OS and FS for a small file hosting service and 1C accounting v7? Jump off kubuntu?

Question

Windows

What are the pros and cons of win server 2012 r2 + win ser 2016 + ws 2019?
(newer versions have more stability and higher security? and if ram = 8gb, then win ser 2019 makes sense to install for 10 users 1c v7 + doc / xls 200gb file storage for 30 machines in the office)

For example, you can both win10 and win put server 2019 on the screw at home (core 2d + nvidia), and then bring it to the system unit to work (i5 + intel gpu) ... will Windows itself deliver new firewood? or will not be loaded or will it be buggy?

Is there any point in making a file hosting service in a small office on Linux? (otherwise they got the viruses on shared disks).

Which Linux is better for file storage? (debian or kubuntu)?

It was on kubuntu ... You need to change the screw from sata 800gb to 1tb.

It had an OS and a database and a file dump... 10 years...
More problems with electricity...
On the file dump, viruses in the windows of teapots write exe pif balls to the root every minute at 100 kb.
Sayliti win32 for example...

Somehow it's very tight with cabbage on the license...

Answer 1

Time with payment of licenses at you hardly - what in general Windows? Install ReactOS, suffer. Pirating at work is not only highly punishable, but generally fu.

Answer 2

Who knows what are the pros and cons of win server 2012 r2 + win ser 2016 + ws 2019?

By and large, it doesn’t matter, I still use 2016 everywhere.

Will Windows deliver new firewood itself? or will not be loaded or will it be buggy?

She will put it. Recent OSes are smart enough in this regard.

Is there any point in making a file hosting service in a small office on Linux? (otherwise they got the viruses on shared disks).

If you rummage better in Linux than Windows - do it on it. But just because of viruses, changing the axis is such a thing. It is better to set up rights, antiviruses and grief not to know.

Answer 3

Take away administrator rights from users on workstations and activate Windows Defender everywhere, it will become easier with viruses.
Which Linux to choose - do not care, the one that you know better, but take the stable distro branch. It is possible without GUI.

Answer 4

For 1C v7 there is absolutely no difference. The main disk and network are faster (SSD, gigabit).
For file cleaning, yes, there are some improvements in 2016+.
Problems with viruses are solved by the domain and tightening the screws by politicians.

Answer 5

and if ram=8gb

for an android phone, the wave is enough.

Answer 6

Better synology in a small office, only synology in a big office.

Answer 7

Umm... first you need to learn the mat part and figure out what B-tree and R-tree are and how they appear in modern DBMS, figure out what "6th normal form" is (second year of university).
If this is a muscle that has "infinite length" plates - from 200GB to 1TB, then it's enough to simply use ENGINE ARCHIVE with an R-tree index. Otherwise (if less than 200GB) you need to (learn the mat part and set your brains) refactor the base. It's better to move from MySQL to PostgreSQL, but with MongoDB there are a lot of problems. Standard DBMS based on B-tree are not suitable for databases from 200GB+. MySQL is an exception due to the modularity of the storage system, meaning ENGINE ARCHIVE, but since there is no T-tree, a normal cache is needed. PostgreSQL will be even worse - you need to pick various extensions like cstore_fdw, etc.
uint64 ID in the form of a hash is a very controversial decision, even if we assume that there are no collisions in some universe, then definitely not in this one, and in addition you need to run the Bloom filter. Although, it's best to just use composite keys and not bother.
You can also try wrapping HBase in Apache Phoenix, everything is already there - both the Bloom filter and indexing, you can even arrange an X-tree. HBase, by the way, scales well for writing, while Cassandra, on the contrary, scales well for reading.
Sharding (partitioning) and replication should be done when the scheme is well normalized, and when it is more or less clear which plates need to be scaled for writing and which for reading - somewhere you need CA, somewhere CP, and somewhere AP ... (CAP theorem)
It's a lot of fun to write C-functions in PostgeSQL to aggregate into materialized views, especially fun with GPGPU.

Answer 8

You have a unique entity id, you don't work with records (only getting/inserting by id), what can be used here besides redis? Which quietly processes hundreds of thousands of requests per second?
try.redis.io - take a look and admire.
redis.io/clients - clients here, whatever you want.
redis.io/documentation - a lot and interesting how to set it up.

Answer 9

Try ArangoDB, I switched to it from MySQL. So far very satisfied.

Answer 10

I agree with Dmitry Avilov - if the selection is by a single key, then rel. DBMS is useless. Monga is not the only possible solution, experiment with other NoSQL databases.

Answer 11

Insert in a crowd and not one record at a time, and if one at a time, then turn off the indexes and then turn them on

Answer 12

Try tarantool. This is a very promising subd, similar to a radish, but with very broad capabilities. Of the minuses - not so popular and widely used.