The server, which is designed to manage the main processes that are running in the local area, will require a fairly large amount of effort. The more server roles are assigned to the managing server, the greater the challenge is to know. For reasons not varto marvel, schogi to the productivity of the server signifi- cantly improve in the same way, as if they are presented to a great job.

The choice of server configuration can be set as at the stage of designing the fence, which allows you to more accurately determine the variability of the creation of the fence, so after the installation of the fence is completed and the food about the choice of method of functioning is checked.

If the choice is made for the selected domain structure, then the server configuration selection step will be mandatory, and the purchase of the server is required.

When choosing the configuration of the server that you manage, you should check the following features:

• uninterrupted work;
• security of authentication of lace coristuvachs;
• collection of all data about the appearance of the records of the correspondents and computers;
• possibility of typing for typing additional roles, for example, DNS- and DHCP-servers;
• the possibility of zastosuvannya for serving web addenda;
• feasibility of winning additional software, for example, a corporate anti-virus system;
• the possibility of connecting the data archiving system, for example, a streamer;
• synchronization to the hour on all computers of the merezhі.

Cream of that We care about nutritionє vybіr variant vykonannya server: okrema installation chi installation at the rack.

Okrema installation transferring the server to the server, which takes an hour to bring the server room to the point where it is occupied by servers of various recognition. In order to improve order, bring vicoristovuvati improvised furniture stacks, as if it is possible to install servers and two or three tiers.

Even more often (especially large meshes are needed) at the server and special server racks, they are used to install servers of a rack type of different recognition. In this case, as a rule, one keyboard with a monitor, that system of KVN jumpers, is zastosovuetsya for keruvannya servers, as it allows the switching of the system and the introduction of that system to the required server. It is necessary to do it manually, the shards of the front panels of the servers are always placed in front of the eye, which allows visual control of their practicality, and the racks themselves can be sized when necessary.

Navit unimportantly to those that a stable server takes less than a month, and may be short of a share with the server, which is worth it - as a rule, only one block of life is victorious. Two blocks of life are practically installed in the server, one of them is a reserve one, allowing to increase the server's efficiency, to put the main block of life in harmony.

If there is a keruyuchy server, then there is a sufficient amount of standard configurations, operational memory, depending on the type of disk subsystem and other characteristics.

One of the typical configurations of a stand-alone server, which can be used to manage a local network of 80-120 computers, is listed in Table 1.

Table 1

Accessories	Decryption
Chip set	Intel 5100
processors	1 or 2 Intel Xeon 5xXX (Up to 8 cores)
System bus speed	1333 MGp
	24 GB Dual Channel DDRII-667 ECC
Extension slots	1x PCI-E 16x, 1xPCI-E 8x, 2x PCI 32/33
Introduced inspectors	8 SAS LSI 1068 ports
Optical storage devices	DVD-R/W
Optional controllers	SAS/SATA controllers with VPS support FibreChanncl and InfiniBand HCA adapters
	4 standard and up to 8 SAS/SATA with hot swap capability
Disk subsystem capacity	Up to 12TB SATA or 3.6 Toant SAS
Merezhev interface	2x Intel Gigabit Ethernet, IOAT
Video controller	ASPEED AST2000, 8 MB
interface	Rear panel: VGA, RS232, 3xRJ-45, 2xUSB, 2x PS2; front panel: 2xUSB
System management	IPMI 2.0, KVM over IP. Virtual Media, Ethernet
Support OS	Microsoft Windows Server 2008 family; Red Hat Enterprise Linux 5.0; Sun Solaris 10
Rosemary (L x W x H), mm	5U, 620 x 430 x 220
Living block	2x550W

Expanding the server, which can cost, allows you to take more, in parallel with a similar stand-alone server, tightness and functionality, the possibility of installing a larger number of overall components. One of the similar typical server configurations is listed in Table 2.

Table 2. Configuration of the key server type type

Accessories	Decryption
Chip set	Intel 5100
processors	1 or 2 Intel Xeon 5xxx (up to 8 cores)
System bus speed	1333 MHz
Maximum operating memory capacity	24 GB Dual Channel DDRII-667 ECC
Extension slots	1xPCI-E 16x, 1xPCI-E 8x
Introduced inspectors	2 LSI SAS 1064 ports
Optical storage devices	DVD-CD-RW
Optional controllers	SAS RAID controllers with BBU support, FibreChannel adapters, 10G Ethernet and InfiniBand HCA
Maximum number of disks	2xSAS 3.5" or 4xSATA/SSD 2.5" with hot swap capability
Disk subsystem capacity	Up to 900 GB SAS/3 TB SATA
Merezhev interface	2x Intel Gigabit Ethernet
Video controller	ASPEED AST2000, 8 MB
interface	Rear panel: VGA, RS232, 3xRJ-45, 2xUSB, 2xPS2; front panel: 2xUSB
System management	IPMI 2.0 KVM over IP, Virtual Media, Ethernet
Support OS	SuSE Linux Enterprise Server 10; Novell Open Enterprise Server; Microsoft Windows Server 200X family; Red Hat Enterprise Linux 5.0; Sun Solaris 10
Rosemary	(W x H x D), mm 1U, 510 x 430 x 44 mm (stack depth not less than 800 mm)
Living block	400 W

Who would have thought for a moment, what for the help of a little drototy headset you can make VoIP calls through a computer or a tablet, or you can learn how to recognize, how to know your master - on the street, in the office, or in the car, and in the fall, you will ignore third-party sounds. Otherwise, the headset can give an intelligent call, call out loudly, and the subscriber can either take it and you can communicate. But all the same, it’s not a fairy tale, but reality, backed up by market propositions, like zayviy times to bring up that it’s a unifying tool that unites all other voice communication skills, there are no equal headsets.

Literature

1. A.I. Vatamanyuk. "Creation, maintenance and administration of the border by 100%"

Before our company, like a postal manager of a server owner, clients often ask for the wrong configuration of the 1C server for the first time.

In case of power failure, it is necessary to protect the impersonal officials: system architecture, software type, nature of the work of the coristuvachіv, comprehension and foldability of the information base, etc. However, for other equal minds, the productivity of the system is completely dependent on the hardware configuration of the server host.

In this article, we share the results of testing the Intel dual-processor server platform in the middle "1C: Enterprise 8" for the additional tool "Standard navantage test The results of the test can be selected as a valid point when choosing the server configuration for 1C in the part of the processor subsystem and the memory subsystem.

The test was awarded by "Alcor Services" company, which has the status of "1C: Franchised" and awarded software security by "1C" company, and by "Tim Computers", "platinum" supplier of Intel server solutions and "golden" OEM partner of Microsoft .

What are we testing?

We will simulate the productivity of the server system for an hour of work with the test information base "1C: Enterprise 8.2" in the configuration "Management of Enterprise Production" in the client-server mode with the DBMS Microsoft SQL Server.

As a server system, we will use the Intel dual processor platform for processors of the Intel Xeon E5-2600 family, on which the Microsoft Windows Server 2012 operating system has an add-on server "1C: Enterprise 8.2" and a Microsoft SQL Server 2012 DBMS.

Productivity is estimated in the "Standard 1C Corylists" for the additional tool "Standard Test of Profitability".

In our opinion, the productivity of the system is dependent on advancing factors:

1. hardware configuration
2. version of the operating system, "1C: Enterprise" and DBMS and its installation

In the course of testing, we will try to find out, which may affect the productivity of the server hardware configuration. Therefore, in all tests, we will test one and the same version of the Microsoft Windows Server 2012 operating system, the Microsoft SQL Server 2012 DBMS, and the "1C: Enterprise 8.2" program server. In addition, in all situations, we will zastosovuvat the same optimal, in our opinion, improved software security.

For the possession of vicorists, it is time to improve, to ensure maximum productivity.

How does the "Standard Test of Interest" work?

"Standard navantazhuvalny test" allows you to designate the maximum number of "standard 1C cores", so that you can service the server system that is being tested with sufficient productivity.

"Standard navigation test" is created on the server system parallel to the launch of a large number of virtual "standard core tests" in 1C thin client mode. Qi coristuvachi are launched on "existing" servers, so as not to take advantage of the system resources, which are tested for client addenda.

Skin such a "standard koristuvach" once for the last few days of the next cycle of the business process "sales in the UPP", which includes the holding of decal documents, the creation of calls and other actions. If so, the skin "koristuvach" works with its own set of data, which makes it difficult to blame blocking.

In the course of the test, the number of corestuvachiv increases due to the given crop dot, the dock productivity of the system is left at the necessary level. Productivity wins over the methodology APDEX (A application P performance In dex), the essence of which lies in the offensive.

For the sale cycle, which is counted by the "standard koristuvach", the number of hours is set Ts- 15 seconds. During the testing process, the actual hour of the skin cycle is frozen, the sale is calculated and the index is calculated APDEX behind the formula:

APDEX = (Ns + Nf/2)/N, de

Ns- number of sales cycles, yakі "invested" in the whole hour Ts(15 seconds)

Nf- number of sales cycles, vykonanih per hour Ts before 4 x Ts(for 15 to 60 seconds)

N- total number of cycles for sale

What is the value of the index, what was it, APDEX If the greater or better value is 0.85, then the productivity of the system is considered to be sufficient, and the test for this number of cores will be passed. The number of coristuvachs increases doti, docks productivity for the index APDEX not fall below 0.85. The result of the test is the maximum number of standard corylists, for which the test was completed successfully.

Dual processor server platform testing

As already mentioned, as a server system for "1C: Enterprise" we will win a dual-processor server platform based on processors of the Intel Xeon E5-2600 family, on which in the middle of the Microsoft Windows Server 2012 operating system, an add-on server "1C: Enterprise 8.2" was installed Microsoft SQL Server 2012

In the course of testing, we are able to take into account the following questions:

What is the maximum number of "standard 1C cores" that can "serve" a dual-processor server based on Intel Xeon E5-2600?

How does the model and the number of processors, the number of cores and their frequency affect the productivity of the system?

How does the inclusion of modes affect the results of the test turbo boost(TB) ta Hyper Threading(HT)?

How does it affect the productivity of the operational memory, the frequency and the number of channels?

How does the configuration of the disk subsystem of the server affect the results of the test?

What is the cost of navigating the meshed interfaces of the server to be tested for the hour of testing?

Testing the server in a configuration with one processor

Servers based on a dual-processor platform in configurations with one embedded processor are required, but chips allow you to save at the server installation stage, so that one processor can get in line with streaming tasks. The productivity of the server can be improved by using a different processor. Therefore, one-processor changes to create interest at the borders of our testing.

We test the system with different processors of the Intel Xeon E5-2600 family. Other parameters of the configuration are permanent: 64GB of DDR3-1600 RAM in four-channel mode, 4 gigabytes of server mesh interface, RAID 1 from two server SSD drives, hardware RAID controller Intel RMS25PB080. The test results are shown on the graph.

It can be seen that the number of "standard cores" correlates well with the "hardness" of the processor - the total frequency of its cores. In other words, productivity growth is directly proportional to the number of cores of those clock frequencies. The amount of gigahertz of the total frequency of the processor adds about 20 "standard clock speeds" to the productivity of the system.

Blame it on the "younger" processor Intel Xeon E5-2609 v2, which obviously lacks productivity (only 14 "standard cores" per gigahertz of the total frequency). This is easy to explain - this cob processor does not support Intel Turbo Boost and Intel Hyper Threading modes.

Let's take a look at these regimes in the report. Turbo Boost technology increases the frequency of one or more processor cores, so that the energy saving and thermal imaging of the cores are kept within acceptable limits, with which the core frequency can be increased from 100 MHz by one or two times for a short time. Hyper Threading technology allows the core to process two streams of instructions at the same time, which is why the operating system "flows" one physical core like two logical CPUs.

We protested the Intel Xeon E5-2650 processor with enabled and disabled Turbo Boost and Hyper Threading modes in various combinations:

CPU model	Intel Turbo Boost	Intel Hyper threading	Standard koristuvannya- TV 1C	Prist for the rahunok TB (%)	Prist for the rahunok HT (%)	Prist for the rahunok TB and HT (%)
Xeon E5-2650 8 cores x 2 GHz	vindicated	vindicated	230	-	-	-
		mentioned	300	-	30%	-
	mentioned	vindicated	270	17%	-	-
		mentioned	340	13%	26%	48%

It turned out that the inclusion of Turbo Boost increased the productivity of the average by 15%, and Hyper Threading - by 28%. The total increase in the number of swed codes with the help of both regimes to become 48%.

Working memory

Nezalezhno vіd navantazhennya vysjag vikoristovuvanої operative їmyatі pіd hour testuvannya not peresuschuvav signs 32 GB. At about 2 GB was hacked by the operating system, no more than 16 GB - by 1C server working processes, and about 12 GB - by the SQL server with 100% caching of the test database in the server's memory.

Depending on the required amount of memory for a specific task, it is necessary to enter 2 GB for the operating system and 4 GB for the skin of the 1C server working process. The amount of memory for the SQL server is responsible for at least 30% of the data base size, and even more shortly, as the entire database can fit in the server's RAM.

Let's move on to the power supply of the operating frequency of the operational memory. Chi vplyaє won on the result of the test? Inject, but only for the oldest of the protests processors Intel Xeon E5-2697 v2. With a memory frequency of 1600 MHz, the whole processor looks like 630 clocks - estimated productivity from the clock is approximately 20 clocks per gigahertz of the total processor frequency (32.4 GHz). In this case, the memory bandwidth becomes 51.2 GB / s (4 channels x 8 bytes per channel x 1.6 GHz). And the memory axis is 1333 MHz (memory bandwidth is 42.7 GB / s), the test result is lower - only 600 cores. You can let go, what to remember, so that you don’t be a narrow place when you are venturing this type, is responsible for ensuring a bandwidth of at least 1.6 Gb / s per gigahertz of the total frequency of the processor. As a matter of fact, for example, for the Intel Xeon E5-2650 processor with an operating frequency of 2 GHz and a total frequency of eight cores of 16 GHz, the required memory bandwidth is 16 GHz x 1.6 GB / s / GHz = 25.6 GB / s. Such a bandwidth can be provided by memory with a frequency of 800 MHz (4 x 8 x 0.8 GHz = 25.6 GB / s). In fact, we protested that the processor from the memory of 800 MHz took away the same result, as from the memory of 1600 MHz - 340 cores. Moreover, from the memory of 1600 MHz in dual-channel mode (and 2 x 8 x 1.6 GHz = 25.6 GB / s), the test also saw 340 short bursts.

However, without delving into such complexity, you can simply change the memory at the maximum allowed frequency, moreover, for the variant of the memory modules 1333 MHz and 1600 MHz, they practically do not work. It is necessary to store all the channels of the controller with memory modules, the stink shards work in parallel. These recommendations guarantee that the memory subsystem will be the bottleneck of the system when this type of interference occurs.

Disk subsystem

The nature of the attack on the disk subsystem for an hour of waiting for the test at 100% cache base in the server's RAM and the daily blocking does not allow the operation of the visnovka to optimize the results of the test. For a significant amount of hard disks, the average recording speed of the disks did not exceed 40 MB / s, and the depth of the cherga became an insignificant value of the order of 0.05 - 0.1. In fact, under an hour of testing, a record is made in the transaction log in the sequential mode, which does not affect the performance of the controller and disks.

In real zastosuvannyah, obviously, the picture is different and the configuration of the disk subsystem can be a big impact on the overall productivity of the system.

Merezhev interface

In our server stand configuration, we combined 4 integrated gigabit network adapters into one aggregation channel for external client connections. Merezhevy traffic did not exceed the peaks of 30 MB / s to reach for the maximum number of carriers. Take advantage of the power and one gigabit tie adapter. In order to switch to this in practice, we tested the "maximum" configuration of the server with Intel Xeon E5-2697 v2 processors in a variant with one wired connection and took the same result from 880 "standard 1s short-range servers".

Testing the server in a configuration with two processors

When testing dual-processor changes, the same processor models were victorious, as in the single-processor version. The results of the test are shown on the graph.

When running with one processor, the test results increase more, the total frequency of installed processors is lower. The dual productivity when adding another processor is also not expected. It can be seen from the offensive table.

Equalization of server productivity with one or two processors

CPU model	E5-2609v2	E5-2630v2	E5-2650	E5-2650v2	E5-2660v2	E5-2697v2
Koristuvachiv, 1 CPU	140	310	340	410	460	630
Koristuvachiv, 2 CPUs	250	540	600	680	740	880
Increment 2 CPU / 1 CPU	79%	74%	76%	66%	61%	40%

The increase in productivity when adding another processor becomes from 79% to 40%, and the more powerful the processor, the lesser the increase.

The effect can be explained as follows. Take a look at the generation of Intel servers inspired by the architecture NUMA (N on- U uniform M emory A architecture). Within the framework of the architecture of the skin, the processor manages its area of ​​operational memory through the introduction of a new memory controller. Interaction of the processor with "one's own" memory is connected more, lower from the memory, connected to another processor, so that the transition to "foreign" memory is connected through the interprocessor interface QPI with the throughput building. The Intel Xeon E5-2600 processors can have two such interfaces with a total throughput of 32 GB/s, while the memory controller transfers data with a bandwidth of 51.2 GB/s.

Through those that take an hour to test, processors can beat both their own and "foreign" memory, more access to another one, lowering the overall speed of operations and not allowing "downgrade" productivity when another processor is installed. More than that, the tighter the processor, the more intensively the memory is broken, and the QPI interface is more powerful.

The maximum result, which we were able to achieve for a dual-processor system, is 880 standard 1C shorts versus 630 for a single-processor system.

How to win the test results for the server selection

In order to win the results of testing for the selection of the server configuration for the basic information base, it is necessary to determine the number of "standard core values" in order to create a real core information base.

Can you work in this way.

It is necessary to choose a period, by stretching some kind of coristuvachi, to create the maximum interest in the information base. Tsikh koristuvachіv need to classify for roles, like stench win, for example, comrade, sales manager just. For a skin role, it is necessary to create a process, as in automatic mode, viconuing a typical business process, typical for this role, that freezes the hour of the vikonanny. Then we will need to calculate the coefficient, which shows how many standard coristuvachs the role is given. The coefficient is calculated according to the following formula:

K \u003d (T / Tp) * (V / 60), de

T- Trivality of a typical business process of participation in a single-entry mode on an explicit setting

Tp- Trivality of the business process of "sales" of the "standard koristuvach" in a single cassette mode on a tangible basis

V- the number of business processes that are worthy of a high-profile role, for 1 year

Hour of vikonannya business process "sales", which vikonu "standard koristuvach", to lay the head rank in the frequencies of the processor and operational memory. Oskіlki operation vykonuєtsya in monopoly mode, the number of processor cores on the swidkіst vikonannya not vplyvaє. Our test has taken the following meaning:

CPU model	E5-2609v2	E5-2630v2	E5-2650	E5-2650v2	E5-2660v2	E5-2697v2
Max frequency TB (GHz)	2,5	3,1	2,8	3,4	3,0	3,5
Hour Tp(Sik)	1,32	1,10	1,24	1,06	1,16	1,08

Having calculated the coefficient for the skin role and multiplying the number of coristuvachіv tsієї roles by the її coefficient, you can take away the number of "standard coristuvačіv", so as to increase the number of coristuvačіv ziєї roles. Squeezing potim tsі meaning for all roles, you can take away the total number of "standard 1C coristuvachiv", as if creating a vanity that is equivalent to the real one.

PC and workstation configurators- A good way to choose a PC or work station for the type of task and tension. The great selection of components, pereverenikh on sum_snіst.

Server configurator and SRS– tse, vlasne, online-constructor of servers and data storage systems (DSS), "imprisoned" according to the specific needs of the client. You can choose from a large variety of ready-made models of computational systems, as shown on our website, and with the help of the designer you can independently choose the own version of the server or SRS on the basis of basic models. The configurators of the basic models are represented by such brands: HP, Lenovo, Dell, SuperMicro, Intel, Asus, Buffalo, Qnap, Synology.

See configurators

For the sake of clarity, our site offers five types of configurators, skins from which you can satisfy the needs of our clients. ix axis short description:

PC configurator

PC configurator is a handy calculator, you can choose a computer for your task. Computers, in the configurator, are divided into two main divisions: Office and Home. In the skin division, the PC was divided into series in staleness and tension. A leather series includes models in different cases and different types processors. Leather series PC may have a short description, so that you could better choose. All components are re-verified by our specialists for summ_sn_st. If you have a problem with your supply, you can always turn to a consultant on the site.

Workstation Configurator

Workstation Configurator hand tool for the quiet, who needs to receive the possession of the owner. Working stations are divided into two types of tasks: for work with graphics and for rozrahunkiv. In the skin series, there are options for working stations, which were divided according to the number of tightness. All configurations are stored by our technical specialists, and you can be assured of the totality of all components. With the help of the configurator, you can also improve the characteristics of the configurations that we propped up.

Software configurator

Handy software configurator to help you choose the software you need. Now you will not be able to shukati programs from the great catalogue. You can do everything in one vein, having only done a few clicks of a bear. Our facsimiles automate and simplify the process of selecting software as much as possible, based on two criteria - comfort and maximum functionality when working with a configurator.

Server configurator

This configurator of the best assignments for those who already have knowledge in the computer room. Tse can be fahіvtsі and IT-engineers, otherwise they are called “put in” coristuvachi. You can design servers for different distributions up to impersonal parameters (CPU; CPU memory; PCI-E slots; IO, SAS, RMM modules) that require the same knowledge. A wide choice of parameters allows you to most accurately adjust the required option that suits your needs.

SRS configurator

Tsey split allows you to configure DAS/NAS-attachment for power consumption. Having selected the basic model of the SRS, you complete it in your own way for additional tasks of the parameters. In the parameters of the configurator, you can change the number of disk storage devices, the skin capacity of them, or select storage devices for the size of the buffer memory. Before all the proponing of the basic models, a report description of their characteristics is added, for example, the stench will help you choose the SHD for such parameters as the form factor, the type of interface and many others.

Advantages of our configurators

· Selecting the complete set, you yourself are responsible for the versatility of the possession.
· Protect the impersonal hour that forces, as if they are being used for the needs of the model.
· Do not boast about those who chose the complete set inadvertently. Our fahіvtsі themselves zberat your nalashtovanny variant and protest yogo.
· You accept all necessary project documentation and certificates for your possession.

• Servers at the counter

  For form factor:

  • Servers 1U
  • Servers 2U
  • Servers 3U
  • Servers 4U/Tower

  By processor platform:

  • On Intel processors
  • On the AMD processors

  For authorities:

  • GPU servers
• Sublog servers

  For form factor:

  • Tower
  • mid-tower
  • mini-cube

  By processor platform:

  • On Intel processors
  • On AMD processors

  For authorities:

  • Low noise
• Servers and savings

  For form factor:

  • 2U storage servers
  • 3U storage servers
  • 4U storage servers

  For controller type:

  • With hardware SAS RAID
  • With software RAID/HBA

  For disk form factor:

  • On 3.5" rims
  • On 2.5" rims
• Working stations
  • Low noise
  • highly productive
  • For help GPU
• Bagatonode servers
  • 1U (2 nodes)
  • 2U (2 nodes)
  • 2U (4 nodes)
  • 3U (8/12/24 nodes)
  • 4U (2 nodes)
  • 4U (4 nodes)
  • 4U (8 nights)

Server configurator

Server configurator

Server configurator for the implementation of all representations on the server site. Below are our recommendations on how to show them the basic knowledge of server topics.
Even if you are well oriented and without our recommendations, the axis of the screws is placed on the main sections, so that the robot can be identified:

Navigation

What is the server configurator and what are the required wines?

Server configuration change - technically foldable, the correct version of what may be on the basis of knowledge of the model range and platforms and components. The current nomenclature of Supermicro platforms (on Supermicro itself, we will be our solutions) today will cover almost 100 positions for different tasks. If you choose not to be surrounded by platforms: the config configuration allows you to combine cases and motherboards, the range of which is even wider. Often vybіr oskladnyuetsya obmezheniem budget, with tsomu tekhnіchnі vymogi to the server for vіdpovіdnіstyu taskіnіhto not skasovuvav. Rise independently with such an obsyaz technical information that price is not easy. Let's go from traditional training to a qualified specialist for the help of a telephone call or e-mail, OSK Techno offers an on-line tool, which, for the obviousness of basic knowledge of the subject, allows:

• Independently choose the optimal platform option as the basis for the system based on key characteristics.
• Correctly select the components of the possible server for the chosen platform.
• Form a price for a ready-made server, save an individual configuration, or immediately replace the server. Registration is registered without registration

Server configurator for implementations of all current Supermicro platforms, and navigate for OSK server line.

Supermicro Server Platforms – The Basis of a Sustained System

Let's look like the vlashtovs model range Supermicro. Basic components of the system є:

• Case (Chassis / Chassis) - the physical basis of the system, which includes baskets for installing hard disks, backplane (false model), living unit (s), rails for mounting at the counter (false model), cooling fan (s) mounting accessories. Enclosures may have part numbers that are based on the CSE prefix (for example, CSE-213AC-R920LPB). The model range of cases is divided into series according to their main characteristics: form factors in the case itself and hard disks (for example, the case can be used up to series 213: for installation in a 19” rack, 2U height, 2.5” form factor disks).
• Motherboard (Motherboard) - is supported by the MP of the stellar desktop icon by zastosovuvannye technologies, which are supported by components, the class of integrating outbuildings, the number of components and the variety of expansion cards that are connected. All points of authority are shaped by the peculiarities of the minds of exploitation and the tasks that stand in front of the server. Supermicro motherboards have part numbers that start with the MBD prefix (for example, MBD-X10DRC-LN4+). The MP model range is divided into the category of processor sockets: the choice of the processor of that series (for example, the MP is brought up to the category of dual-processor sockets for installation in it Intel processors of the E5-2600 v4 / v3 series).

Promoted to choice and configuration of the platform є factory selection of cases motherboards. The Supermicro platform nomenclature has SuperServer/SuperStorage and may have part numbers prefixed with SYS/SSG (for example, SYS-2028R-C1R4+). New rules You can look over the platform models. At the transfer of the components of the skin platform, it is clearly indicated that the case and MP are included in the warehouse (the platform is set in the butt of the choice of the case and MP, which was also pointed in the butts). Yakschko in the transfer of proper platforms of the municipality of the Vіdpovіdna Rishennya, Maє Sens is brought out to SP_V_SK TECHNO for Formuvina Іndivіdivnya Specitіkatsії at the basis of the Kombіnatsii MP Ta Coruscia: Mozdlyostі such Combinumanna Posіntly Shirshі, Nіzh Standard Model Row of Platforms Tom Zdatni Titted Vishennia Personally beware.

Varto signify that high-powered servers (Ultra, Twin, MicroCloud series) are no longer used as a platform: their layout in the case and the MP is rarely practiced due to technological features, the availability of components, or it is impossible to be out of the box. Moreover, a number of technically folding platforms are supplied only in a complete set, which does not come with a minimum specified picker. Such platforms may be marked with the “Complete System Only” sign from the designated minimum configuration (processor, RAM, disks thin).

Which processors to choose? Key moments

The table of series of server processors has been created
	Intel						AMD
	xeon E3-1200v6	xeon E5-1600v4	xeon E5-2600v4	xeon E5-4600v4	xeon E7-4800v4	xeon E7-8800v4	Opteron 6300
Name code	Kaby Lake	Broadwell	Broadwell	Broadwell	Broadwell	Broadwell	Abu Dhabi
Number of cores	4	4 - 8	4 - 22	8 - 22	8 - 16	4 - 24	4 - 16
base frequency	3.0 - 3.9 GHz	3.2 - 3.7 GHz	1.6 - 3.5 GHz	1.8 - 2.5 GHz	2.0 - 2.1 GHz	2.1 - 3.2 GHz	1.8 - 3.5 GHz
Pidtrimka rich processor	One- processors	One- processors	Up to two processors	Up to four processors	Up to four processors	Up to eight processors	Up to four processors
A type support memory	DDR4 2400 MHz ECC	DDR4 2400 MHz ECC RDIMMs	DDR4 2400 MHz ECC RDIMMs LRDIMM	DDR4 2400 MHz ECC RDIMMs LRDIMM	DDR4 2400 MHz ECC RDIMMs LRDIMM	DDR4 1866 MHz ECC RDIMMs LRDIMM	DDR3 1600 MHz ECC RDIMMs
Max. obligatory memory (per processor)	64 GB	1536 GB	1536 GB	1536 GB	3072 GB	3072 GB	384 GB
Number of channels in memory	2	4	4	4	4	4	4
Budova graphics	For some models	-	-	-	-	-	-
Number of PCIe lanes	16	40	40	40	32	32	42

The main characteristics of processors that determine the choice:

• Number of cores (threads)
• Clock frequency
• Number of processor sockets required

Moreover, in the okremi vipadka:

• Characteristics of operative memory, which is considered (type, volume, frequency, etc.: reports about those, how the characteristics of the processor and VP are shown in the “Random Access Memory” section)
• The presence of the video

To the right of your respect is a table with the main parameters of server processors. Most of the processor market is occupied by Intel products. AMD has recently updated its line of server processors - the first time in 5 years, while for the current class the Opteron 6300 is still left with good decisions. Until the delivery of new Intel Xeon Scalable and AMD EPYC processors to the Russian Federation has not been adjusted, the stench is in the whole table (div. below)

We won’t be able to see desktop processors, however, we can say that a number of single-processor servers and workstations support the installation of Core i3 / i5 / i7 series processors or Atom processors may be installed. As a rule, servers on such processors are suitable for simple tasks, for example, an Asterisk server or a border screen.

The cob line of Xeon server processors is the E3 family. Qi processors are powered by a small number of cores and high frequency; vstanovlyuyutsya strictly single-processor systems. The dual-channel mode supports DDR4 ECC memory with a frequency of up to 2400 MHz and a volume of up to 64 GB. Servers on such processors, as a rule, run like a server email, merging screen, domain controller, WEB-server, server for saving cob level (file-server), server backup chi archiving.

The great homeland of E5 processors is divided into a sprinkling of lines. E5-1600 v4 - processors with high frequency and up to eight cores for installation in single-processor systems. In the choti-channel mode, DDR4 ECC Registered OP is supported (LRDIMM is not supported!) Up to 1.5 TB.

The most and most popular E5-2600 v4 line consists of decal categories: Basic, Standard, Advanced, Low power, Frequency optimized, Segment optimized, Workstation and Low Power. E5-2600 v4 can be played both in dual-processor and in single-processor configurations. In the chotyricanal mode, up to 1.5 TB of DDR4 ECC Registered/LRDIMM RAM is supported. The range of tasks deployed on servers with E5-2600 v4 processors is even wider: from simple tasks to a domain controller's desk to super-productive solutions to the class of great business.

The E5-4600 v4 line processors are powered by a low frequency, but by a large number of cores, moreover, they are selected for installation in different processor configurations. Xeon E7-4800 v4 and E7-8800 v4 also win for rich processor solutions and are compatible with E5-4600 v4 for multiple QPI links (3 versus 2). Application areas: highly productive systems of corporate level, including for organizing virtual environments, machines for calculating models of molecular chemistry or seismic and other.

AMD processors in the server segment are currently represented only by the Opteron 6300 line (on the way to EPYC 7000 - follow the updates on the site). We guess that the Opteron 6300 has no technology similar to Hyper-Threading.

The filter server platforms have a series of processors that can be installed, as well as the number of processor sockets in the system, the introduction of different selection parameters, depending on the points in which it is possible to differentiate the required platforms according to the nomenclature for further selection parameters for other:

A table of new series of server processors has been added
	Intel Xeon Scalable				AMD EPYC
	Bronze	Silver	Gold	Platinum	1 socket	2 socket
Name code	skylake				Naples
Number of cores	6 - 8	4 - 12	4 - 22	4 - 28	16 - 32	16 - 32
base frequency	1.7 GHz	1.8 - 2.6 GHz	1.9 - 3.6 GHz	2.0 - 3.6 GHz	2.0 - 2.4 GHz	2.0 - 2.4 GHz
Pidtrimka rich processor	Up to two processors	Up to two processors	Up to four processors	Up to eight processors	One- processors	Up to two processors
A type support memory	DDR4 2133 MHz ECC RDIMMs LRDIMM	DDR4 2400 MHz ECC RDIMMs LRDIMM	DDR4 2666 MHz ECC RDIMMs LRDIMM	DDR4 2666 MHz ECC RDIMMs LRDIMM	DDR4 2666 MHz ECC RDIMMs LRDIMM	DDR4 2666 MHz ECC RDIMMs LRDIMM
Max. obligatory memory (per processor)	768 GB				2048 GB
Number of channels in memory	6				8
Budova graphics	-
Number of PCIe lanes	48				128

The table below shows the relative characteristics of the new lines of server processors announced by Intel and AMD in the middle of 2017. It can be seen from the tables that new line Intel Xeon is called to replace the current Xeon E5 v4 and E7 v4 series on the market. AMD at the presentation announced the superiority of EPYC productivity over E5 v4 up to 70% fallow in the i segment, perhaps becoming a good competitor to that Xeon Scalable line. Supermicro has already introduced updates to its server lines on new platforms that support Intel Xeon Scalable and AMD EPYC. The company OSK Techno is updating the model range of proponated servers presented on the site and forming quotas for placing new products so that our managers can estimate the budget and buy a new generation server already today. detailed look We have seen new processors and server platforms in the same article.

In the on-line configurator of the implementation servers, automatic control of memory types is available depending on the platform: Necessary obsyag OP in the server is assigned to the managers / winning programs.

• When choosing the memory bar of the VP, the number of the memory carded by the head of the memory card, the following is the number of memory slots in the system (for building up in the future, as well as zastosovno), as well as the number of memory channels in the system: for optimal productivity, the memory needs to be repeatedly replaced channels (for example, for systems based on E5-2600 v4 processors, the optimal mode is when the OP bars are installed in a number that is a multiple of four per processor)

• Table of frequency occurrence of OP by type and type and installation method for E5-2600 v4
  VP type	VP ranks (Ranks)	Plank capacity		Speed ​​(MT/s); slots per channel (SPC); Plank per channel (DPC)
  				1 slot per channel	2 slots per channel		3 slots per channel
  				1 DPC	1 DPC	2 DPCs	1 DPC	2 DPCs	3 DPCs
  RDIMM	SRx4	8 GB	16 GB	2400	2400	2133	2400	2133	1600
  RDIMM	SRx8	4 GB	8 GB	2400	2400	2133	2400	2133	1600
  RDIMM	DRx8	8 GB	16 GB	2400	2400	2133	2400	2133	1600
  RDIMM	DRx4	16 GB	32 GB	2400	2400	2133	2400	2133	1600
  LRDIMM	QRx4	32 GB	64 GB	2400	2400	2400	2400	2400	1866

  As a task, which is kept on the server, the highest frequency of the operation of the operative memory is critical (database server, 1C server, etc.), pay attention to the support of the processor tobto. like, for example, when installed processor E5-2620 v4 put a 2400 MHz VP, її the operating frequency will be 2133 MHz itself through the interface on the side of the processor. Of course, there are no other factors that affect the operating frequency of the airfoil: the number of bars, their placement (SPC, DPC), type, rank and nominal frequency. For example, for dual-processor MPs on the E5-2600 v4, when installing memory, you can use the following table:

• For systems based on E5-2600/E5-4600/E7 processors, it is allowed to install the LRDIMM: memory type with reduced electrical power on the memory bus, which ensures high capacity without reducing the operating frequency. Zastosovuєtsya, if it is necessary to secure a great deal of OP at the maximum frequency (for example, a database server for the creation of a ram-disk).

Server Disk Subsystem: A Brief Overview and Recommendations

The disk subsystem consists of three elements:

• Backplane - backplane board, with sockets for connecting hard drives (SATA / SAS / NVMe) from one side and the controller from the other. The backplane enters the warehouse building i, fallow in the model, can support different interfaces of the accumulators. Backplane can mother expander(s) - it allows you to connect more disks, lower ports from the side of the controller. V simple systems backplane can be daily: disks are connected directly to the controller.
• Disk subsystem controller - for integrating on MP attachments or implementing a fixed card into a PCIe slot, which allows you to handle the disk subsystem of the server. There are different implementations of the controller: - hardware (“hardware”, hardware, HW) okremiya processor for the form of RAID logic and special functionality that supports it, - software (“software”, software, SW) activate the host resource of the central processor and HBA (Host Bus) Adapter), controller that translates commands to disks without organizing RAID arrays
• Drives: vary by type (HDD / SSD), interface (SATA / SAS / PCIe), form factor (3.5” / 2.5”), volume, speed parameters (rpm / IOPS) and int. Configurator, disks are classified as corporate and are approved for operation in 24x7 mode.

To other characteristics of the disk, you can see the availability or the availability of hot swap disks.

For otrimannya transfer of systems required characteristics In the disk subsystem, check the "disk drive" and "disk form factor" sections of the filter, as well as the "authority" section in the part that should be used in SAS RAID/HBA:

If the system does not have a built-in SAS controller, if it is necessary, you can add it with a different card at the system configuration stage.

• The type of controller, disk space, necessary security and other parameters depend on the nature of the task. Choose RAID, as it is necessary to ensure the capacity of the disk subsystem independent of the OS, or as the software/OS may have a little bit of totality in the part of the SAS controllers (VMware ESXi). HBA is used for systems with software / operating systems, which means direct connection to disks (Microsoft Storage Spaces, VMware VSAN) or for connecting external systems to save data.
• Choose, secure the possibility of a further expansion of the disk subsystem. For a simple stockpile of cables, one of the expansion options is to install a SAS HBA controller with name ports (SFF-8644) or cascading from the backplane, which can also expand the SAS name for a further JBOD connection.
• If you choose a RAID controller with a cache, it is recommended to include the cache protection module in the specification at once (also called "battery" is no longer a relevant term) - you can not use the cache instead of the cache in case of failure in the controller's work mode to write to the cache.

• Pay attention to the possibility of integrating add-ons, such as SAS HBA/RAID, 10Gb/s controllers (the right column of the server platform filter): For example, as a matter of fact, it depends on the presence of a full-fledged striped RAID controller and mesh connection 10Gbіt / s, V_dzdnіdnі Подениi, vi otrimіть приліка минировый platforms, in the pleasant ci attrolation, Yakschko і інданки: at once, Yakschko Svіdishkovі Іrteg "Svіdomo is not stopped, Sch_d Vuschi Z Vіantіv, Scho Dyshtiy їSh плат Place: Reflecting for an unpleasant attachment.
• With installed processors in the number of smaller, lower number of processor sockets (for example, one in a two-processor configuration), may be able to install the rest of them in the future, next protect a few important points: the memory slots of empty processor sockets will not be available; fallow vіd specific model MP may not be available for integration of attachments (such as a SAS controller, a 10Gb/s network controller, etc.) and expansion slots (PCIe, NVMe slots). You can clearly see behind the structural diagram of the MP, on the image of the PCIe bus, which links the processor socket with the integration of devices. For a detailed consultation on nutrition, it is recommended to contact the specialist of our company.

Krіm already described parameters, the filter allows you to see the system behind such authorities:

• NVMe support - servers, ready to install storage SSD with a PCIe interface in the form factor U.2 On the front side of the primary SSD and disks, they show the highest productivity of the bus over the PCIe bus instead of the SATA / SAS interface.
• GPU support - servers optimized for installing GPU cards (Tesla, GTX, etc.) Such servers are stationed in these blinds (quantum chemistry, molecular dynamics, visualization, rendering, seismic, etc.)
• The sub-unit of life is servers, in which the super-world of life is implemented to increase the stability to the spirit.
• Small depth - servers, which can be installed in telecommunication racks or cabinets of small depth from fastenings behind the ear on one 19” flat without rails
• In case of 4-port LAN 1/10 Gb/s, 2-port LAN 10 Gb/s - for these powers you can choose a server, which can be integrated with a network of controllers, in addition to a 2-port gigabit, which is standardly equipped with a server

Today we will look at the choice of a server "hall" for a small organization for 25-30 employees, with a separate infrastructure (trade points, a warehouse), which kind of terminal server is needed and the program "1C: Enterprise". These services serve all practitioners.

More small companies, for cheaper valets, will allow minimizing the amount of equipment that they buy, and asking administrators to "pump" all the services they supply into one physical server. Bazhannya is more sensible and provocative, but here there are nuances.

You can organize a terminal server and win the file version of 1C there, but for such a large number of coristuvachiv, the retail company recommends switching to the client-server version. That's why we need another server called "1C: Enterprise" and a database server. Let's clarify how to organize a terminal server, SQL server and 1C server on one operating system Possibly, ale, with a look at the safety and stability of the robot services, it is not recommended. And if you still want to win one physical server for all three roles, then it is recommended to win virtualization, for example, VMWare ESXi or Hyper-V.
In this order, there are three options:

1. One server from file 1C. Poor variant, we can't see it.
2. One server from two virtual machines.
3. Two physical servers, one terminal, the other from the database and 1C.

To complete these tasks, you can request the next server configuration:

Wu vipadu z one physical server we reuploaded the choice on a Dell R710, with two six-core Xeon X5650 processors, 64 GB of RAM and six drives: two SSDs for RAID 1 and two SAS drives for RAID 10.

Wu vipadu z two physical servers We've zipped vibir on the following configurations:

• Terminal server: IBM x3550 M3 with one Xeon E5620 processor, 32 GB of RAM and two SSDs in RAID 1, with an additional tiering card for two gigabit interfaces. This server also has a lot of room for upgrade, two-processor oscillators, can have 18 slots for memory modules and support up to 288 GB of RAM.
• Database server: IBM x3250 M5 with one Xeon E3-1220v3 processor, 16 GB RAM, SAS/SATA add-on RAID controller with RAID 10 SAS disks, add-on 2 GB tiering card.

Why did we choose such configurations? For the purpose of providing food, it is necessary for us to provide comfortable work for our small organization for 25-30 employees. So that there was no confusion: only one of the examples of inexpensive zastosuvannya 1C, and in rich ways, choose other changes.

Processor

From a glance of the processor hour, terminal sessions take up a small part. For the provision of terminal solutions in various organizations, for support of comfortable work 30 cores, 4-6 physical processor cores will be enough, one core for 6-8 sessions.

For a small database, SQL Server needs one core. But let's focus on the expansion of the database in the future (or increase the number of databases) and take two cores in SQL.

For the 1C: Enterprise server, the number of cores is not so important, as clock frequency that bus frequency. Therefore, we put two more cores on the 1C server.
And let's not forget that in times of virtualization one or two cores we need to secure the robotic host operating system.

Together we have to enter:

• For a server with two virtual machines, 12 physical cores are required. It can be less, but it’s always to blame for the supply of tension. A server with two six-core processors is ideal for one.
• for a terminal server, one Xeon E5620 processor with six cores is sufficient, for a database server - a Xeon E3-1220v3 processor with four cores.

Working memory

We wonder a little, how much operational memory is needed for services:

• Operation Windows system Server only needs 2 GB of RAM.
• For SQL, that small base 1C will be enough for 4-6 GB of RAM.
• The server "1C: Enterprise" has 2-3 GB of RAM.
• Rozrakhovuєmo, scho skin koristuvachiv need 700 MB of RAM in the terminal session, then 30 koristuvachiv need 21 GB.

Now zastosuєmo tse up to our options.

• For one server and two virtual machines, about 40 GB of RAM is required.
• For a terminal server, 24 GB or 32 GB of RAM will suffice (it will be taken from the reserve, allowing for future expansion). For a server with data bases, at least 8 GB is required, but 16 GB in reserve. Memory at a time is one of the found server components.

Disk subsystem

Ce traditional plyashkov neck bagatioh systems. Correct choice hard disks are important for securing swedcode servers. When working with 1C with a SQL base, there are impersonal read / write operations per second (IOPS). How do you work on the terminal server from thin clients (that's why the terminal server is completely corrupted like a working medium), it strongly interferes with the disk system of the server. For example, 30 terminal server boxes on RAID 1, SATA 3 Gb/s, with WD Velociraptor disks will feel uncomfortable for an hour of work while actively surfing the Internet. For terminal servers, we recommend that you use SSD drives. For database servers - SAS-disks, selected in stack to drive array.

Krim accumulating, slid to give respect to the disk controller. Modern servers can do good controllers on board, for example HP SmartArray and DELL PERC. However, it is not correct to choose the "onboard" solution in case of serious ambition, if the maximum productivity is required. Having spared the troubles, you can easily take away a hard server, which is not a burden. The controller is to blame for this, but the hardware, not the software, because of its independent memory.

Let's take a look at the options for rozvyazannya of this task.

• For one server with two virtual machines, it is necessary to create two RAID arrays: on one, the files of the virtual machine of the terminal server will be stored, on the other, the files of the virtual machine of the data base server and "1C: Enterprise". For the creation of the first array, it is best to vikorate two SSD-accumulators in RAID 1 (mirror).

  Another array is better created from two SAS disks in RAID 10 (mirror + stripe), or it is possible to use two SSD-storage disks in RAID 1. Choose to store only a few disks of that server model.

• For two servers, all the same, only arrays will be separated by servers. On the terminal - RAID 1 of two SSDs, on the database server - RAID 10.

One or more servers

As it is said above, small organizations should have a great opportunity to spread all the services on one server.

Until the victory of one server, that virtualization can be reduced to a lower power supply and more resources have been allocated between virtual machines. That porting of virtual machines, in a different way, is richer than the porting of physical OS.

However, two servers may have more room to upgrade. For example, in our version, an inexpensive IBM x3550 M3 with the addition of one more processor and RAM is transformed into an elegant short terminal server for 50 and more cores.

One more "high school" in our mind, as it is necessary to be safe when choosing two physical servers, the exchange of data between them is as good as possible. For virtual servers, data exchange goes through a virtual switch. Right there, for the benefit bandwidth Merezhi, you can install in a skin server using a single card with two gigabit interfaces, so you can aggregate between yourself and without intermediary connection between servers and aggregated 2 gigabit links. Otherwise, you can try different cards with SPF + 10GBASE, but on the way you will be satisfied.

Reserve for pushing

When rozrahunka that choice of server is necessary to take to respect the peak of interest. Also, it is necessary to keep in mind that the data base will be less “swollen”, the data on the terminal server will grow, and the number of cores can be increased. A lot of undertakings save on stocks of tension and for pіvroku-rіk stick with interruptions in work and skargami koristuvachіv. For that change, if the transcendental economy is to bring up new expenses for the future, it is stingy to pay two. We have chosen the options of insurance coverage from the reserve of tightness and the possibility of an upgrade. Guaranteed that DELL R710 can add two more hard drive and RAM, as well as replace the processor with more productive ones.

So, if you need to expand, or the number of services will increase, then there are great prospects, and long hour Effectively meet your goals. Possibly, through the river, we will not be able to increase the number of coristuvachiv vdvіchі, up to 60 osіb? Please.

If you win one DELL R710 server, then you can buy an inexpensive IBM x3550 M3, upgrade to a new hypervisor, transfer a virtual machine with a database and a 1C server to it, and all the resources to DELL virtual machine from the terminal. Tse will be quick, and it will not be necessary to “throw everything away and buy new”.
Well, well, you won two IBM servers, then the x3550 M3 with the addition of another processor, that small amount of RAM is transformed from a middling machine to an exhaustive machine. And in the x3250 M5, you can upgrade the processor from E3-1220v3 to E3-1285v3.

Visnovok

Obviously, proponated configurations are far from the only options for possessing, for example, for the implementation of the same 1C. Here it’s already rich to lay a lot of money in the streaming infrastructure, equaling the profitability of that її zbіlshennya from the future, and navit the budget. Therefore, you can always choose either similar models of servers, or more expensive ones.