About choosing hard drives for servers
As the practice of communicating with customers, considering the issue of increasing server performance, most people think about replacing the processor, expanding memory or increasing the useful bandwidth of I / O systems. Unfortunately, at the same time, drives are either not considered at all as the main performance factor, or they are chosen according to the residual principle.
In this article I will talk about what predetermines the choice of drives for the server, and what type of them will be suitable for different cases. The main reason for writing is the feeling that customers with whom they have to communicate on duty do not bother with this at all, and if they think about it, then nothing more than an intuitive level. This article is an attempt to summarize the available facts by relying on some internal corporate documents. In fact, it contains an overview of the technologies used in Fujitsu PRIMERGY servers and other manufacturers.
The drives used in the server determine how well the server can “serve” the corresponding application or network. The requirements for them include not only speed and performance, but also reliability, low latency, low power consumption, in addition, they should be easily adapted to various client applications.
')
It is no secret that the hard disk function in the server differs from the disk functions used in desktops, laptops and other computers, and is determined and determined mainly by the task of the server, which in turn integrates into some network. Accordingly, hard drives in servers must withstand a greater load and serve more users than standard PCs. A data transfer request to a user or device can be issued at any time and must be processed with as little delay as possible. This means that the hard disk in the server must be functional and active at any time, while the hard disk of a regular desktop PC can be put into “standby” mode when online access is not required. And this is not just a “one” hard drive: servers are never equipped with just one drive (HDD or SSD), at least these are two drives that are installed in a RAID array in order to ensure greater performance and reliability.
The three main factors influencing the choice of hard drives are as follows:
â– applications installed on the system
â– data stored on them
â– significance of applications and data for the company
From a technical point of view, the emphasis can be placed on the performance of the drive, which has three main elements:
The following typical server usage scenarios can be distinguished:
Mail servers - are responsible for all types of communication, which include mail traffic and other types of messages. Mail servers are just “hard disk servers,” and the processor utilization level for them is not so significant. This is where reliable hard drives are required. The speed of circulation is a necessary condition, but not so significant, except for very large mail databases, where low latency is very important.
Application servers whose main task is the execution of user programs. This may be a few people from the sales department, as well as several million Internet users. This scenario requires the fastest and most reliable drives.
Storage servers are used to store various files and, as a rule, contain not only their own hard drives, but are also associated with external disk arrays. One of their top priorities is reliability. Data stored on these servers can be important for the company in its production or other business processes. Storage servers are typically associated with backup devices, such as tape libraries, optical disc recorders, or online storage services. This provides a different speed of access to information from fast “online” to slow to data in archives. This does not mean that archival information is not so important, it just does not need it so often, and, therefore, it does not need to be stored on very fast disks. However, if access to archived data becomes regular, then perhaps companies will need faster disks or a combination of reliable and fast disks.
Database servers are perhaps the most popular scenario for using servers that are repositories of databases accessible through a local network or the Internet. In addition, database servers can be connected to special application servers. Database servers must simultaneously handle multiple parallel requests from different users, which makes the requirements for drives for low latency and high reliability important.
Streaming servers provide multimedia data to employees of a company or its customers. Users can access such data for a limited period of time, and have round-the-clock access (if systems are available via the Internet). This scenario definitely requires fast disks: for large files to be always available, the servers must have the necessary speed and performance.
Servers for virtualization are the most growing use of servers. If five years ago it was necessary to prove that the future is behind such technology, now everything comes down only to correct sizing. Putting a hypervisor and running multiple virtual machines becomes a common script. Servers in this case play an important role - in fact, they are application servers, but due to the fact that they run several applications and operating systems, the requirements for the disk subsystem are even higher. For example, many manufacturers of virtualization software put special demands on the type of disks.
Of course, such a script division is rather conditional, and many organizations use a combined scheme on servers. In this situation, for each scenario, depending on its requirements for the disk subsystem, it is recommended to allocate a separate disk group.
As mentioned earlier, not all drives designed for ordinary PCs or laptops can be used in servers, since they have special requirements. The differences between server disks can be in the type of interfaces, capacity and internal components used. This determines the performance, reliability and energy efficiency of servers.
The servers have long used hard drives (HDD), but now an increasing number of manufacturers, including Fujitsu, use solid-state drives (SSD). Each type has its advantages and disadvantages.
From the outset, hard disks consisted of several disks (“pancakes”) fitted with a magnetic layer, the read / write mechanism accesses data on each such disk (“pancake”). At present, depending on the size and type of hard disk, up to 4 TB of data can be stored on it. This classic type of data storage provides a satisfactory degree of reliability, and its performance depends on various factors (speed, interface, cache), which we will talk about later.
Relatively new solid-state drives are based on flash memory, almost the same as that used in SD cards or non-volatile memory in mobile devices. There are two major differences between SSD and SD cards. SD cards use a different type of controller and they are considered by the operating system as a removable drive, and the most important difference is reliability. SD cards are suitable only for short-term data exchange, while solid-state drives have much greater reliability and are suitable for long-term data storage.
The SSD bandwidth level is much higher than that of the HDD, but their efficiency largely depends on the type of access. Direct comparison of hard drives and solid-state drives in tests have mixed results. Traditional HDD performance tests are usually aimed at those applications where hard drives have difficulties: rotational delays and search time. SSDs do not have these problems, and it can be said that SSDs offer much higher performance than hard drives in most usage scenarios. If earlier SSD had a very short life expectancy, in recent years it has been constantly increasing. Now we can say with confidence that the solid-state drives supplied by Fujitsu are ideal for long-term use in servers.
Of the many interfaces available on the market, only three types are important for servers.
SATA (Serial Advanced Technology Attachment) is a successor of parallel ATA (PATA) systems. Modern SATA drives are standardized to operate at a speed of 600 MB / s, which gives a bandwidth of 6 Gb / s to the hard disk. SATA are ideal in terms of price / quality ratio, and reliable in reading / writing.
SAS (abbreviation for "Serial Attached SCSI"). This term refers to SCSI - the basis of modern server interfaces (Small Computer System Interface). SAS uses the same sequence of commands as SCSI, being adapted to a fast serial connection. This interface offers bandwidth up to 12 Gb / s and bandwidth 1200 Mb / s per disk. Thus, it provides the advantage of a more stable and fast protocol with a higher speed, improved interference resistance and the possibility of dual-port media connections for cluster operation. Using SAS is always necessary when the focus is on high bandwidth and high reliability.
PCIe-SSD are not connected via regular SATA or SAS controllers, but directly to the PCIe bus on the server system. This exceeds the data transfer rate of SAS / SATA interfaces. PCIe-SSD drives upgraded by Fujitsu allow you to achieve a speed of 1.5 GB / s. However, due to the fact that PCIe solid-state drives are connected directly to the system bus, they cannot be bootable (for more information on PCIe SSDs ioDrive2 and I recommend reading them here ).
A number of Fujitsu PRIMERGY generation S8 server models announced last week also have support for 2.5 "PCIe SSDs, which means there is an opportunity to install a special PCIe backplane in the HDD cage, by connecting directly to the PCIe bus of the standard 2.5" SSDs in the usual disc baskets.
Also recall that SAS is “backward compatible” with SATA, i.e. SATA and SAS drives can be connected to the SAS controller. However, this does not work in the opposite direction - SAS drives cannot be connected to host SATA adapters.
The form factor of the disks installed in the server determines the capacity of its storage system and at the same time its energy efficiency. Currently only two sizes of disks are available and relevant in servers: 3.5 "and 2.5".
3.5 "is the most widely used size. They allow you to place the maximum amount of data - 4 TB. 3.5" hard disks are usually recommended for solutions with high storage requirements. But at the same time, they consume more energy than smaller 2.5 "disks, and SSDs are not available in this form factor. The main feature of 3.5" hard drives is high capacity at a more favorable price. We can say that 3.5 ”wheels offer the best prices per GB.
2.5 "HDD and SSD frequent size. Although these drives are just one inch smaller in size, they use much less power than their 3.5" brothers. Currently, the maximum amount of such a disk is 2TB. Compared to their larger satellites, 2.5 "hard drives are not only economical, but also offer better read / write speeds when working on a network of several carriers. Thus, they are ideally suited for use in low-power systems or for systems with the required maximum performance.
Now let's take a look at the differences in hard drives in terms of internal components.
The data in the HDD is stored on "disks" - on glass or aluminum disks with a layer of magnetizable material (chromium dioxide). The higher the density of these “disks” and the more such disks are contained in one HDD, the higher its capacity. All discs rotate at a certain speed, which determines its throughput, access time, as well as its power consumption. Servers with high performance requirements, as a rule, are equipped with hard drives with a rotation speed of 7200, 10 000 or 15 000 revolutions per minute. Faster drives not only consume more power, but also require improved cooling mechanisms. Another problem is due to vibrations. When the discs in the case rotate at different speeds, they can interfere with each other while in different write cycles. All disks that work in the same place must have the same rotational speed (i.e., the same interface). If disks with different speeds or interfaces are combined into a server with different applications (see the table at the end), they should always be placed in different places in the chassis.
Solid-state drives have no moving parts, and the data is stored in flash memory cells. Tests conducted by Fujitsu revealed that SSD input / output performance was improved by 100 times compared to a hard disk. And at the same time, they use only one fifth of the power consumption of HDD, because they do not have the need for an electric motor. Due to the absence of moving parts, these drives are not subject to mechanical wear and are not sensitive to temperature and vibration. However, one of the disadvantages of SSD-drives is their limited lifespan. If the life of hard drives is limited due to their mechanical wear, then SSD wear is due to an electrical effect. The essence of which is that the number of write operations on flash memory has a limitation from 3000 to 100000 depending on its quality. Upon reaching these values, reading the information will not be reliable. This rapid deterioration of SSDs created problems at the very beginning of their appearance. To combat uneven wear, balancing schemes are applied by storing information about the number of times, which blocks have been rewritten and periodically changing the recording sequence. This distributed procedure at the controller level leads to an increase in the lifespan of SSDs to periods of similar or even longer than normal hard drives.
The main parameter of each drive (HDD or SSD) is its capacity. As noted earlier, 3.5 "hard drives can store up to 4 TB, and 2.5" hard drives are now available with a maximum capacity of 2 TB, and solid state drives are available up to 600 GB. It is not always advisable to choose drives of maximum volume in order to get a certain capacity. The hard disk should not be installed in the server in the singular, but should always be combined into groups of two (at a minimum) to ensure reliable operation. If, for example, you plan to equip a server with a disk space of a total of 2 TB, then it is much better to buy four 1 TB drives and then configure them into a RAID array. So that any failure in one of the disks does not lead to data loss. This is one of the reasons why the capacity of server hard drives is usually lower than their desktop counterparts.
When choosing the right drives for the server, be sure to take into account the following important parameters:
To ensure that all HDDs and SSDs meet these requirements, they must be rigorously tested and appropriately certified. This applies to all drives installed by Fujitsu in their PRIMERGY servers.
In order to summarize all of these aspects and simplify customer choices, Fujitsu defines several classes for hard drives:
Economic (ECO) . Discs in this class have a low unit price. The level of performance and reliability of these drives determines their purpose for entry-level systems. They should be used in non-critical areas with low I / O traffic and moderate speed requirements. High loads may impair their reliability. ECO drives run at 5400 or 7200 revolutions per minute and have an SATA interface.
Business Critical (BC) or Nearline . Disks in this class offer high capacity with a minimum cost per 1 GB. They are designed to provide good performance and suitable reliability. Depending on the server implementation, BC-disks can be equipped with SAS or SATA interfaces and have a speed of 7200 revolutions per minute.
Enterprise (EP) . Disks of this class provide maximum performance and reliability. They are designed to cope with the maximum workload. This class uses the SAS interface and has a rotation speed of 10,000 and 15,000 revolutions per minute.
SSD Enterprise Performance / Mainstream . These drives offer the best performance and durability in the SSD segment and are therefore ideally suited for use in systems with high I / O price / capacity requirements. Enterprise Performance SSDs (SLC or MLC technologies) offer better I / O performance using the SAS interface. In contrast, Enterprise Mainstream SDD (MLC technology) has SATA interface and is more affordable.
Now we will try to reduce all the above to the table:
“Is it possible to buy your server, and the disks for it separately?”, “And why do not you sell the sled separately?”, “We bought your server, and our disks do not work in it!”, Etc. It's nice that every year there are less and less such questions. It's nice that our consciousness is gradually turning from "Yes, I will collect everything myself!" To "We need to take a complete and tested device." We all remember how our fathers repaired their “Lada” buying parts “where it will turn out”. The concept of "I will assemble myself", "I will do it myself" in our brain sits with mother's milk. With the advent of world car manufacturers to us, we began to get used to buying cars that do not need to be modified. We get used to maintain and upgrade them in specialized services. So with the servers.
You may ask why you cannot use disks from Fujitsu servers, which are sold in computer stores and on radio markets. Indeed, in the first approximation, they are no different from those hard drives and solid-state drives that come bundled with the Fujitsu server. They are so much alike. The answer to this question is simple: the security of your data. That is what should be the decisive factor determined by the invisible differences.
The main differences are really not visible at first sight, because The most important is that all new disks for Fujitsu servers are tested in the smallest detail and, accordingly, will be tested and adapted to PRIMERGY servers.
The first step is to adapt the microcode (firmware) of the hard disk together with its manufacturer to the requirements of the PRIMERGY servers. This includes setting the basic parameters and flashing the firmware that was specifically designed for Fujitsu. For example: all problems in the firmware, which are determined during the preliminary tests, are corrected and integrated into the new modified firmware, and, accordingly, the disks from the same batch will contain the same firmware.
The second step is that the drives pass the input tests on Fujitsu stands. These tests are different: shock, vibration and temperature tests. The challenge of these tests is to detect any weak points. The test center also simulates long-term operation of storage media. Tests are also conducted on the claimed and required read and write data rates. After the discs are tested on the stands, they are checked in real systems and certified.
Once the disks are successfully certified, they are subject to additional input tests (random tests using the AQL method). This test is subjected to 100% of the delivered media. If malfunctions do not occur during the first test, the proportion of products tested from the lot is reduced to 10%. If errors begin to occur during subsequent tests, the proportion for testing increases.
And even after the disks are delivered to the conveyor, Fujitsu continues to identify the causes and error rates based on reports from Fujitsu Services. The problems identified at this stage allow to correct them by adjusting the firmware.
If you decide to increase the volume of the disk subsystem a few years after the purchase, then you can get a fully compatible HDD and SSD for at least three years. And these discs will also be tested and certified for your system, as well as previously released.
Choosing the right drives for a server is quite a challenge. In this article, in spite of its volume, I was able to highlight this only quite superficially. If you decide to purchase a Fujitsu PRIMERGY server, you can always get expert advice and assistance in sizing with our authorized partners.
In this article I will talk about what predetermines the choice of drives for the server, and what type of them will be suitable for different cases. The main reason for writing is the feeling that customers with whom they have to communicate on duty do not bother with this at all, and if they think about it, then nothing more than an intuitive level. This article is an attempt to summarize the available facts by relying on some internal corporate documents. In fact, it contains an overview of the technologies used in Fujitsu PRIMERGY servers and other manufacturers.
The drives used in the server determine how well the server can “serve” the corresponding application or network. The requirements for them include not only speed and performance, but also reliability, low latency, low power consumption, in addition, they should be easily adapted to various client applications.
')
Hard drive function in server
It is no secret that the hard disk function in the server differs from the disk functions used in desktops, laptops and other computers, and is determined and determined mainly by the task of the server, which in turn integrates into some network. Accordingly, hard drives in servers must withstand a greater load and serve more users than standard PCs. A data transfer request to a user or device can be issued at any time and must be processed with as little delay as possible. This means that the hard disk in the server must be functional and active at any time, while the hard disk of a regular desktop PC can be put into “standby” mode when online access is not required. And this is not just a “one” hard drive: servers are never equipped with just one drive (HDD or SSD), at least these are two drives that are installed in a RAID array in order to ensure greater performance and reliability.
Typical server applications and requirements
The three main factors influencing the choice of hard drives are as follows:
â– applications installed on the system
â– data stored on them
â– significance of applications and data for the company
From a technical point of view, the emphasis can be placed on the performance of the drive, which has three main elements:
- Speed. Decisive in this sense is the parameter of the number of IOPS (I / O operations per second) that the hard disk can process, as well as the amount of raw data (useful information without headers) transmitted during a certain period of time.
- Delays, that is, the time that passes between the data request and the moment when the data reaches the users.
- Reliability - the lifespan of a data carrier can be a decisive factor if data are to be stored for a long period of time.
The following typical server usage scenarios can be distinguished:
Mail servers - are responsible for all types of communication, which include mail traffic and other types of messages. Mail servers are just “hard disk servers,” and the processor utilization level for them is not so significant. This is where reliable hard drives are required. The speed of circulation is a necessary condition, but not so significant, except for very large mail databases, where low latency is very important.
Application servers whose main task is the execution of user programs. This may be a few people from the sales department, as well as several million Internet users. This scenario requires the fastest and most reliable drives.
Storage servers are used to store various files and, as a rule, contain not only their own hard drives, but are also associated with external disk arrays. One of their top priorities is reliability. Data stored on these servers can be important for the company in its production or other business processes. Storage servers are typically associated with backup devices, such as tape libraries, optical disc recorders, or online storage services. This provides a different speed of access to information from fast “online” to slow to data in archives. This does not mean that archival information is not so important, it just does not need it so often, and, therefore, it does not need to be stored on very fast disks. However, if access to archived data becomes regular, then perhaps companies will need faster disks or a combination of reliable and fast disks.
Database servers are perhaps the most popular scenario for using servers that are repositories of databases accessible through a local network or the Internet. In addition, database servers can be connected to special application servers. Database servers must simultaneously handle multiple parallel requests from different users, which makes the requirements for drives for low latency and high reliability important.
Streaming servers provide multimedia data to employees of a company or its customers. Users can access such data for a limited period of time, and have round-the-clock access (if systems are available via the Internet). This scenario definitely requires fast disks: for large files to be always available, the servers must have the necessary speed and performance.
Servers for virtualization are the most growing use of servers. If five years ago it was necessary to prove that the future is behind such technology, now everything comes down only to correct sizing. Putting a hypervisor and running multiple virtual machines becomes a common script. Servers in this case play an important role - in fact, they are application servers, but due to the fact that they run several applications and operating systems, the requirements for the disk subsystem are even higher. For example, many manufacturers of virtualization software put special demands on the type of disks.
Of course, such a script division is rather conditional, and many organizations use a combined scheme on servers. In this situation, for each scenario, depending on its requirements for the disk subsystem, it is recommended to allocate a separate disk group.
Technical details
As mentioned earlier, not all drives designed for ordinary PCs or laptops can be used in servers, since they have special requirements. The differences between server disks can be in the type of interfaces, capacity and internal components used. This determines the performance, reliability and energy efficiency of servers.
HDD and SSD
The servers have long used hard drives (HDD), but now an increasing number of manufacturers, including Fujitsu, use solid-state drives (SSD). Each type has its advantages and disadvantages.
From the outset, hard disks consisted of several disks (“pancakes”) fitted with a magnetic layer, the read / write mechanism accesses data on each such disk (“pancake”). At present, depending on the size and type of hard disk, up to 4 TB of data can be stored on it. This classic type of data storage provides a satisfactory degree of reliability, and its performance depends on various factors (speed, interface, cache), which we will talk about later.
Relatively new solid-state drives are based on flash memory, almost the same as that used in SD cards or non-volatile memory in mobile devices. There are two major differences between SSD and SD cards. SD cards use a different type of controller and they are considered by the operating system as a removable drive, and the most important difference is reliability. SD cards are suitable only for short-term data exchange, while solid-state drives have much greater reliability and are suitable for long-term data storage.
The SSD bandwidth level is much higher than that of the HDD, but their efficiency largely depends on the type of access. Direct comparison of hard drives and solid-state drives in tests have mixed results. Traditional HDD performance tests are usually aimed at those applications where hard drives have difficulties: rotational delays and search time. SSDs do not have these problems, and it can be said that SSDs offer much higher performance than hard drives in most usage scenarios. If earlier SSD had a very short life expectancy, in recent years it has been constantly increasing. Now we can say with confidence that the solid-state drives supplied by Fujitsu are ideal for long-term use in servers.
Interfaces
Of the many interfaces available on the market, only three types are important for servers.
SATA (Serial Advanced Technology Attachment) is a successor of parallel ATA (PATA) systems. Modern SATA drives are standardized to operate at a speed of 600 MB / s, which gives a bandwidth of 6 Gb / s to the hard disk. SATA are ideal in terms of price / quality ratio, and reliable in reading / writing.
SAS (abbreviation for "Serial Attached SCSI"). This term refers to SCSI - the basis of modern server interfaces (Small Computer System Interface). SAS uses the same sequence of commands as SCSI, being adapted to a fast serial connection. This interface offers bandwidth up to 12 Gb / s and bandwidth 1200 Mb / s per disk. Thus, it provides the advantage of a more stable and fast protocol with a higher speed, improved interference resistance and the possibility of dual-port media connections for cluster operation. Using SAS is always necessary when the focus is on high bandwidth and high reliability.
PCIe-SSD are not connected via regular SATA or SAS controllers, but directly to the PCIe bus on the server system. This exceeds the data transfer rate of SAS / SATA interfaces. PCIe-SSD drives upgraded by Fujitsu allow you to achieve a speed of 1.5 GB / s. However, due to the fact that PCIe solid-state drives are connected directly to the system bus, they cannot be bootable (for more information on PCIe SSDs ioDrive2 and I recommend reading them here ).
A number of Fujitsu PRIMERGY generation S8 server models announced last week also have support for 2.5 "PCIe SSDs, which means there is an opportunity to install a special PCIe backplane in the HDD cage, by connecting directly to the PCIe bus of the standard 2.5" SSDs in the usual disc baskets.
Also recall that SAS is “backward compatible” with SATA, i.e. SATA and SAS drives can be connected to the SAS controller. However, this does not work in the opposite direction - SAS drives cannot be connected to host SATA adapters.
Form Factor (Size)
The form factor of the disks installed in the server determines the capacity of its storage system and at the same time its energy efficiency. Currently only two sizes of disks are available and relevant in servers: 3.5 "and 2.5".
3.5 "is the most widely used size. They allow you to place the maximum amount of data - 4 TB. 3.5" hard disks are usually recommended for solutions with high storage requirements. But at the same time, they consume more energy than smaller 2.5 "disks, and SSDs are not available in this form factor. The main feature of 3.5" hard drives is high capacity at a more favorable price. We can say that 3.5 ”wheels offer the best prices per GB.
2.5 "HDD and SSD frequent size. Although these drives are just one inch smaller in size, they use much less power than their 3.5" brothers. Currently, the maximum amount of such a disk is 2TB. Compared to their larger satellites, 2.5 "hard drives are not only economical, but also offer better read / write speeds when working on a network of several carriers. Thus, they are ideally suited for use in low-power systems or for systems with the required maximum performance.
Internal components
Now let's take a look at the differences in hard drives in terms of internal components.
The data in the HDD is stored on "disks" - on glass or aluminum disks with a layer of magnetizable material (chromium dioxide). The higher the density of these “disks” and the more such disks are contained in one HDD, the higher its capacity. All discs rotate at a certain speed, which determines its throughput, access time, as well as its power consumption. Servers with high performance requirements, as a rule, are equipped with hard drives with a rotation speed of 7200, 10 000 or 15 000 revolutions per minute. Faster drives not only consume more power, but also require improved cooling mechanisms. Another problem is due to vibrations. When the discs in the case rotate at different speeds, they can interfere with each other while in different write cycles. All disks that work in the same place must have the same rotational speed (i.e., the same interface). If disks with different speeds or interfaces are combined into a server with different applications (see the table at the end), they should always be placed in different places in the chassis.
Solid-state drives have no moving parts, and the data is stored in flash memory cells. Tests conducted by Fujitsu revealed that SSD input / output performance was improved by 100 times compared to a hard disk. And at the same time, they use only one fifth of the power consumption of HDD, because they do not have the need for an electric motor. Due to the absence of moving parts, these drives are not subject to mechanical wear and are not sensitive to temperature and vibration. However, one of the disadvantages of SSD-drives is their limited lifespan. If the life of hard drives is limited due to their mechanical wear, then SSD wear is due to an electrical effect. The essence of which is that the number of write operations on flash memory has a limitation from 3000 to 100000 depending on its quality. Upon reaching these values, reading the information will not be reliable. This rapid deterioration of SSDs created problems at the very beginning of their appearance. To combat uneven wear, balancing schemes are applied by storing information about the number of times, which blocks have been rewritten and periodically changing the recording sequence. This distributed procedure at the controller level leads to an increase in the lifespan of SSDs to periods of similar or even longer than normal hard drives.
Volume (Capacity)
The main parameter of each drive (HDD or SSD) is its capacity. As noted earlier, 3.5 "hard drives can store up to 4 TB, and 2.5" hard drives are now available with a maximum capacity of 2 TB, and solid state drives are available up to 600 GB. It is not always advisable to choose drives of maximum volume in order to get a certain capacity. The hard disk should not be installed in the server in the singular, but should always be combined into groups of two (at a minimum) to ensure reliable operation. If, for example, you plan to equip a server with a disk space of a total of 2 TB, then it is much better to buy four 1 TB drives and then configure them into a RAID array. So that any failure in one of the disks does not lead to data loss. This is one of the reasons why the capacity of server hard drives is usually lower than their desktop counterparts.
Quality factors
When choosing the right drives for the server, be sure to take into account the following important parameters:
- Estimation of the life of the hard disk. Hard drive manufacturers use a parameter known as MTBF to evaluate it. This abbreviation stands for “mean time between failures” (average number of working hours between downtime) and speaks of the time after which the hard drive will probably fail. MTBF is measured in hours, and this value is always quite large. However, this is only a calculated value, and it does not guarantee that the disk will not fail before this time is reached.
- In addition to physical capacity and performance values, it should also be noted that most disks in the server should work continuously or more precisely: disks should be able to work continuously without any interruptions. Fujitsu offers only drives for its servers that meet these conditions. At the same time, MTBF for hard drives of desktop devices is indicated based on an 8-hour working day. SSDs currently sold by Fujitsu are always covered by the warranty period for the entire server system. Another interesting parameter is the DWPD (the number of drive rewrites per day), which can be considered to estimate the lifespan of solid-state drives. Fujitsu's SSDs based on MLC technology (multi-level cells) have a value of 10 DWPD over five years. Solid state drives with SLC technology (single-level cells) are defined by higher characteristics in the amount of 50 DWPD for five years. The trend of the latest developments in flash technology is a move towards more cost-effective MLC technology.
- Another feature of SSD is their limited ability to store data in the off state. If the SSD is removed from the server, and, for example, put into the cabinet as a backup, the information stored on it will remain available at best for ten years. At the same time, various factors (type of SLC / MLC flash technology, previous usage intensity or ambient temperature) reduce the shelf life. It is determined that the minimum shelf life due to appropriate life expectancy projections is six months for SLC SSD and three months for MLC SSD.
To ensure that all HDDs and SSDs meet these requirements, they must be rigorously tested and appropriately certified. This applies to all drives installed by Fujitsu in their PRIMERGY servers.
Hard Drive Classification
In order to summarize all of these aspects and simplify customer choices, Fujitsu defines several classes for hard drives:
Economic (ECO) . Discs in this class have a low unit price. The level of performance and reliability of these drives determines their purpose for entry-level systems. They should be used in non-critical areas with low I / O traffic and moderate speed requirements. High loads may impair their reliability. ECO drives run at 5400 or 7200 revolutions per minute and have an SATA interface.
Business Critical (BC) or Nearline . Disks in this class offer high capacity with a minimum cost per 1 GB. They are designed to provide good performance and suitable reliability. Depending on the server implementation, BC-disks can be equipped with SAS or SATA interfaces and have a speed of 7200 revolutions per minute.
Enterprise (EP) . Disks of this class provide maximum performance and reliability. They are designed to cope with the maximum workload. This class uses the SAS interface and has a rotation speed of 10,000 and 15,000 revolutions per minute.
SSD Enterprise Performance / Mainstream . These drives offer the best performance and durability in the SSD segment and are therefore ideally suited for use in systems with high I / O price / capacity requirements. Enterprise Performance SSDs (SLC or MLC technologies) offer better I / O performance using the SAS interface. In contrast, Enterprise Mainstream SDD (MLC technology) has SATA interface and is more affordable.
Configuration selection
Now we will try to reduce all the above to the table:
Why use original Fujitsu discs?
“Is it possible to buy your server, and the disks for it separately?”, “And why do not you sell the sled separately?”, “We bought your server, and our disks do not work in it!”, Etc. It's nice that every year there are less and less such questions. It's nice that our consciousness is gradually turning from "Yes, I will collect everything myself!" To "We need to take a complete and tested device." We all remember how our fathers repaired their “Lada” buying parts “where it will turn out”. The concept of "I will assemble myself", "I will do it myself" in our brain sits with mother's milk. With the advent of world car manufacturers to us, we began to get used to buying cars that do not need to be modified. We get used to maintain and upgrade them in specialized services. So with the servers.
You may ask why you cannot use disks from Fujitsu servers, which are sold in computer stores and on radio markets. Indeed, in the first approximation, they are no different from those hard drives and solid-state drives that come bundled with the Fujitsu server. They are so much alike. The answer to this question is simple: the security of your data. That is what should be the decisive factor determined by the invisible differences.
The main differences are really not visible at first sight, because The most important is that all new disks for Fujitsu servers are tested in the smallest detail and, accordingly, will be tested and adapted to PRIMERGY servers.
The first step is to adapt the microcode (firmware) of the hard disk together with its manufacturer to the requirements of the PRIMERGY servers. This includes setting the basic parameters and flashing the firmware that was specifically designed for Fujitsu. For example: all problems in the firmware, which are determined during the preliminary tests, are corrected and integrated into the new modified firmware, and, accordingly, the disks from the same batch will contain the same firmware.
The second step is that the drives pass the input tests on Fujitsu stands. These tests are different: shock, vibration and temperature tests. The challenge of these tests is to detect any weak points. The test center also simulates long-term operation of storage media. Tests are also conducted on the claimed and required read and write data rates. After the discs are tested on the stands, they are checked in real systems and certified.
Once the disks are successfully certified, they are subject to additional input tests (random tests using the AQL method). This test is subjected to 100% of the delivered media. If malfunctions do not occur during the first test, the proportion of products tested from the lot is reduced to 10%. If errors begin to occur during subsequent tests, the proportion for testing increases.
And even after the disks are delivered to the conveyor, Fujitsu continues to identify the causes and error rates based on reports from Fujitsu Services. The problems identified at this stage allow to correct them by adjusting the firmware.
If you decide to increase the volume of the disk subsystem a few years after the purchase, then you can get a fully compatible HDD and SSD for at least three years. And these discs will also be tested and certified for your system, as well as previously released.
Instead of conclusion
Choosing the right drives for a server is quite a challenge. In this article, in spite of its volume, I was able to highlight this only quite superficially. If you decide to purchase a Fujitsu PRIMERGY server, you can always get expert advice and assistance in sizing with our authorized partners.
Source: https://habr.com/ru/post/195798/
All Articles