OVH New SP-32 and SP-64 2017 with E3-1270v6 4c/8th 3.8/4.2GHz and vRack!

exception0x876

Basically repost of their tweet. vRack availability together with cheap NVMe SSDs are interesting.

SP-32 2017 price - €69.99/mo

SP-64 2017 price - €94.99/mo

Link - https://www.ovh.ie/dedicated_servers/enterprise/

Harambe

EG-16 & 32 are now also 1230v6 & 1270v6. With a 3x4TB and 2x4TB + 2x450GB NVMe on the latter.. interesting.

willie

Whole dedi line and public cloud seems to have been revamped though not earth-shakingly.

pbgben

We've already got NVMe in our BHS location. Soo fast ;P

Luke007

Price in the US site is much higher than that in the IE site.

Bopie

@Luke007 said:
Price in the US site is much higher than the IE site.

Don't forget the VAT for IE site, the prices shown are EX vat

Harambe

@pbgben said:
We've already got NVMe in our BHS location. Soo fast ;P

What, no promo? Hit us with a deal.

Luke007

Bopie said: Don't forget the VAT

VAT is applicable only for European Customers.

Falzo

@Luke007 said:

VAT is applicable only for European Customers.

that's probably more wrong then right. if you are dealing with a branch of OVH located in europe themselves VAT applies to all customers unless they are valid/proven business clients.

sin

@Falzo said:

@Luke007 said:

VAT is applicable only for European Customers.

that's probably more wrong then right. if you are dealing with a branch of OVH located in europe themselves VAT applies to all customers unless they are valid/proven business clients.

You can just open a ticket with OVH validation proving you live in the US (or wherever else) and they'll take off VAT and reimburse you for any amount that already charged you. Worked great on my SYS dedi and with Kimsufi once I entered my US billing address it automatically took VAT off.

I started buying from the North American branches anyways though because their NA support is WAY better.

Luke007

Falzo said: unless they are valid/proven business clients

Any non European customer can get VAT exemption by sending necessary validation documents.

Falzo

@Luke007 said:

Falzo said: unless they are valid/proven business clients

Any non European customer can get VAT exemption by sending necessary validation documents.

if that's the way OVH handles it I won't argue.. but after all probably not the right way according to actual tax laws around here. maybe they do some tricks of internal reassignment to non-eu branches or whatever to get around it.

gleert

pbgben said: We've already got NVMe in our BHS location. Soo fast ;P

Which model NVMe SSD did you get?

willie

I have a Kimsufi through their .ie office and am not getting charged VAT. I think they just went by my billing address which is not in EU.

pbgben

@gleert said:

pbgben said: We've already got NVMe in our BHS location. Soo fast ;P

Which model NVMe SSD did you get?

I have centos installed and can't get any cmd to tell me about them. Its the 450GB Option.

teamacc

you forgot the conv=fdatasync in that last cmd

pbgben

@teamacc said:
you forgot the conv=fdatasync in that last cmd

Ah, what does that bit do?

Should update your sig to say you work for http://lowendtaIk.com/ Has the same effect :P Capital i

teamacc

@pbgben said:

@teamacc said:
you forgot the conv=fdatasync in that last cmd

Ah, what does that bit do?

Should update your sig to say you work for http://lowendtaIk.com/ Has the same effect :P Capital i

This tells dd to require a complete “sync” once, right before it exits. So it commits the whole 256 MB of data, then tells the operating system: “OK, now ensure this is completely on disk”, only then measures the total time it took to do all that and calculates the benchmark result.

source: https://romanrm.net/dd-benchmark

Aka you're testing ram speed :P

pbgben

So your way comes back with a slower result :P

teamacc

@pbgben said:
So your way comes back with a slower result :P

True, but it's the actual disk write speed. Your way is filling the disk buffer in ram, without actually writing to disk. Kernel caches it and writes it to disk in its own time.

MikeA

@pbgben said:
So your way comes back with a slower result :P

v

@teamacc said:

@pbgben said:

@teamacc said:
you forgot the conv=fdatasync in that last cmd

Ah, what does that bit do?

Should update your sig to say you work for http://lowendtaIk.com/ Has the same effect :P Capital i

This tells dd to require a complete “sync” once, right before it exits. So it commits the whole 256 MB of data, then tells the operating system: “OK, now ensure this is completely on disk”, only then measures the total time it took to do all that and calculates the benchmark result.

source: https://romanrm.net/dd-benchmark

Aka you're testing ram speed :P

Normal SSD gets 1.5 GB/s with the method in his screenshot, fdatasync gets 240 MB/s, so the screenshot he gave from the bench likely is using fdatasync which is showing around 450 MB/s on the NVMe. Plus it's /dev/zero.

pbgben

@MikeA said:

@pbgben said:
So your way comes back with a slower result :P

v

@teamacc said:

@pbgben said:

@teamacc said:
you forgot the conv=fdatasync in that last cmd

Ah, what does that bit do?

Should update your sig to say you work for http://lowendtaIk.com/ Has the same effect :P Capital i

This tells dd to require a complete “sync” once, right before it exits. So it commits the whole 256 MB of data, then tells the operating system: “OK, now ensure this is completely on disk”, only then measures the total time it took to do all that and calculates the benchmark result.

source: https://romanrm.net/dd-benchmark

Aka you're testing ram speed :P

Normal SSD gets 1.5 GB/s with the method in his screenshot, fdatasync gets 240 MB/s, so the screenshot he gave from the bench likely is using fdatasync which is showing around 450 MB/s on the NVMe. Plus it's /dev/zero.

Marketing is all lies! I want my 1GB/s SSD

teamacc

@pbgben said:

@MikeA said:

@pbgben said:
So your way comes back with a slower result :P

v

@teamacc said:

@pbgben said:

@teamacc said:
you forgot the conv=fdatasync in that last cmd

Ah, what does that bit do?

Should update your sig to say you work for http://lowendtaIk.com/ Has the same effect :P Capital i

This tells dd to require a complete “sync” once, right before it exits. So it commits the whole 256 MB of data, then tells the operating system: “OK, now ensure this is completely on disk”, only then measures the total time it took to do all that and calculates the benchmark result.

source: https://romanrm.net/dd-benchmark

Aka you're testing ram speed :P

Normal SSD gets 1.5 GB/s with the method in his screenshot, fdatasync gets 240 MB/s, so the screenshot he gave from the bench likely is using fdatasync which is showing around 450 MB/s on the NVMe. Plus it's /dev/zero.

Marketing is all lies! I want my 1GB/s SSD

Read https://romanrm.net/dd-benchmark

teamacc

Also, testing a 1+gbyte/s ssd with 84mbyte of test data? Meh.

pbgben

@teamacc said:

@pbgben said:

@MikeA said:

@pbgben said:
So your way comes back with a slower result :P

Read https://romanrm.net/dd-benchmark

Yeh, I have. It comes back slower.

rootbeer@hybrid:~# dd bs=1M count=256 if=/dev/zero of=test conv=fdatasync
256+0 records in
256+0 records out
268435456 bytes (268 MB) copied, 0.525429 s, 511 MB/s
rootbeer@hybrid:~# dd bs=1M count=1024 if=/dev/zero of=test conv=fdatasync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 1.86997 s, 574 MB/s
rootbeer@hybrid:~# dd if=/dev/zero bs=1M count=1024 | md5sum
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 1.86613 s, 575 MB/s
cd573cfaace07e7949bc0c46028904ff  -
rootbeer@hybrid:~#

teamacc

raid0? What do they do individually?

pbgben

@teamacc said:
raid0? What do they do individually?

Thats a R1 (I have yet to trust customer data with a R0)

Also, This server was setup last week, so not 100% sure its what was mentioned in the OP

teamacc

What does sudo hdparm -Tt /dev/sda return? (might have to adjust to /dev/md0/1)

teamacc

Also, those 560ish speed should be expected, intel 600p does about that: http://ark.intel.com/products/94924/Intel-SSD-600p-Series-512GB-M_2-80mm-PCIe-3_0-x4-3D1-TLC

pbgben

@teamacc said:
What does sudo hdparm -Tt /dev/sda return? (might have to adjust to /dev/md0/1)

Neither of those paths are valid

[rootbeer@ns511813 ~]# hdparm -Tt /dev/md3

/dev/md3:
 Timing cached reads:   36118 MB in  2.00 seconds = 18095.42 MB/sec
 Timing buffered disk reads: 1620 MB in  3.00 seconds = 539.81 MB/sec
[rootbeer@ns511813 ~]# hdparm -Tt /dev/md2

/dev/md2:
 Timing cached reads:   35842 MB in  2.00 seconds = 17957.25 MB/sec
 Timing buffered disk reads: 254 MB in  0.87 seconds = 292.93 MB/sec

[rootbeer@ns511813 ~]# hdparm -Tt /dev/nvme0n1

/dev/nvme0n1:
 Timing cached reads:   37122 MB in  2.00 seconds = 18599.47 MB/sec
 Timing buffered disk reads: 2032 MB in  3.00 seconds = 677.07 MB/sec

EDIT: 37GB in 2 Seconds, 18.6GB/s Read seems suspect

teamacc

That's again cached reads, so from ram. Real values are the buffered ones. They seem a tad low though, compared to intel specs.

Then again, linux drivers.