13 ways to reduce the cost of cloud computing

Cloud computing was originally an economic proposition for enterprises . Servers were expensive at the time , The large number of Internet connections required for data centers to flourish are also expensive . Why not rent directly and stop paying high fees for hardware and employees ? Operating a large cloud data center can save a lot of money .

Today, more than ten years later , The cost of cloud computing is beginning to increase . It's easy for businesses to spend millions of dollars on smaller instances .

Here are some ways to reduce the cost of Cloud Computing , Many methods involve improving communication , For developers and devops The team can make more informed financial decisions . But none is perfect , All these methods involve trade-offs . Although no one method can be applied to all situations , But these methods are worth considering .

1. Turn to commercial cloud providers

Large cloud computing companies offer a range of options , Including some first-class services , These services are suitable for ordinary work such as data storage , It is also suitable for complex work such as machine learning or computer vision . however , If you're not going to use all the best API And the service , You can find commercial cloud providers , They more or less sell the same product or service at a lower price .Linode、DigitalOcean、Vultr、Hetzner、Scaleway and LightSail And other companies may offer cheaper , Especially for simple loads .

2. Find major cloud discounts

Users can take advantage of the price discounts offered by cloud computing providers . for example ,AWS The company offers discounts for long-term commitments by reserving examples . When prices fall , Their spot market may be a good place to run background processing . Google cloud platform offers automatic discounts , The longer the user's service runs , The higher the discount . The price is set for enterprises that want to meet and start machines immediately . Planning ahead and looking for discount options is one of the easiest ways to reduce the cost of major Cloud Computing .

3. Using a hybrid cloud

Not every enterprise has standby servers , But companies that own them can use increasingly complex software packages to build secure subnets that connect them together . This can reduce the cost , Maybe even safer .

Tailscale、Submariner and Traefik Just some options for building this secure grid using public and private clouds . Some users are building Kubernetes colony , Other users just host back-end services that do not require low latency . Although this is not a solution for customer workloads , But combining some useful background processing tools is easier than ever .

4. Explore the use of bare metal servers

Not everyone needs the flexibility of modern cloud instances . Some vendors are renting servers , Without the need for additional hypervisors and virtualization layers . Enterprises can use bare metal machines , It also reduces security concerns , Will get root password . Enterprises may need to tidy up the operating system (OS) Do more work on layers , But sometimes this brings the opportunity to better adjust the machine .IBM、OVHcloud、GTHost and Vultr It is one of many companies renting bare metal servers .

5. Break down the bill and pay attention to each part

Some cloud computing users crash when they receive monthly bills , Because their applications target a large number of data users , Some cloud platforms charge for data leakage , This may increase costs , Especially in the case of enterprise data leakage .Cloudflare And other companies are building networks that don't charge penetration fees . therefore , Enterprises take a closer look at the configuration files of their applications . It's the Internet 、 Users of a particular item on a data store or bill ? Other cloud computing providers are likely to provide better answers .

6. Stop paying extra for data storage

Backblaze or Wasabi And a few other cloud computing providers focus on providing data storage at a low price , It charges less than major cloud storage providers 80%. The cost savings may also be more , Because they have different data mobility cost structures , So that enterprises can save costs . These providers also provide services with Amazon S3 Compatible API, Make switching simple .

7. Use static web pages

If an enterprise's application needs a lot of constant information , Then an application built from a static web page may be the least expensive way to reach a wide audience through a content distribution network . For tasks that require a lot of back-end computing and customization , They are not a good solution , But maybe you can split the process into a single service . Good static frameworks include Jekyll、Gatsby、Middleman and Hugo. When programmers can transfer all content to CDN when , May significantly reduce costs .

8. No server is adopted

Cloud computing is usually better suited for applications with surging demand . If an enterprise's applications are idle for hours or even days , Then rewrite it as in Cloudflare Workers or AWS Lambda Running in a serverless framework , This will mean no payment for all downtime . When the load is low , It won't pay so much .

9. by Arm recompile

AWS And some other cloud computing providers are adding around Arm CPU Built server . these RISC The chip runs faster 、 Cheaper , But they usually need to recompile binary applications . The cost savings depend on the type of application and the load on the machine . for example ,AWS The company estimates that Graviton The chip may be more efficient than x86 Chip high 25% To 100% Even higher . because AWS The company has just launched a new Graviton3 processor , Therefore, the processing speed may increase , The company estimates that the processor may be faster than Graviton2 Three times faster , At least based on bfloat16 The value of machine learning application benchmarking is like this .

Unless the enterprise's servers are running at full capacity , Otherwise, it may not be easy to capture this value , This makes some developers nervous . One solution is to start using Arm The chip performs background processing . Another way is to turn to providing ARM Event driven computing services for processors ( for example AWS Lambda), Programmers have found that this solution may save money 20% Cost of , But only for certain languages and Applications .

10. Adopt micro server

Arm The number of options is growing rapidly , Sometimes it even grows in amazing ways . Many people use small machines with little cost Raspberry Pi or Jetson Machine hacking . They are building embedded machines 、 Internet of things applications or just ideal for experimentation . They can now be part of the enterprise's cloud computing infrastructure . for example ,Raspberry Hosting and miniNodes Will rent a small enterprise in a professional data center Raspberry Pi or Jetson Nano, And there are a lot of Internet connections , In this way, basic software can be run at a low price .

11. Observe the default configuration

The default size of many cloud computing instances makes their configuration easier . In many cases , Enterprise code doesn't need all the storage space , But we need to continue to pay for it . What's worse is , Cloud computing providers often have difficulty narrowing this unused space . If you remain vigilant during the creation process , You can usually have enough space to run their applications .

12. Let everyone know the cost

Some enterprise developers have never seen any costs , because devops The team dealt with all these details .

Cost sharing helps engineers understand how to save costs , When certain Kubernetes When the cluster expands to take on some load , Will rapidly increase costs . It may not be easy for developers to reduce costs , But if they know the cost , They are less likely to use excess cloud computing resources .

13. Hire a cost engineer

Some teams designate some developers as “ Cost engineer ”, Their job is to find ways to save the cost of Cloud Computing . Maybe this means adjusting some microservices to use less memory . Perhaps this means moving some occasional tasks to a serverless configuration . Let developers know that their work indicators will be associated with cloud computing , This is a good way to shift the focus to cost savings .