Storage VPS in Amsterdam - 4TB - 10Gbit port.

lowprofile

Hi,

looking for a VPS with minimum 4TB storage in Amsterdam - (must be able to write and read min. 600-700MB/s) - Port must be 10Gbit and speed must be close to 10Gbit

4 vCPU
8GB ram
10Gbit port
30TB bandwidth/monthly

Usecase: Backup

What's the best offer out there?

TrK

Budget?

allthemtings

@NovaCloudHosting comes to mind but they offer 2.5Gbps afaik

AndroidDeals

I think it will be a big cost

luckypenguin

My suggestion is, for that kind of IOPS, r/w, get a dedi. Your specs are minimal.
With a VPS, you will always be limited to what your neighbors are doing.
The provider can allocate you more "priority", but what fkn backup is it to saturate 10g?

rpqu

@host_c , any backup only deals ?

luckypenguin

@rpqu said:
@host_c , any backup only deals ?

I think you misread "Amsterdam" for "Bucharest".

rpqu

@luckypenguin said:

@rpqu said:
@host_c , any backup only deals ?

I think you misread "Amsterdam" for "Bucharest".

Oops

lowprofile

Budget 50eur

The reason for 10gbit is when to recover is must recover/transfer fast. Will be using to do incremental backups, but if the shit hits the fan, I need FAST recovery (each backup file is approx 300G)

jsg

@lowprofile said:
Budget 50eur

The reason for 10gbit is when to recover is must recover/transfer fast. Will be using to do incremental backups, but if the shit hits the fan, I need FAST recovery (each backup file is approx 300G)

With that kind of budget I think you should be able to get some good offers. This is how I'd approach it:

• @Hybula - top connectivity, good HW performance - but at a price.
• @AirnodeHosting - a very attractive provider who actually offers good pricing. I don't know about 4TB storage, but I'd definitely contact and ask them.
• @HostDZire - known good quality (themselves or Leaseweb resold) at really attractive prices - but the same 4TB storage caveat as above, but I'd definitely contact and ask them.

Independent of that I strongly suggest to also have a good look at @host_c who IMO is THE european storage/backup provider. In Romania though (which IMO in 95+% of cases isn't a problem). Not super-cheap anymore, sadly, but you get top quality, reliability, and support (and I think still within your budget).

host_c

@rpqu said:
@host_c , any backup only deals ?

Like rclone or ftp only ?

If yes, cost wise it would do nothing, I prefer to hand out a full vps and let the customer decide what he uses.

Or you are referring to some thing else ?

@jsg, thx for the mention

@lowprofile , why the 10g link on a 4tb storage ? If you do backup , that would imply a compression of some sort, I highly doubt you can push more then 1-2gbps from source read + compression , also, a backup solution will not write at 10g, maybe if raid 10 of NVME…..

lowprofile

I have 3 servers - each of them having a proper 10Gbit interface (dedicated) - They offload mission critical backups incrementals... if the shit hits the fan, i would need to recover as fast as possible - We are talking about 300GB per restore file. The bottleneck in my test has been mainly the 1GB interface.

5Gbit would also work i assume.

itzsenu

@lowprofile said: 5Gbit would also work i assume.

The important part is getting 600+ of sustained storage throughput, not just 10G port.

host_c

@itzsenu said: The important part is getting 600+ of sustained storage throughput, not just 10G port.

@lowprofile

600 MB/sec sustained write, even if sequential, is NVME territory ( SSD tops out at ~550MB/SEC, also I am not talking any type of cache into consideration here, just the pure raw performance of the underlying storage type)

125mb/sec-256mb/sec ( 1-2 GBPS ) on HDD is doable in raid 6/60, above that is also doable on HDD, but in Raid 10 Setups with large cache, yet that will cost you premium.

host_c

-

farsighter

Wait for @host_c BF yearly deals $7/tb 🚀

itzsenu

@host_c said: 600 MB/sec sustained write, even if sequential, is NVME territory ( SSD tops out at ~550MB/SEC, also I am not talking any type of cache into consideration here, just the pure raw performance of the underlying storage type)

125mb/sec-256mb/sec ( 1-2 GBPS ) on HDD is doable in raid 6/60, above that is also doable on HDD, but in Raid 10 Setups with large cache, yet that will cost you premium.

@lowprofile

real bottleneck for 300gb restores is usually small file/metadata overhead not sequential throughput. hdd raid 6/60 hits 600mb/s+ sequential read fine with enough spindles, but seek latency on non sequential data (fragmented backups, lots of small files) tanks it way below theoretical regardless of link speed. single restore capped by storage at 1-2gbps

A massive water pipe cannot fix a sputtering water pump

EDIT: go with NVMe + 10Gbps port

host_c

@itzsenu said:

@host_c said: 600 MB/sec sustained write, even if sequential, is NVME territory ( SSD tops out at ~550MB/SEC, also I am not talking any type of cache into consideration here, just the pure raw performance of the underlying storage type)

125mb/sec-256mb/sec ( 1-2 GBPS ) on HDD is doable in raid 6/60, above that is also doable on HDD, but in Raid 10 Setups with large cache, yet that will cost you premium.

@lowprofile

real bottleneck for 300gb restores is usually small file/metadata overhead not sequential throughput. hdd raid 6/60 hits 600mb/s+ sequential read fine with enough spindles, but seek latency on non sequential data (fragmented backups, lots of small files) tanks it way below theoretical regardless of link speed. single restore capped by storage at 1-2gbps

A massive water pipe cannot fix a sputtering water pump

EDIT: go with NVMe + 10Gbps port

host_c

@farsighter saidhost_c BF yearly deals, $7/tb/Month🚀- There, fixed the typo

Yet.........

HOST-C

rpqu

@host_c said:

@itzsenu said: The important part is getting 600+ of sustained storage throughput, not just 10G port.

@lowprofile

600 MB/sec sustained write, even if sequential, is NVME territory ( SSD tops out at ~550MB/SEC, also I am not talking any type of cache into consideration here, just the pure raw performance of the underlying storage type)

125mb/sec-256mb/sec ( 1-2 GBPS ) on HDD is doable in raid 6/60, above that is also doable on HDD, but in Raid 10 Setups with large cache, yet that will cost you premium.

might as well get warm node + HA @oloke AMS s3 when

lowprofile

@host_c said:

@itzsenu said: The important part is getting 600+ of sustained storage throughput, not just 10G port.

@lowprofile

600 MB/sec sustained write, even if sequential, is NVME territory ( SSD tops out at ~550MB/SEC, also I am not talking any type of cache into consideration here, just the pure raw performance of the underlying storage type)

125mb/sec-256mb/sec ( 1-2 GBPS ) on HDD is doable in raid 6/60, above that is also doable on HDD, but in Raid 10 Setups with large cache, yet that will cost you premium.

I can accept 500MB/s as well - whats the best you can offer?
Reason for i stated Amsterdam as location was due to best possible utilization of the 10gbit interface

jsg

@lowprofile said:
I have 3 servers - each of them having a proper 10Gbit interface (dedicated) - They offload mission critical backups incrementals... if the shit hits the fan, i would need to recover as fast as possible - We are talking about 300GB per restore file. The bottleneck in my test has been mainly the 1GB interface.

5Gbit would also work i assume.

.

@lowprofile said:
I can accept 500MB/s as well - whats the best you can offer?
Reason for i stated Amsterdam as location was due to best possible utilization of the 10gbit interface

You've built yourself a mind trap it seems. "If SHTF ultrafast recovery is absolute priority". Don't get me wrong, in theory that's a good approach - in real life though that's either very expensive (and still not easy to find) or a balancing act of compromises; in fact you yourself introduced the first barrier ("Budget €50", per month presumably).

For a start, optimal 10 Gb/s pipe utilization almost certainly won't be your problem/bottle neck, nor will location, at least not within 50 ms. Besides, professionals WANT at least a couple of hundred kilometers distance.

This is how I see it and the relevant main factors:

• your budget (vs. your approach and real-world)
• reliability (a backup that isn't reliably available is next to worthless)
• total time of recovery (good enough pipe and fast enough storage, mainly reading)
• real importance of data.
• organization of backup (not all data are equally speed sensitive)

Which in terms of technology translates to (as @host_c has explained): the first, and highly likely most influential, factor is realistic disk speed in real life usage and situation. One and the same storage device will show quite different results on a dedi vs on a VDS vs on a VPS (and so are prices).
That kind of problem complex exists pretty much since storage exists, plus some others like e.g. medium lifetime, failure rate, etc.

Actually IMO even more important though is an entirely different point: what to back up, and in what kind of situation to restore? Restoring some hobby VPS is very different from restoring the database core of an enterprise which is losing tens or hundreds of thousands each and every minute. Besides: can you really be sure that the backed up server is fully and reliably available/operational after whatever "disaster" hit it?

So I think you should think well about that whole thing, check your premises and design the right (and realistic and feasible) backup strategy for you!

For example requesting "x Gb/s" is meaningless if your budget can't cover the "perfect theory". Also look closely at your premises and assumptions, what e.g. does "SHTF" mean? Because in quite a few SHTF scenarios you may not even be able to "restore quickly".

But OK, looking at what you probably although vaguely have in mind I suggest to simply buy two(!) 500 GB NVMes, put them into a mirror and solve the "as quickly as possible" problem by having a LOCAL backup, done. Next, (a) for "religious" reasons ("sound principle"), (b) in case the "SHTF" is very different from what you expect and more grave, and (c) peace of mind, add AT LEAST one outside backup, preferably a reliable one like @host_c and don't worry that much about "quick" because if the actual "SHTF" kills or makes unavailable your local backup, "quick" isn't decisive, "works and is available" is.

In other words: I suggest you split what actually is TWO problems into two approaches, namely "quick" (it's hard to beat "local") and "available, even if here an earthquake strikes". And having two backups is the reasonable minimum anyway, Amen.

concept

hmm ovh does have SYS-1 in France with 2x SSD NVMe 1.92TB Datacenter Class Soft RAID
But it is 70 eur/m + One time setup fee and limited to 1G

bobsburgers

Go get a server from hostingbydesign or hostdzire with 4x drives to get 600mbs speeds

zorker

Why not use Backblaze B2/S3 in AMS?

lowprofile

@zorker said:
Why not use Backblaze B2/S3 in AMS?

What is your experience with the speed?