Any pointers on who might be a good VC outfit for SaaS type startup that actually needs to run their own clusters as opposed to throwing everything onto AWS?
> Any pointers on who might be a good VC outfit for SaaS type startup that actually needs to run their own clusters as opposed to throwing everything onto AWS?
Very curious, why do you think you need to run your own clusters, to the point where you think you need to seek unusually risk prone capital?
If your justification is that you are are optimizing for cost (e.g. maximizing your performance per dollar), then you must think the capex dramatically lowers opex at scale if you move from AWS to specialized hardware (say, an FPGA bare metal cluster instead of a GPU or CPU cluster in AWS). In such a case I would argue that you should just get enough paying customers first using AWS, to build up enough cash-flow that would then clearly justify the capex for scaling up and going colo. And at that point, you would not need to shop around for wild risk taking capital, the CBA could easily be explained to any VC and in fact you could probably even just get a bank loan to finance the capex if it is a clear winner in reducing costs.
The only scenario I can see where you really need to run your own hardware from day 1 is if you need to terminate some dedicated circuits for specialized data feeds (e.g. real-time market data)... In that scenario, I would recommend looking into co-locating the data appliance in a third-party data center that is physically nearby to US-EAST and can run a Direct Connection into AWS for you.
Otherwise, you may want to consider the possibility that your desire to own your own cluster may be a case of premature optimization and scaling up too early/too fast. It's a very expensive thing to do and an easy way to accidentally run out of money if you don't have client contracts signed beforehand.
We connect storage nodes via infiniband for lower latency and higher throughput. Think RDS type service where you can have significantly larger instances like say up 40TB DB running on
4 way box (soon 8 way) box.
Couldn't you get at least similar storage performance from EBS Provisioned IOPS SSDs in AWS? And maybe "scale out," instead of "scaling up," to maintain performance within AWS constraints? The advantage obviously is that you avoid the capex and if you don't achieve the sales that you foretasted then you can just rapidly scale down your AWS footprint, thus avoiding financial catastrophe. You could spring for the fancy custom bare metal hardware when you reach a "hyperscale" amount of workload where it makes sense to switch to your own hardware, like Google or Facebook. And if your customers really want the fancy Infiniband clusters right now, why not license your software and make them put up the capital to buy the hardware themselves? Just my two cents, maybe I'm missing something and your approach makes sense.
Nope you really can't that's why RDS, Azure SQL etc. have pretty bad limitations on max instance sizes. The service itself is integrated storage layer (think PostgreSQL + ElasticSearch + Event Bus) that can scale far beyond what is possible with current options.
