Are startups being drawn into the allure of Kubernetes and hype surrounding server-less?
Total cloud 'elasticity' (or 'serverless' as it's commonly being referred to) is very young, and can come at immense costs to a startup. Counter intuitively exacerbating the very problems it promises to solve; Increasing complexity, cost of ownership, as well as reducing time to market. If you don't have the right skills inhouse, and that's all but the very largest tech-firms, these costs quickly swamp any promised 'cloud savings' of per second billing
.
Make sure you are making the decisions for the right reasons. Don't end having to hire a £150K a year devops engineer, or extremely rare full stack engineers, just to avoid keeping 20 cheap servers (docker VMs) running at a cost of £5 a month each.
Advice for startups
Not running on Kubernetes does not automatically mean you cannot elastically scale. It may simply mean you may have to manually log in to a control panel, or run a prepared script to upgrade or downgrade the number of servers allocated to handling load for you.
(Put an entry in a diary to manually downgrade the number of servers after BlackFriday.)
It's definitely not as sexy as Kubernetes, however, investors are not going to value your business more highly because you have the worlds most complex CICD processes.
It's much cheaper to set configurable alerts to notify the right team when thresholds are approaching
Instead of embracing complexities associated with Kubernetes, encourage a culture of simplicity, and rather design the services so that if load spikes beyond your budget to handle, that non critical services throttle themselves, leaving critical services operational
If you're a startup, then by the time your startup has the problem of how to handle more load than some well engineered and simple (uncomplicated services) running on cheap servers can handle then you should already have exited with a trade sale. And if your challenger startup is disruptive
and your goal happens to be to sell to one of the market leaders who do care about running on Kubernetes at scale, then they can use their experts to migrate your system to be cloud native. Having a simple set of message driven services will allow any acquirer to easily integrate your systems into their platform.
Deliberately shape your business so that it's survival is not tied to any 1 founder
This is relevant to Kubernetes
; since if you're the technical co-founder or CTO, then you need to make sure you're not using Kubenetes simply because you have a wealth of Kubernetes skills due to your background. And if you are a non-technical founder, you need to watch out for freelancers or sub-contractors from using overly complicated and expensive tech for similar reasons.
If you're a business founder, and you're one of the few lucky founders to have a technical founder WITH kubernetes skills, then perhaps this advice doesn't apply to you? But in my experience, that's extrelely rare. Make sure you deliberately shape your business so that it's survival is not tied to any 1 founder.
Acquirers will apply a steep discount to the value of any business with key-man risks that are costly or time consuming to address. Whether they CAN address it or not, is irellevant, they'll simply use it to force you to accept a significantly lower valuation; just because they can.
What I'm really talking about is just keeping your tech simple. Yes, Kubernetes is sexy, and comes with many promised benefits but with significant trade-offs and risks for non big-enterprise with dedicated teams with specialist skills.
Investors care about product market fit
Optimise your business and technology for speed to market, simplicity and reliability, keeping team size as small as possible.
The simpler your technology choices and architecture, the easier it is to guarantee low MTTRs.
Mean time to repair. The average time required to repair a failed component or service.
By focusing on delighting users and fast feedback cycles, (Rapid high quality feature releases) a startup will reduce the time it needs to find product market fit, and increase it's chances of successful exit or acquisition.
Start small, grow fast, bootstrap from profits.
Want to chat further?
Book some free time with me over a coffee, I'm always happy to talk to new clients looking for full or part time help.
Resources
If your team is seriously considering Kubernetes because you're currently overwhelmed with all the micro-services
that make up your offering, then chances are you might be able to simplify your stack by making use of some of the new services targeting Saas businesses.
Back end as a Service
| <a href=https://directus.io>directus.io</a> | Total back end as a service provider |
|---|---|
| <a href=https://temporal.io>temporal.io</a> | Open source distributed transactions banking and finance services in a box. |
| <a href='https://knock.app'>knock.app</a> | PAYG push notifications |
| <a href='https://appwrite.io'>Appwrite.io</a> | Back end as a service - OAuth - S3 storage API - database |
| <a href='https://render.com'>render.com</a> | Define all your services and dependencies in yaml and under source control. |
| <a href='https:/supabase.com'>Supabase</a> | Open source Firebase alternative |
| <a href='https:/upstash.com'>UpStash.IO</a> | Serverless Kafka - PAYG queues - eventing. |
Low Code (locode)
| <a href='https:/airtable.com'>Airtable | Low‒code platform to build next‒gen apps. Use sparingly - Gets expensive with larger datasets. Excellent for stage before product market fit or before scaling. Only replace airtable when you need to scale. Alternatively keep track of when you approach any limits and plan to replace critical scale tables with self managed services one at a time. (Just in time). |
|---|---|
| <a href='https:/budibase.com'>Budibase</a> | RAD tool for creating Enterprise and SaaS spike or POC projects - forms - custom workflows etc. |
| <a href='https:/retool.com'>AWS Retool</a> | Suitable for creating complex internal tools and dashboards. Retool is often used to build CRUD (Create - Read - Update - Delete) apps quickly. |
If there's something obvious missing from this list, please ping me and email or message on LinkedIn or Whatsapp and I'll add it to the list.