4 ms·
Azure, EC2, and Google Cloud are overpriced in most cases. You can do the same with cheaper (and often similarly reliable) VPSes managed with tools like Puppet/
by api 11y ago
Azure, EC2, and Google Cloud are overpriced in most cases. You can do the same with cheaper (and often similarly reliable) VPSes managed with tools like Puppet/Chef, Consul, Vault, Docker, etc. Plus you avoid lock in. Your stuff is yours and can be deployed anywhere.
I don't see the allure of the costlier cloud other than the "nobody ever got fired for" factor so common in enterprise purchasing. Amazon is the only one that might have a stronger case for it based on its huge managed service stack, but much of that is not too terribly hard to duplicate with other tools and more a la carte services. There also really isn't a reason you can't use some of Amazon's stuff while also using more commodity options.
On a more principled level I'm starting to see huge proprietary cloud as a potential threat to the open Internet. It's not quite there yet but at some point I could see it, especially with the walled garden plays you see around IoT.
- patio11 11y agoOne is welcome to use staff that cost $20k per month (e.g. DevOps engineers who understand those four technologies well enough to use them in production) to shave ~50% off one's AWS bill, but one needs minimally two to three of them, so your friendly neighborhood insurance company should probably pay their $15k or whatever a month without blinking.
- api 11y agoIt's not that hard, and if you are so huge that your devops takes a team you have a good problem. If you are a startup then architect your software so it can be scaled but otherwise just stick it somewhere and worry about product market fit. You can decompose and refactor and distribute once your product has enough users for it to matter. A lot of the difficulty also comes from over engineering and premature scalability obsession. You often just don't need all that. I swear over engineering is the bane of software and devops these days. We've gone from java factoryfactorysingletons to "how many distributed systems fads can I use in one stack?"
- AdieuToLogic 11y agoIn many (perhaps most) areas of the US, DevOps staff do not cost $20k per month. For example, 92% of Boston rates are less than half that[1]. This implies that a $40k per month bill for AWS[2] could pay for three DevOps engineers and save approximately $10k per month in the vast majority of the US. 1 - http://www.indeed.com/q-Devops-Engineer-l-Boston,-MA-jobs.html http://www.indeed.com/q-Devops-Engineer-l-Boston,-MA-jobs.ht... 2 - derived from your statement that a staff of $20k per month would save "~50% off one's AWS bill"
- kasey_junk 11y agoNot to quible, but fully loaded costs and salary costs can be dramatically different.
- AdieuToLogic 11y agoQuite true regarding company (fully loaded) costs. IMHO, a reasonable estimation for fully loaded cost per employee (excluding facilities expendetures) is approximately 1.4 * ES, where "ES" is the employee salary. The "three DevOps engineers and save $10k" estimation was based on working backward from the 92% of available jobs in Boston being less than half of the stated $20k per month cost. Assuming a Gaussian distribution where 0.5 * $20k per month represents the high end of two standard deviations (since Boston ranks quite highly in S/W salary nationally), most DevOps engineers will be paid roughly half of that as well. This yielded an estimation of $6.5k per month per DevOps employee or $19.5k per month for three. Since all of this was off-the-cuff, I figured it best to throw in a bit of "fudge factor" and present a $10k per month savings. As always, YMMV and I could be completely wrong about all of this :-).
- vgt 11y agoOne unique thing about Google Cloud is that most managed services like Load Balancer, PubSub, Datastore, BigQuery etc do not charge you for variability and high-availability. AWS and Azure WILL charge you 10x to scale up and another 3x for redundancy. Because Google's managed services are often based on Google's internal stack, they just scale. Good luck scaling Kafka to millions of messages per second - with PubSub you get it out of the box. PubSub, BigQuery, and others are geographically high-available out of the box. These things are difficult to replace on EC2, and nearly impossible on players like Digital Ocean. Edit: BigQuery, for example, allows you to rent 10,000 cores for 5 seconds at a time. This type of stuff is impossible to do with VMs at all.
- kasey_junk 11y agoNot really related to your bigger point, which I have no opinion on, but Kafka & PubSub have different delivery contracts, Kafka's are generally more strict. Therefore comparing the scalability of the 2 is somewhat problematic.
- vgt 11y agoCan you elaborate on that? PubSub is a fully-managed service, which means that Google SREs are on call making sure things are up. In addition, Pubsub has "guaranteed at-least-once message delivery". In a sense, Google's SREs guarantee delivery. PubSub is also a GLOBAL service. Not only are you protected from zone downtime, you are protected from regional downtime. Is there an equivalent to this level of service anywhere in the world? I'm not too familiar with Kafka's fully managed service, but Kafka-on-VM is a whole other ball game. YOU manage the service. YOU guarantee delivery, not Kafka.
- kasey_junk 11y agoKafka promises strictly ordered delivery, PubSub promises mostly ordered. The differences between those promises are what drive PubSubs ability to scale throughput and global availability. From an availability standpoint, I don't disagree with anything you mention, but the difference between the consistency models means that PubSub is solving a different set of problems than Kafka, thus my opinion that comparing them is problematic.