anyone using samsung evo 850 in servers ?

cociu

Hello there ,
can anyone paste here a I/Q speed who use samsung evo 850 1 tb ?

Thanks in advance,

miTgiB

cociu said: I/Q speed

[root@e3clt08 ~]# dd if=/dev/zero of=/vz/test.1g bs=1G count=1 oflag=dsync
1+0 records in
1+0 records out
1073741824 bytes (1.1 GB) copied, 1.91212 s, 562 MB/s
[root@e3clt08 ~]# dd if=/dev/zero of=/vz/test.1g bs=1G count=1 oflag=dsync
1+0 records in
1+0 records out
1073741824 bytes (1.1 GB) copied, 1.77893 s, 604 MB/s
[root@e3clt08 ~]# dd if=/dev/zero of=/vz/test.1g bs=1G count=1 oflag=dsync
1+0 records in
1+0 records out
1073741824 bytes (1.1 GB) copied, 1.7738 s, 605 MB/s
[root@e3clt08 ~]# 

mdadm raid1 for a pair of 1tb 850 EVO

NexHost

@miTgiB said:
mdadm raid1 for a pair of 1tb 850 EVO

iops?

rds100

miTgiB said: 605 MB/s

So somehow it beats the theoretical maximum on the SATA3 interface, which is 600MB/s?

black

rds100 said: So somehow it beats the theoretical maximum on the SATA3 interface, which is 600MB/s?

SATA 3 is 6 Gbps, which is 750 MB/sec, right?

dragon2611

root@pve:~# dd if=/dev/zero of=/mnt/850_evo/test.1g bs=1G count=1 oflag=dsync
1+0 records in
1+0 records out
1073741824 bytes (1.1 GB) copied, 3.72171 s, 289 MB/s

Single 512GB 850 Evo, probably plugged into a Sata2 port instead of Sata3. (It's a C2750-D4I board)

rds100

black said: SATA 3 is 6 Gbps, which is 750 MB/sec.

It is clocked at 6Gbps, but then the data is transferred with 8b/10b encoding (so for transferring 8 bits of real data, 10 bits are clocked on the wire). This results in real theoretical throughput of 4.8Gbps, or 600MB/s.

Riz

[root@xen7 ~]# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 2.90005 seconds, 370 MB/s
[root@xen7 ~]# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 3.79181 seconds, 283 MB/s
[root@xen7 ~]# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 3.3518 seconds, 320 MB/s


SATA3, 256G Evo

black

rds100 said: It is clocked at 6Gbps, but then the data is transferred with 8b/10b encoding (so for transferring 8 bits of real data, 10 bits are clocked on the wire). This results in real theoretical throughput of 4.8Gbps, or 600MB/s.

Learning something new everyday.

cociu

dragon2611 said: root@pve:~# dd if=/dev/zero of=/mnt/850_evo/test.1g bs=1G count=1 oflag=dsync 1+0 records in 1+0 records out 1073741824 bytes (1.1 GB) copied, 3.72171 s, 289 MB/s

Single 512GB 850 Evo, probably plugged into a Sata2 port instead of Sata3. (It's a C2750-D4I board)

i thinck i have the same issue here

black

sata 3.1 500 GB Evo ( on my laptop )

dd if=/dev/zero of=disktest bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 2.45795 s, 437 MB/s

dd if=/dev/zero of=disktest bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 2.5249 s, 425 MB/s

dd if=/dev/zero of=disktest bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 2.42858 s, 442 MB/s

FVPS

Not going to say much but probably Tim is using SATA 3.2 (released in 2013)?

https://www.sata-io.org/sites/default/files/documents/SATA_v3 2_PR__Final_BusinessWire_8.20.13.pdf

I don't know how recent his servers are. But the chance might be there that he has SATA 3.2. There are some SATA 3.2 mainboards from ASUS and other companies. No sure how it's about servers though.

miTgiB

FVPS said: I don't know how recent his servers are. But the chance might be there that he has SATA 3.2. There are some SATA 3.2 mainboards from ASUS and other companies. No sure how it's about servers though

You are all gonna call me a cheater, so I'll fess up before anyone gets the chance, the system I made my post from is a Supermicro x9 e3 class board X9SCD for the microcloud supermicro offered many moons ago, then it is running Proxmox 4.1 and I tested from a KVM system on it running a SolusVM OpenVZ slave. Same system many in Charlotte have their openvz 128/256mb VPS on, it is a live, in-production system. The disk caching I use is writeback on a qcow2 disk.

The KVM systems that have been converted to SSD and are native CentOS 7 with mdadm raid10 using the 2 SATA-III ports on Supermicro boards, with a PCIe 2.0 x2 SATA-III 4 port card for 2 additional ports, here is a look from within a system running on that. Again, a full production node. This is the VM hosting hostigation.com, it is not as impressive, but I spoke with @cocui with some suggestions how he can tweak his setup.

root@command [~]# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 4.35285 s, 247 MB/s
root@command [~]# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 4.15506 s, 258 MB/s
root@command [~]# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 4.55854 s, 236 MB/s
root@command [~]#

dragon2611

I wasn't overly worried about Sata2/Sata3 it's considerably faster than the Spinning Rust it replaced So I just plugged it into whatever port that particular cable went to.

adendrag

personally would never use EVO's - always go pro

J1021

1600+0 records in
1600+0 records out
104857600 bytes (105 MB) copied, 0.577748 s, 181 MB/s

Sometimes it tops 200MB/s. Just a single EVO connected running Debian, no virtualisation or tricks.