{"id":6422,"date":"2020-11-17T07:00:27","date_gmt":"2020-11-17T06:00:27","guid":{"rendered":"https:\/\/lcloud.pl\/?p=6422"},"modified":"2024-12-10T12:36:07","modified_gmt":"2024-12-10T11:36:07","slug":"cost-optimization-in-google-cloud","status":"publish","type":"post","link":"https:\/\/lcloud.pl\/en\/cost-optimization-in-google-cloud\/","title":{"rendered":"Cost optimization in Google Cloud"},"content":{"rendered":"

Managing expenses in the cloud is one of the most important issues when designing a cloud solution. The application of appropriate principles of good design will allow not only to ensure high availability of the solution but also to meet the demand and optimize the control of operating costs of the environment.<\/span><\/h4>\n

Monitoring costs is not an easy task, even if it seems so obvious that we are paying for cloud resources available on demand. To properly optimize your expenses, you first need to understand the essence of cost generation. For this, cost management and accounting tools<\/a><\/span> will be useful. Billing reports<\/a><\/span> that give you an insight into costs are also useful.<\/span> An important activity is the appropriate tagging<\/a><\/span> of assigned costs for individual departments. The use of labels allows you to create personalized dashboards<\/a><\/span> with the necessary information. You can also use quota<\/a><\/span>, budgets and alerts<\/a><\/span> to accurately control costs. They make it possible to monitor trends and forecast them over time, taking into account possible differences in the budget.<\/span>\u00a0<\/span><\/span><\/p>\n

\"\"<\/p>\n

Once you have knowledge of the tools, the next step is to identify resources that are not delivering business value to your organization.\u00a0<\/span><\/p>\n

The first thing that will lower costs will be identifying idle virtual machines (VMs) and disks<\/span><\/strong> and cutting back on resources that are not being used. To this end, Recommender<\/a><\/span> was created, which identifies inactive virtual machines (VMs) and hard disks based on usage rates. Keep in mind the consequences of removing resources; deleting an instance removes the underlying disk (s) and all of its data. It is good practice to snapshot the instance and the alternative is only to stop the virtual machine, which in turn will save resources.<\/span><\/span><\/p>\n

The second consideration is scheduling the startup and shutdown of virtual machines<\/span><\/strong>. Production systems usually run 24\/7. But virtual machines in development, test or personal environments are usually only used during working hours. Therefore, their exclusion may contribute to considerable savings. A solution specially prepared by GCP Team, based on the Serverless technology<\/a><\/span>, will be useful here, which will allow automating the process of automatic shutdown of virtual and large-scale management machines.<\/span><\/span><\/p>\n

The third issue is the appropriate type (size) of virtual machines (VMs)<\/span><\/strong> – the so-called rightsizing. Custom Machine Types<\/a><\/span> is a functionality of the Compute Engine<\/a><\/span> service that allows you to easily create machine types tailored to your current needs. One last issue related to virtual machines; you should also consider instance options, which operate on a 24-hour basis and are generally up to 80% cheaper than long-term solutions. Preemptible VMs<\/span><\/strong> are suitable for fault-tolerant workloads such as big data, genomics, media transcoding, financial modelling, and simulation. Thanks to the appropriate plan and the Flexible Resource Scheduling<\/a><\/span> tool, you can reduce the costs of stream and batch analysis.<\/span><\/span><\/p>\n

\"\"<\/p>\n

Storage, data storage, and cost management can all disappear into the total cost of infrastructure management. However, in the cloud, where storage is billed as a separate item, paying attention to its usage and configuration can result in significant savings. <\/span><\/p>\n

Here are 3 rules that can help you save space, time, and storage costs.\u00a0<\/span><\/p>\n

    \n
  1. Cloud Storage classes.<\/span><\/b> Each of the available classes has a specific use and price list. Appropriate use of the class for its intended purpose allows you to optimize expenses. By using, for example, the coldline<\/a> <\/span>or archive type<\/a><\/span>, you can significantly reduce the cost of storing data, access to which is not needed often<\/span>.<\/span><\/span><\/li>\n
  2. Lifecycle policies.<\/b><\/span> Appropriate use of classes in conjunction with the automation of facility lifecycle management can reduce the costs of data archiving. Object lifecycle configuration should be adapted to the object’s storage class, based on a\u00a0set of conditions<\/a><\/span>, or include its complete removal if not needed.<\/span><\/span><\/span><\/li>\n
  3. De-duplication<\/b><\/span>. Another cost-generating issue is the duplication of stored data. Of course, there are situations when duplication is required, e.g. by a team working in different regions, who needs immediate access to data. So it makes sense to create a multi-region bucket<\/a><\/span>, rather than having multiple copies in different containers. You should also read about object versioning<\/a><\/span> in Cloud Storage. Appropriate configuration prevents overwriting or accidental deletion of data.\u00a0<\/span><\/span><\/li>\n<\/ol>\n

    Another useful service when optimizing costs is BigQuery<\/span><\/strong>, which allows for a modern approach to data analysis and appropriate configuration, and as a result allows you to reduce costs. The first point is query control<\/span><\/em>, which will allow you to monitor your accruals. For this, it will be useful to set the limits (maximum number of bytes)<\/a><\/span>. This will allow for supervision, and exceeding the set threshold will not cause additional costs – as it is shown in the example below:<\/span><\/span><\/p>\n

    \"\"<\/a><\/span><\/p>\n

    It’s also worth taking advantage of table partitioning and clustering. Partitioning allows a table to be partitioned based on: processing time, date, time stamp, or a column of an integer range. By using the Require partition<\/a><\/span><\/em> filter, you can make sure that your queries and tasks use partitioned tables. An additional advantage of using partitions in BigQuery is the ability to reduce data storage costs by 50% for each partition or table. Non-editable tables are moved to long-term memory<\/a><\/span><\/em>, and simply moving a table or partition has no effect on performance, durability, availability, or degradation of functionality. At this point, let’s pay attention to data transfer\/upload<\/span><\/em>. Data can be uploaded to the BigQuery service in two ways:<\/span><\/span><\/p>\n