How to optimally configure the 1C server?

Question

1C

Good day!
In production, a 1C-enterprise server was down (there was a user system unit on i7 and 32 RAM), the authorities gave the go-ahead to do "everything is right and kosher." Now everything rests on a regular system unit with i5 and 16 RAM, where the SSD is for the base and that's it.
The extreme budget figure is 300,000-340,000 rubles.
I would like to ask for advice on the proper organization of the 1C server and MS SQL with a peak load of 50 clients, but in the future there will be more users.
Variants that have been studied, without minor technicalities. iron details.
1. One server, where the database is SQL and 1C, SQL on SSD in RAID1, the rest is RAID10.
2. Two servers. The first under 1C, the second - SQL.
3. Cluster of 2 servers. I haven’t come across anything like this, is it possible to assemble a productive solution with my budget?
Everything goes to the fact that I organize 1 solution, but for sure there is a more elegant option, correct, most productive and fault-tolerant.
I would love to hear your experience/advice/recommendations.
Thank you.

Answer 1

> Cluster of 2 servers. I haven’t come across anything like this, is it possible to assemble a productive solution with my budget?
A cluster of two servers is not optimal. There must be a quorum. (those three)
For the cluster, you also need storage, and then the point of failure will be this same storage.
Of course, this would be the most correct option to do a cluster, but it is very expensive.
From my own experience, it is good to drive clients into terminal sessions. Here you need to look at how much the 1C network is loaded and how slow everything is.
It is better to use SSD + hdd in a hybrid using storage space technology, you can also make a software raid there.
Completely workable item.
In the end, I probably would have done this, and many here will twist at the temple and say that I'm crazy and full of loss of productivity. But with your 50 people and more, these are trifles.
Get a good server. Under the system, take 256gb disks in the number of three. Make a mirror + hot spare.
The rest is to finish the SSD + HDD at the rate that for about 1 tb of data, it is better to take 256gb of SSD cache. (at least).
Using Microsoft StorageSpace to make Raid + caching.
SSD is better to take based on the mirror.
Create two virtualization machines on Hyper-V. What for ?
1) It's more convenient to backup the whole machine at once
2) If everything dies, it will be easier for you to deploy these machines anywhere, just drag them where you need them.
3) If suddenly you run into performance, you can easily spread and cluster it all.
You don't have those loads, for example, to create a separate ClusterFileSystem (it's true that only MSSQL and Hyper-V machines can be stored there)
The budget is really not very clear, is it on a server from a license?

Answer 2

The above is true, you will not have enough for two servers.
Your option is all on the 1st. Don't take the first result on google
as an advertisement , here is the price configurator. -The processor will be expensive (the higher the frequency, the better) and SSD drives (you need to choose based on what users do more often, write information to the database or read it, or according to the most critical resource-intensive operation). -You can win 1-2 tr on RAM, it can be critical with your budget)), play around with 4x16 or 2x32 bars and you will see the difference.
-If you are later, but it must be for sure, they will be able to give money to purchase additional. components (argument with the growth of users in the future), then first take a server with 1 CPU but the possibility of installing a second one, with 1 power supply but with the possibility of installing a second one, for example 32 GB of RAM (not enough of course but tolerable) later expand to 64 GB minimum. Under the base you need at least a RAID1 SSD, you can later add and rebuild.
-Further on the hardware, if there is a high criticality of data loss, then a RAID controller with a cache and a battery)) is a must.
-1C is not able to parallelize queries in the database! Therefore, you need to chase not the cores, but the CPU frequency (only if you are not a 1C mega programmer and know how to write queries directly to the MS SQL database (MS SQL itself "parallel" them perfectly)).
-The database should be on a separate disk, tempdb should be placed on a separate disk, the database logs should also be placed on a separate disk (we are talking only about SSDs), since the last two are very intensively exchanged.
-But the OS can be installed on ordinary HDDs (of course, preferably on 1 RAID).
-Network interfaces are usually at least 2 x 1 Gb / s, can be combined using NIC Teaming, there will be 2 Gb / s (look at the current load, you may not need more).
-Why 1C server and database on the same server? All because of the cherished Shared Memory protocol (shared Shared Memory and 1C).
Ask if there's anything else you're interested in.

Answer 3

Literally a week ago I made myself the first option - a budget one.
- Server HPE DL120 Gen9 operative 64, two power supplies.
- 240Gb Intel SSD (Interprise series, those for databases, with a large write cycle per day) - 2 pieces in a mirror under Windows server 2016
- 240Gb Intel SSD (the same) - 1 piece, for placing MSSQL database files.
- HDD WD RE 1Tb - for temporary database backups.
Works like clockwork, loads instantly. Users are happy, 45 people work.

Answer 4

For 50 users, 1 server will be faster. Well, to ensure the pace of the service user 1c and / srvinfo / also on fast disks.

Answer 5

It's a normal configuration, but you'd be better off having a second SSD in reserve in case your MSSQL database or system drive fails than keeping it in RAID.
And make copies to an external device.