Righting the Wrongs of Cloud Analytics
The cloud has led to a major shift in how we think about deploying technology, bringing agility that enables companies to jump on new opportunities as they arise and flex compute to meet varying demands.
Consumption-based or pay-per-query pricing at first glance seems to be a great fit to match cloud elasticity. As data warehouse workloads grow, across multiple teams and multiple locations and workloads become more complex, many organizations find that consumption models can become an innovation inhibitor – with teams actively scaling back analytics efforts to stay within budget. This is clearly not the desired outcome.
As analytics workloads grow, organizations need to be able to take advantage of additional efficiencies and economies of scale. Once you are locked into a consumption model, the vendor has little incentive to provide discounts or help you to reduce costs.
The Easy Button Works… Until It Doesn’t
As workloads grow or become more complex, consumption-based businesses are incentivized to help you achieve your goals by making it easier to increase your consumption. With Snowflake, for example, you can turn on auto-scale to add entire new clusters to your virtual data warehouse or turn on features like query acceleration to temporarily add more resources.
Need to make sure different teams don’t get in each other’s way? Sounds simple, add new virtual warehouse clusters. Moving teams to a new cluster adds complexity to existing reports and tasks need to be updated. Each new cluster or scale event results in cold caches which impact efficiency. The more scale-up or scale-down events, the higher the inefficiency.
Whichever way you turn, relying solely on the easy button will constantly be driving up cost. This quickly becomes unsustainable, particularly as data initiatives succeed in attracting new teams and workloads.
Yellowbrick is different. We give you the scale and multiple-cluster options for when you need to scale. However, we also allow you to apply common sense to let Yellowbrick know which teams and workloads are a higher priority for your business. We offer the flexibility to enable BI users to run tens of concurrent simpler queries with fewer resources assigned to maximize the interactive query performance, e.g., in Tableau or Power BI. Of course, you want to enable these scenarios without automatically driving up cost. At Yellowbrick, we call these Workload Management Rules – they can be as simple or complex as you need.
A little common sense goes a long way to improving user experience, resource consumption, and minimizing cost.
The Predictability of Fixed Subscription with the Flexibility of On-demand
Yellowbrick’s software pricing model is transparent and easy to understand. Subscriptions are available on 1- or 3- year terms based on the number of vCPUs you need. For on-demand pricing, you pay per vCPU hour you use billed by the second. If deploying in the cloud, infrastructure comes out of your normal cloud bill or if deploying our hardware in your data center, hardware is also available on a subscription billed per vCPU and per TB deployed.
Fixed subscriptions give budget certainty and provide value where you are using the system more than 20% of the time. The split between software and cloud infrastructure costs means that you can build elasticity into your fixed-cost subscription. Don’t need all your vCPUs? Reduce the number of nodes (and vCPUs) and immediately start saving on cloud infrastructure.
Buying Cloud Twice
You went to great lengths to negotiate a great deal with your cloud vendor, only to buy a SaaS DW platform that re-sells those very same cloud resources at a price they chose. They don’t even tell you what resources you are paying for.
Yellowbrick is different. We run our software in your cloud account, meaning you get the benefit of any negotiated price breaks. No intermediary marking up cloud resources and full transparency of the resources Yellowbrick uses. Use standard cloud provider reserved capacity discounts to further reduce costs.
Match Resources Needs More Closely to Demand
Some cloud data warehouse solutions require you to add an entire cluster at a time. The new cluster must be the same size, causing resources and costs to double, triple, or quadruple.
Another dimension vendors provide is to grow your cluster size. In Snowflake, for example, clusters are constrained to increase by a power of 2, i.e., 2x, 4x, 8x, etc. Synapse is constrained to cluster sizes that are factors of 60 and only a single cluster.
With Snowflake, the more complex your query, the bigger the cluster size. Each cluster can handle up to eight simple queries – the more concurrent queries you run, the more clusters you need. A moderately complex BI dashboard may issue 20 or 30 queries for one dashboard refresh for one user – eight gets you nowhere.
Serverless options may seem like a good way to match supply and demand. However, these resources come at a premium and take variable amounts of time to spin up. These use cases work well for occasional experiments, but they are terrible for interactive user experience and disastrous for cost management.
Yellowbrick enables you to finetune your clusters one node at a time to improve performance. Add in a little common sense with workload management and you can run hundreds of concurrent queries and loads in a single cluster.
SaaS’s Security Dilemma
Using a SaaS or PaaS platform usually requires giving data to a third-party provider, or at the very least allowing them to broker access to data. This requires you to rely on promises of security and access from your platform providers, delivered through compliance certification and third-party audits.
Yellowbrick is different. The entire Yellowbrick platform runs in your cloud account, eliminating third-party access concerns and hugely simplifying integration into your network and compliance with internal InfoSec teams. All this removes implementation bottlenecks which often delay enterprise data platform deployments by months.
The simplicity of consumption-based pricing and capacity provisioning drives ultimate agility but comes at a significant cost premium. Capping your consumption spend means capping your analytics ambitions. Fixed subscription models with a little common sense applied through workload management can help your organization benefit from economies of scale and keep different teams from getting in each other’s way.
At Yellowbrick, our friendly customer success teams help you to actively make the most of existing capacity and optimize the performance and cost of your data warehousing services. Get in touch with us today and we’ll help you understand how Yellowbrick’s approach translates into exceptional user experience, deeper business insights, architectural simplicity, and lower analytics costs.