Overdue Servers/Network Build Log

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

kidchunks

Member
Sep 28, 2014
39
18
8
35
Latest Pictures!

New edgerouter 1U Bracket


ATP USBe fits like a glove!



Wiring === fun


Keeping that fiber card nice and cool..


Rack completed..for now. Ignore the workstation and amp :).



Also, considering another ESXi host with the following:

Mobo: Supermicro X10DRL-i
CPU: 1xE5-2630v3
RAM: 4x16GB DDR4 ECC Samsung
SSDs: Still searching..
Chassis+PSU: Still searching..
 
Last edited:

neo

Well-Known Member
Mar 18, 2015
672
363
63
Gotta get me one of these !

Gonna check what shipping is to denmark :)

Edit: yikes.. $45 to ship.. guess i'll have to go play around with some bending metal myself.
It shouldn't cost more then $25 or so if they use USPS

If you really want it, I can probably reship it to you.
 
  • Like
Reactions: T_Minus

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,640
2,058
113
Very high end USB sticks :) May I ask where you got them? I couldn't find any info other than Manufacture and digiKey. SLC USB? MMMhmm :D
 

kidchunks

Member
Sep 28, 2014
39
18
8
35
Very high end USB sticks :) May I ask where you got them? I couldn't find any info other than Manufacture and digiKey. SLC USB? MMMhmm :D
I used the manufacturers flyer as a guide. I went with the SLC model, look at that TBW! :)
http://www.atpinc.com/downloadlog/16ab9ce2fcfcee0b

Highest capacities are the 16GB and 8GB. Digikey was the best place I could find the 8GB model.
AF8GSSGH-OEM ATP Electronics, Inc. | AF8GSSGH-OEM-ND | DigiKey

They also sell the 16GB...but you need to purchase........1,000 :(
AF16GSSGH-OEM ATP Electronics, Inc. | AF16GSSGH-OEM-ND | DigiKey


I'm happy with the speed and performance. Definitely recommend it as a boot drive if you don't need much space and want reliability.
 
  • Like
Reactions: T_Minus

maze

Active Member
Apr 27, 2013
576
100
43
It shouldn't cost more then $25 or so if they use USPS

If you really want it, I can probably reship it to you.
Thanks neo, i'll check up on it when im back from vacation! - Apreciate it :)
 

kidchunks

Member
Sep 28, 2014
39
18
8
35
Finally had some time to update this thread. Only managed to snag these while parts being delivered and building the NAS.

No introduction needed :)


Favorite Fans (noctua industrial 120mm)


PM863s with 3M thin sata cables (these are fragile!)


Semi-tight fit but enough air-flow


In the meantime, here is some benchmarks for my new VM storage (4xPM863 480GB in raidz on a flashed dell h200)


Test was done with the VM attached to the storage ( NFS, 10gbe and MTU 9000).

zpool dataset Settings:
compression=off
sync=disabled
primarycache=metadata

Arc cached was not used

-----------------------------------------------------------------------
CrystalDiskMark 5.1.2 x64 (C) 2007-2016 hiyohiyo
Crystal Dew World : Crystal Dew World
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

Sequential Read (Q= 32,T= 1) : 1096.495 MB/s
Sequential Write (Q= 32,T= 1) : 362.453 MB/s
Random Read 4KiB (Q= 32,T= 1) : 47.691 MB/s [ 11643.3 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 11.788 MB/s [ 2877.9 IOPS]
Sequential Read (T= 1) : 346.657 MB/s
Sequential Write (T= 1) : 259.825 MB/s
Random Read 4KiB (Q= 1,T= 1) : 5.866 MB/s [ 1432.1 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 3.041 MB/s [ 742.4 IOPS]

Test : 4096 MiB [C: 45.8% (14.6/31.9 GiB)] (x5) [Interval=5 sec]
Date : 2016/05/20 11:35:45
OS : Windows 7 Ultimate SP1 [6.1 Build 7601] (x64)
4K looks a little slow..curious if anyone with these drives in a similar configuration experienced something similar?

Update: Just tested locally (using bonnie and nmon for monitoring iops)..Something else is going on..either the network or the VM...4k speeds should not be that low.


Update #2:

Originally when I did my first round of testing I had too many variables in place. MTU changes, different blocksizes, only testing VM's disk. So, I honed my test and after a few days of testing (wont show it all)...decided on using ISCSI and 64k blocksize for the zvol datastore.

NAS Specs/Config:
OS: Debian Jessie
iSCSI Target: iSCSI Enterprise Target
RAM: 8GB
CPU: Xeon-D 1518
Disks: 4xPM863
Network: 10GBe (MTU 1500)

Pool/Dataset Config:
RAIDZ
sync=always
compression=off
primarycache=all

Testing pure disk speeds (Reduced max arc to 2GB to avoid hitting it)
I used arcstat.sh to make sure I wasn't hitting the arc.

IOMETER Config:
Test unformatted iscsi disk
4GB disk size
2 Workers, 16 Outstanding I/Os (per worker) 2 mins per test

Test Patterns
  1. VDI - 4K 20% Seq/80% Random (80% Write / 20% Read)
  2. SQL Server 16K - 100% Random (34% Write/ 66% Read)
  3. SQL Server 64K - 100% Random (34% Write/ 66% Read)
  4. Other patterns were included in with IOMETER.

Results:

LAT(ms) -- Avg IOPS -- Avg MBps

VDI
3.24 -- 4934 -- 19
SQL Server 16K
2.78 -- 5746 -- 89
SQL Server 64K
3.17 -- 5050 -- 315
64K Seq Read
1.29 -- 12404 -- 775
64K Seq Write
2.55 -- 6263 -- 391
256K Seq Read
4.78 -- 3344 -- 836
256K Seq Write
8.53 -- 1876 -- 469


Takeaways:
  1. I didn't see much performance gains with striped mirrors (2 spread vs 3 spread raidz). That's a given I would think as one would need a few more striped vdevs to really see the power of striped mirrors. If you have only 4 SSDs..stick to raidz.
  2. zvol blocksize makes a difference. I tested 8k, 16k, 64k, 128k and found 64k to be the right balance for my storage needs (16k close second).
  3. NFS fell short when sync=always was set where as iSCSI performed very well. Stick to iSCSI(with sync=always) if you can.
  4. Need more ram but will wait till ARC miss ratio hits < 85%

Look forward to hearing others experiences and feedback.
 
Last edited: