RSS
Логотип
Баннер в шапке 1
Баннер в шапке 2
2020/06/02 18:31:47

Check list: What is important to consider when using object stores?

TAdviser continues a series of publications in a new format - "Check List." In these materials, we share useful application information containing tips and instructions on the use of various technologies. What is important to pay attention to when using S3 object repositories was helped to understand by Georgy Megrelishvili, Executive Director of SBC Laud

Content

Prior to use

Features of object repositories:

Object stores do not have a directory hierarchy. The data in them is stored as separate objects, hence the name. Each object has a unique address. Access to data is provided through the API, the API standard is de facto S3, although a number of providers have their own APIs.

Check list: what is important to consider when using object stores

Advantages of Object Storage: Unlimited expandability; Simple organized replication of data to different sites, which is often used as a content delivery network (CDN); Support of versivity of objects; Easy access to data, since the http (s) protocol is used, then compatibility issues with various OS or equipment do not arise: you can get data simply from the command line, from the script or from the browser; Simple migration, both between cloud provider storage and between in-house storage and cloud provider storage; An important advantage is the ability to add arbitrary metadata to each object, which makes the application architecture easier and its logic more transparent.

Disadvantages:

Low performance (still familiar to storage engineers). Compared to block or file stores, a separate operation will be performed for an extremely long time, both due to a sufficiently high delay in performing the operation and due to the inability to perform the operation on a part of the object. Impossible or unreasonable use of legacy applications already in use.

The most appropriate scenarios for using object stores are:

Storage of media content with access to it for a wide range of people; Storage of unstructured heterogeneous information; Long-Term Retention Archives

What object vaults are not at all suitable for

For classic relational databases.

Fear "novelty disease"

Object storage is a powerful, but by no means universal tool for solving storage problems associated with analytics, CDN and geo-distributed data security. As part of the development of your service, application or task, you should never forget about the scale factor. If yesterday's classic vertical scaling of storage and application servers met your needs, this does not mean that maintaining the existing architecture and the next step in vertical scaling will be the most profitable option.

CapEx в OpEx

Investing in your own IT infrastructure, including storage, is always a significant cost to your business that doesn't pay off quickly. You should also not forget the direct and indirect costs of operating and supporting all deployed systems. Now this is especially relevant in connection with foreign exchange risks.

File:Aquote1.png
Moving from CAPEX to OPEX - and object storage is primarily a service - significantly reduces costs. Firstly, you have to pay only for what is really necessary to support the business. Secondly, cooperation with the provider allows you to provide a reserve of resources that can be vital when peak loads arise or new business tasks arise. It is simply impossible to keep unlimited resources "in stock" in your own server. Thirdly, the transition to the cloud is the flexibility of consumption management: volumes can be increased or reduced, depending on the needs of the business. Object storage, of course, is not a panacea - but if used correctly, it can save significant funds to the business.
Georgy Megrelishvili, Executive Director of SBC
File:Aquote2.png

When selecting a supplier

What affects the cost of service

The cost of consuming the provider's storage is usually composed of the following parameters:

Traffic to storage. Note that some providers charge both inbound and outbound traffic, others only outbound, third - limit bandwidth, the fourth put their idea of ​ ​ the average use case by market in the price of storage. It is important to note that traffic between the provider cloud and its object store is usually free. Price of storage space. The amount will have to be paid inevitably, the price depends on the number of data replicas, the method of replication (copy of data or Erasure Coding), the number of data centers in which the data will be located, and the time of data access. For example, foreign providers for archive data offer repositories for archives, objects from which are uploaded to the level of operational interaction within a maximum of 12 hours. The intensity of interaction with the vault (the number of put, get, select, copy operations, etc.). Most operators pay for the number of operations, although there are exceptions.

Thus, not only the amount of data is important, but also the use case, so it is necessary to "understand" your application well for pricing. If you know in detail the model of interaction of your application with objects in the repository, then it is worth paying attention to the most detailed tariffs, where the fee is taken for all kinds of parameters, this is likely to be economically profitable. If you are only launching applications and services related to the use of object storage, on the contrary, it is worth considering the least detailed tariffs, paying attention to the provider's ability to provide detailed statistics.

File:Aquote1.png
SBClaud adheres to perhaps the simplest model: we charge only for the storage volume, without charging requests or traffic, "comments Georgy Megrelishvili. - So we don't have hidden payments or resource usage restrictions. At the same time, we help the customer's specialists with migration or configuration so that the effect of using our resources for business is maximum.
File:Aquote2.png

What Affects Performance

Performance is determined not only by the performance of the storage itself, but also by the bandwidth and latency of Internet access to it, if required. The COSBench and s3tester utilities are widely used for testing.

When access to storage data is assumed from a certain territory, it is important to understand that the laws of physics operate everywhere the same, so the closer the data is to the user, the faster it works.

Russian or foreign provider

When personal data is stored in the object store, keep in mind the requirements of the regulators. In other words, if you work with personal data, one of the most important qualities of your cloud - or the provider cloud with which you cooperate - should be compliance with the Russian 152FZ and other requirements of domestic regulators related to personal data. The ability of the provider to maintain the technical complexity and unique of your service is certainly also important.

During migration

Migration is often cost effective only for cloud native applications

Simple migration of existing information systems is only possible for a small list of tasks, so it is worth assessing the benefit of implementing cloud native applications in connection with unconditional cost-effectiveness of object storage.