Dear Evan Spiegel, Bobby Murphy, and whoever manages Snap’s cloud infrastructure,
We have a proposition for you. We can save you $80 million on your cloud bills.
See, when you filed for IPO a few weeks back, one snippet of information that caught our eye was your use of public cloud – specifically your $400-million-annual four-year public cloud deal with Google Cloud Platform. We never doubted that your cloud infrastructure would be huge, see. After all, saw Netflix’s cloud spend rise to some $800 million a year in 2016 after they completed a near-total migration to AWS. These huge infrastructures are the most important to optimize – particularly as you grow.
As cloud waste reduction engineers here at ParkMyCloud, we are passionate about automating that optimization – and doing so quickly and simply. Obviously, in your case, you will continue to scale your infrastructure to deal with the exponential customer growth and daily peaks in usage (and by the way, my kids love Snapchat and I had to get an unlimited data plan – thanks!). Based on our analysis, we know that the largest item on public cloud customers’ monthly bills is compute instances/VMs (typically about 70% of a cloud bill). Research has shown that industry-wide, the amount of non-production in this infrastructure is about 44%. These non-production instances are the number one place to start hunting for optimization opportunities.
Your rate of innovation is certainly impressive and we see that you spent some $185MM last year on R&D. We would be willing to bet that although your teams are incredible, that in the haste to deliver better and better product, your cloud waste is likely enormous. Based on what we have seen in similar set-ups, we think we can save close to half this spend by simply automating the turning off of your cloud instances when not being used.
Here’s what we propose: you need to put parking schedules in place on your non-production instances. Snap, you need to ensure that your public cloud resources are only used when needed and turned off when not. Based on this alone we typically see our customers saving upwards of 65% off their compute spend. If you add additional optimization approaches that address industry rates of over-provisioning of compute instances (55%) and large scale inventory waste (15%) i.e. spend on resources that are no longer required savings grow even further, you will save even more on non-production (dev, test, QA etc.) workloads.
So when we see huge monthly spend numbers like yours, what gets us excited is thinking about how just how big your savings could be. And the truly wonderful thing with these type of savings is that everyone’s a winner – the DevOps team wins as they help the enterprise deliver more for less, the shareholders benefit from reduced Op-Ex and increased profits. (You probably don’t care, but your cloud providers will also benefit as they can better utilize their own datacenters.)
So Evan Spiegel, Bobby Murphy, and the rest of the team – shoot us a note. We are happy to talk whenever you are.
Jay Chapel & the ParkMyCloud Team