to a CPU cycle basis. In experimental situations this remains true — researchers and analysts can spin up a thousand servers, do a job then spin them down again, potentially paying only tens of pounds for the privilege.
But for ongoing workloads, this model has been superseded. A virtual machine running in the cloud across a longer period will incur costs for just being there. Mundane, complex reality kicks in: poorly (in hindsight) architected applications will still burn CPU, even when doing nothing; and will have been procured on a by-instance basis, sometimes (in the case of reserved instances) trading off higher costs against longer-term stability.
And it’s not in the interests of cloud providers to question these choices. Companies like Densify occupy a space that nobody wanted to exist: poor performance of virtualised workloads in what was originally sold as an elastic, pay-per-use environment. But exist it does. Densify’s service is predicated on the need to have assistance in cloud sizing. In principle, this should be a non-problem but in practice it is a challenge, and the company will be there as long as it is.
While cloud providers could do more to keep their customers’ costs down, this is not only a challenge for cloud-based environments; more, it comes from a failure to close the loop between those defining systems and those operating them. Another big idea of the times – DevOps – is predicated on the notion that developers and operational management can work together on creating optimised services for their users: maximum benefit at minimum cost.
Densify’s current raison d’être comes from a failure of this closed loop, as much as the company’s future perhaps depends on moving it from theory to reality.