[Translation] 10 commandments that storage manufacturers want to hide
Propaganda still drives the data mining systems market and vendors create numerous myths that the client needs to be able to ignore in order not to be trapped with a system that does not meet the requirements of the business.
There are a lot of things that storage vendors want to believe in customers, but there are some that no one wants to talk about. These are the ten commandments that you want to forget about.
A few days ago I read with great interest the theses that Ben Rossi published in one of the blogs (the original ). Some of them seemed prejudiced to me, some were fair, and some I was not ready to accept at all. But all of them, of course, deserve to think at least once. Below is my free translation.
1. Five-year warranty should be standard
Most storage providers include a 12-month warranty and may increase it to three years, if you put a little pressure on them. But it is worth talking about a five-year warranty, as the price of the solution will grow very significantly. Why? The reason is that the manufacturer would prefer that you plan to replace equipment every 3 years. Perhaps, following Moore's law, it is worth changing servers at such intervals, but this is not the case for storage systems. Storage should not only live 5 years, but also meet the performance requirements all this time - this is the basis for dialogue with the manufacturer.
')
2. You must be able to use all the capacity of the system you paid for.
As soon as we talk about any other product, no one wants to use only a part of the purchased goods. If you bought a house with five bedrooms, then you will want to be able to use all five times. If you bought a car with five seats, then you plan that you can put 5 people in it. If you buy a storage system, then you should be able to use 100% capacity without sacrificing performance.
Nevertheless, many suppliers accompany their product with a special “best practice” document, in which you are warned of a significant decrease in performance if you use more than 75% of the available volume. It is possible to create a storage system that will work when using the entire volume without loss of performance. But why aren't most existing architectures designed for this?
3. The upgrade should not be difficult and too expensive
Large-scale upgrades or changes in part of the existing infrastructure often develop into a complex and long-term project, which is accompanied by downtime, investment in new equipment, and a decent bill for work. Online migration is quite possible, but with a true scalable storage architecture and it can be a thing of the past.
4. The storage system does not need maintenance personnel.
It is known that the human factor is most often the main cause of failures in data centers. If we can limit the interaction of people and technology, making the equipment maintainable at the place of operation, we will significantly limit the possibility of failure in the data center. About 70% of the returned warranty discs do not have defet and a simple rebuild was enough to restore the system. The goal is to find such a storage provider who can repair and replace disks on site, which will allow your own maintenance staff not to approach the storage system for at least five years.
5. When it comes to disks, quality matters
Whether we choose hard drives or solid-state drives, the general rule works — beware of consumer-grade equipment. There is a huge difference in the quality of the components used, testing, and most importantly, in contrast to the percentage of failures for regular disks and corporate class disks.
Conventional “consumer” drives can be noticeably cheaper, but the operating costs and risks are likely to be much higher than for a reliable enterprise-level product. When choosing discs, make sure that the upcoming capital costs, operating costs and risks are balanced.
6. Flash is not a panacea
Flash drives are not the solution to any problems. Yes, they can be used to significantly improve application performance for a particular type of load, but this technology has its limitations. For example, if we are talking about sequential recording of large amounts of data, then hard drives are much better suited. Flash drive and hard disk are two different tools that can and should be used for relevant tasks.
In order to achieve a balance in performance, it is worth talking to vendors that do not limit the client to one type of disk.
7. All-Flash arrays are not always the most energy efficient.
Many people believe that storage systems that use only Flash are more energy efficient than hard drive systems. However, in order to obtain a complete picture, it is necessary, in evaluating energy consumption, to sum up the consumption of all modules, and not just disk shelves.
It turns out that some systems with hard disks consume less power than other All-Flash systems. The decisive influence is how exactly vendors use energy-hungry components in their systems, such as processors and cache memory.
8. Vibration can reduce the estimated performance.
Strong vibration can cause problems not only with reliability, but also seriously affect performance, although many do not know about it.
Although it is quite possible to obtain predictable performance results by eliminating vibration, it is not so easy and requires significant investment in the development of such a storage system. It sounds crazy, but the easiest way to check the system is to scream at it well and see how the performance changes at this time. You may be greatly surprised by the result.
9. You do not need all these bells and whistles.
The choice of corporate data storage systems is often made on the basis of a long list of capabilities, and not at all on reliability and performance indicators. With the advent of SDS (software-defined storage), a significant part of the functionality implemented at the array level is no longer so necessary, as it passes into the competence of operating systems, hypervisors and applications.
For many customers, the acquisition of an enterprise storage SHD ends in “slavery” from a supplier of proprietary software, which is supported on one single platform. The implementation of SDS provides significant flexibility in choosing a platform, without thinking about the hardware that is actually used - it strengthens your position as a customer and allows you to focus on performance and reliability.
10. Scalable systems have their own characteristics.
Many manufacturers offer scalable "up" storage systems, but most of them do not explain that for performance-demanding applications, much depends on the central controller in the system. Often this can cause a significant investment in the upgrade to increase the capacity or performance of the system.
However, there is an alternative to traditional vertical scaling. Scale-out systems can be combined into a single storage pool, combining both large capacity and high performance.
Instead of performance and reliability, the storage market is now, unfortunately, largely based on idle talk and abstruse terms. What customers really need is an honest, no-fuss conversation, in which the vendor helps them choose the best system for their business. So far there are few such vendors, but they still exist.
PS I have a certain suspicion that everything was started for the last two or three paragraphs, but I sincerely want to believe in the author's best intentions. In addition, some of the statements are largely fair and really worth paying attention to when choosing a storage system.
There are a lot of things that storage vendors want to believe in customers, but there are some that no one wants to talk about. These are the ten commandments that you want to forget about.
A few days ago I read with great interest the theses that Ben Rossi published in one of the blogs (the original ). Some of them seemed prejudiced to me, some were fair, and some I was not ready to accept at all. But all of them, of course, deserve to think at least once. Below is my free translation.
1. Five-year warranty should be standard
Most storage providers include a 12-month warranty and may increase it to three years, if you put a little pressure on them. But it is worth talking about a five-year warranty, as the price of the solution will grow very significantly. Why? The reason is that the manufacturer would prefer that you plan to replace equipment every 3 years. Perhaps, following Moore's law, it is worth changing servers at such intervals, but this is not the case for storage systems. Storage should not only live 5 years, but also meet the performance requirements all this time - this is the basis for dialogue with the manufacturer.
')
2. You must be able to use all the capacity of the system you paid for.
As soon as we talk about any other product, no one wants to use only a part of the purchased goods. If you bought a house with five bedrooms, then you will want to be able to use all five times. If you bought a car with five seats, then you plan that you can put 5 people in it. If you buy a storage system, then you should be able to use 100% capacity without sacrificing performance.
Nevertheless, many suppliers accompany their product with a special “best practice” document, in which you are warned of a significant decrease in performance if you use more than 75% of the available volume. It is possible to create a storage system that will work when using the entire volume without loss of performance. But why aren't most existing architectures designed for this?
3. The upgrade should not be difficult and too expensive
Large-scale upgrades or changes in part of the existing infrastructure often develop into a complex and long-term project, which is accompanied by downtime, investment in new equipment, and a decent bill for work. Online migration is quite possible, but with a true scalable storage architecture and it can be a thing of the past.
4. The storage system does not need maintenance personnel.
It is known that the human factor is most often the main cause of failures in data centers. If we can limit the interaction of people and technology, making the equipment maintainable at the place of operation, we will significantly limit the possibility of failure in the data center. About 70% of the returned warranty discs do not have defet and a simple rebuild was enough to restore the system. The goal is to find such a storage provider who can repair and replace disks on site, which will allow your own maintenance staff not to approach the storage system for at least five years.
5. When it comes to disks, quality matters
Whether we choose hard drives or solid-state drives, the general rule works — beware of consumer-grade equipment. There is a huge difference in the quality of the components used, testing, and most importantly, in contrast to the percentage of failures for regular disks and corporate class disks.
Conventional “consumer” drives can be noticeably cheaper, but the operating costs and risks are likely to be much higher than for a reliable enterprise-level product. When choosing discs, make sure that the upcoming capital costs, operating costs and risks are balanced.
6. Flash is not a panacea
Flash drives are not the solution to any problems. Yes, they can be used to significantly improve application performance for a particular type of load, but this technology has its limitations. For example, if we are talking about sequential recording of large amounts of data, then hard drives are much better suited. Flash drive and hard disk are two different tools that can and should be used for relevant tasks.
In order to achieve a balance in performance, it is worth talking to vendors that do not limit the client to one type of disk.
7. All-Flash arrays are not always the most energy efficient.
Many people believe that storage systems that use only Flash are more energy efficient than hard drive systems. However, in order to obtain a complete picture, it is necessary, in evaluating energy consumption, to sum up the consumption of all modules, and not just disk shelves.
It turns out that some systems with hard disks consume less power than other All-Flash systems. The decisive influence is how exactly vendors use energy-hungry components in their systems, such as processors and cache memory.
8. Vibration can reduce the estimated performance.
Strong vibration can cause problems not only with reliability, but also seriously affect performance, although many do not know about it.
Although it is quite possible to obtain predictable performance results by eliminating vibration, it is not so easy and requires significant investment in the development of such a storage system. It sounds crazy, but the easiest way to check the system is to scream at it well and see how the performance changes at this time. You may be greatly surprised by the result.
9. You do not need all these bells and whistles.
The choice of corporate data storage systems is often made on the basis of a long list of capabilities, and not at all on reliability and performance indicators. With the advent of SDS (software-defined storage), a significant part of the functionality implemented at the array level is no longer so necessary, as it passes into the competence of operating systems, hypervisors and applications.
For many customers, the acquisition of an enterprise storage SHD ends in “slavery” from a supplier of proprietary software, which is supported on one single platform. The implementation of SDS provides significant flexibility in choosing a platform, without thinking about the hardware that is actually used - it strengthens your position as a customer and allows you to focus on performance and reliability.
10. Scalable systems have their own characteristics.
Many manufacturers offer scalable "up" storage systems, but most of them do not explain that for performance-demanding applications, much depends on the central controller in the system. Often this can cause a significant investment in the upgrade to increase the capacity or performance of the system.
However, there is an alternative to traditional vertical scaling. Scale-out systems can be combined into a single storage pool, combining both large capacity and high performance.
Instead of performance and reliability, the storage market is now, unfortunately, largely based on idle talk and abstruse terms. What customers really need is an honest, no-fuss conversation, in which the vendor helps them choose the best system for their business. So far there are few such vendors, but they still exist.
PS I have a certain suspicion that everything was started for the last two or three paragraphs, but I sincerely want to believe in the author's best intentions. In addition, some of the statements are largely fair and really worth paying attention to when choosing a storage system.
Source: https://habr.com/ru/post/231723/
All Articles