Despite the fact that the program complex 1C has existed for quite a long time, still around it constantly boil battles on forums. Many people believe that several HDDs and a formed hardware or software RAID is enough to ensure full functionality of the software system, others prefer to use faster drives. Today we’ll talk about SSD for 1C server, and to make it easier to navigate, we’ll form a small list of requirements that we’ll focus on:
software is used by all employees, sometimes at the same time;
interacts with the website and other software, e.g. uploads price list and prices;
Let’s not go into technical details, we are looking at SSD for 1C server;
Postgre or MSSQL is used as a database;
two servers are used when the load is heavy.
If the users are from 5 to 10
Standard set for a small business, several people work with 1C constantly, others occasionally, because the scenario here is extremely simple and does not require excessive costs. Several possibilities, however, can be highlighted:
Use SAS hard disks. Expensive and useless. Even though it is faster than SATA, it is also twice as expensive. Building a mirrored RAID will eat up a lot of money without much benefit. This is the era of SSD, hard disks for 1C servers are no longer relevant.
Using SATA HDDs. Mirror RAID will be extremely slow, RAID 10 will require 4 hard disks. Also expensive, and the “bottle neck” in the form of a disk subsystem can sometimes make itself felt. And yes, RAID 0 has low fault tolerance, so if the drive fails, your data will be damaged, possibly irretrievably.
The interface is still the same SATA, but we use solid state drives. They are noticeably faster and are suitable for building a mirrored RAID, which will maximize performance and data security. It is better to use TLC or 3D NAND QLC memory, the price will be reasonable.
NVMe interface. Very fast but quite expensive would be an unwarranted waste of money in this case. The previous configuration will be sufficient for these tasks.
Now let me tell you about the types of memory used in SSDs. MLC has at its disposal about 3000 rewrite cycles of an individual cell, then it is highly likely to fail, read/write speeds are also quite high, about 600 MB/s for reads and 200 MB/s for writes. TLC is practically the same in terms of speed parameters, but reliability is noticeably lower, a single cell lasts about 1000 rewrite cycles.3D NAND is noticeably more reliable and higher in speed, while allowing more memory to be crammed into the same drive form factor. 3D NAND TLC has 3 to 10 thousand rewrite cycles, 3D NAND TLC has 3 to 10 thousand rewrite cycles, and MLC 3D NAND approximately 10-15,000. You can find out more information here.
To kill a SSD like this quickly, you’d have to completely fill and wipe it several times a day. Otherwise, it will quietly last for several years, if not more. Modern SSDs are very reliable.
But do not forget about the NVMeinterface . Architecturally flash memory with this interface has no differences, but the point is that the SSD performance is limited by buses such as SAS and SATA. NVMe, unlike previous types, connects directly to the PCI-E slot, which leads to lower latency in responding to requests to the drive, and also structurally PCI-E has more lanes to pass data, which leads to a significant increase in speed. As an SSD for 1C server for 5-10 users in the use of NVMe makes little sense, because there are no such serious needs in speed, but the price will bite, and seriously.
The most cost-effective investment is a simple SATA SSD with 150 TBW or more or 0.5 DWPD. Note the following companies: Crucial, Micron. They put out some pretty good products at not the most expensive price.
There are also some good solid state drives from kingston. I love this company with all my heart.
If the users are from 10 to 30
In this case HDD will definitely turn into a bottle neck, so it is not recommended to use hard disks. A SATA or SAS SSD, on the other hand, will cover your needs to the fullest. The greatest efficiency will show NVMe-drives, but such SSDs for 1C server with a decent volume may be unreasonably expensive.
Your task is as follows:
- to estimate the potential size of the future database;
- calculate the scope of the application and operating system;
- Allocate a sufficient number of disks for the planned RAID.
Typically, organizations with this many users have at least 100 gigabytes of active database capacity, but rarely exceed 1 terabyte. RAID 1 for data integrity requires 2 drives, so look at the following models:
- SSD from Crucial with up to 1 terrabyte, costing around 5-7k.
- SSD from Kingston. Volume is the same, reliability is about the same. No different in price than the previous option. So choose according to availability.
- Samsung SSD. The price is noticeably higher, reliability is also a class above. If you have the extra money, Samsung is the best choice.
Of course, this is just an offhand guess, you can choose a different volume and also other models, but make sure TBW is not below 150, higher is better.
SSD for 1C server for 30-50 users
Good old SATA will still work, but increase the task to increase not only reliability but also speed, so calculate so that you can build RAID 10, which requires 4 drives. You can expect a lot of mixed accesses to the disk subsystem, a lot of users are able to seriously overload the hardware and even on SSD to create a bottle neck. Therefore, either use SATA on RAID 10, or there is an option to opt for multiple separate servers and a separate storage system for the database. But this option is very expensive.
If speed is very important, build your disk subsystem on NVMe interface, look out for Intel and Samsung SSDs with M.2 interface. But for 1C server this solution may be redundant. Of course, there are better models, but the price will get much higher.
50-100 users
The situation becomes an order of magnitude more dire. We will have to make a distributed database, which will be divided into an active part (which is constantly accessed) and an archive part (kept in reserve, because it may come in handy). The drives must be performant and reliable. Preference is better to give NVMe-drives, or use SATA, but on several servers, which will later combine the data into a single pool.
100 users and more
The system is pretty much the same, but will need performance SSDs for the 1C server on each server. It’ll do Intel Optane, the drive is expensive but very reliable. Also can use Intel SSD DC P4500 Series, it has a resource of 1, 34 PBW, the price is certainly high, but I think for a company with so many 1C users it will be acceptable. The best option is to use a cluster with at least two devices. More is better, for both an application server and a database server are required. Each has its own arrays, programmatically pooled together.
Conclusion
As you can see, choosing an SSD for a 1C server can be difficult. But it’s pretty easy, more users – higher reliability of the individual drive. To get the speed you will have to build complex RAID, sometimes using clusters of servers. If the task is so complex, it is better to hire a competent expert or contact the professionals whose link I left above.
