Do you wonder if your IT business is truly profitable or if you should raise your prices? Learn how to calculate your overhead burden using our free interactive tool and use it to determine the right price for your IT services. Start calculating Now!
ZFS: Nexentastor best practises for ESXi 5
Hi,
I just configured my HP Microserver 16GB NL 40 with Nexenta. There are 6 SSD's inside: 2 for OS (mirrored, 60 GB), 4 SSD's 120GB (Kingston, enterprise version), 3 SATA 250 GB 5400 RPM.
Before setting this up, I actually only was looking for a fast storage NAS in my lab. I ended up spending some weeks but actually I don't want to spend any more now. I want to focus again on Microsoft products which run on my vm's.
Therefore this question.
I saw loads of articles for best practises: using NFS3 instead of 4 for ESXI?, make cache/log with mirrored drives?, backup the storage using snapshots?, use ISCSI over NFS?
Now, as mentioned, I don't want to spend much more time on configuring, checking best performance, just have a best practice "jump start".
So please advise (configured ZRAID on 4 SSD's of 120GB, NFS; 3 SATA's in ZRAID too).
Note: I'll probably replace the SATA's with 3 or more barracuda 2TB 64 MB cache, green disks for having more storage.
J.
I just configured my HP Microserver 16GB NL 40 with Nexenta. There are 6 SSD's inside: 2 for OS (mirrored, 60 GB), 4 SSD's 120GB (Kingston, enterprise version), 3 SATA 250 GB 5400 RPM.
Before setting this up, I actually only was looking for a fast storage NAS in my lab. I ended up spending some weeks but actually I don't want to spend any more now. I want to focus again on Microsoft products which run on my vm's.
Therefore this question.
I saw loads of articles for best practises: using NFS3 instead of 4 for ESXI?, make cache/log with mirrored drives?, backup the storage using snapshots?, use ISCSI over NFS?
Now, as mentioned, I don't want to spend much more time on configuring, checking best performance, just have a best practice "jump start".
So please advise (configured ZRAID on 4 SSD's of 120GB, NFS; 3 SATA's in ZRAID too).
Note: I'll probably replace the SATA's with 3 or more barracuda 2TB 64 MB cache, green disks for having more storage.
J.
Do more with
EXPERT OFFICE® is a registered trademark of EXPERTS EXCHANGE®
Best practice is to get rid of ESXi entirely with Nexenta. You want to run native Nexenta on the physical machine.
As for using 'cuda green drives those disks are low duty cycle, and not designed for 24x7x365. If you DO use them, don't even think of doing anything other than RAIDZ2.
As for using 'cuda green drives those disks are low duty cycle, and not designed for 24x7x365. If you DO use them, don't even think of doing anything other than RAIDZ2.
you may find better performance using iSCSI LUNs, if your NFS performance is not as expected.
you will need to test performance in the VMs.
Benchmark software listed as before.
your current config looks okay. experiment with adding a ZIL log SSD to your SATA three, for increased performance, also SSD for cache
you will need to test performance in the VMs.
Benchmark software listed as before.
your current config looks okay. experiment with adding a ZIL log SSD to your SATA three, for increased performance, also SSD for cache
You are wasting the SSDs on O/S disk.
Use one SSD for cache. Get two cheap, cheap desktop or even laptop SATA disks and use them mirrored for root pool. As for the rest, it really depends on what you are trying to do, but I would generally make it easy and go with single RAIDZ2 pool with 4 high capacity SATA disks, 3 SSDs, (mirrored pair for ZIL), 3rd for cache.
unmirrored ZIL is foolish and you risk major data loss if the ZIL disk fails. Don't do it. Losing the cache SSD does not risk data.
Buy less SSD and more RAM, as system memory is used for read cache and Nexenta/Solaris will use every bit of it.
Use one SSD for cache. Get two cheap, cheap desktop or even laptop SATA disks and use them mirrored for root pool. As for the rest, it really depends on what you are trying to do, but I would generally make it easy and go with single RAIDZ2 pool with 4 high capacity SATA disks, 3 SSDs, (mirrored pair for ZIL), 3rd for cache.
unmirrored ZIL is foolish and you risk major data loss if the ZIL disk fails. Don't do it. Losing the cache SSD does not risk data.
Buy less SSD and more RAM, as system memory is used for read cache and Nexenta/Solaris will use every bit of it.
Thanks for the input dlethe!
I do want to make it easy, general.
However, I do have the ssd s installed and don t like to start all over installing Nexenta afain on other disk (might do that some day but for now I want to play with my new born nas :-)).
I don t want more then 1 disk for redundancy (it s a lab after all, if 1disk down, I could shutdown whole nas and go to shop for other disk) UNLESS performance proves drastically.
That would mean raidz, not raidz2 I guess.
So you wouldn t make a raidz of ssd s then and use a mirror for zil and 1 disk for cache. Then I have 1 spare ssd 120 gb. Could use that in my laptop.
Note: microserver has already 16 gb ram which is max it can hold.
I do want to make it easy, general.
However, I do have the ssd s installed and don t like to start all over installing Nexenta afain on other disk (might do that some day but for now I want to play with my new born nas :-)).
I don t want more then 1 disk for redundancy (it s a lab after all, if 1disk down, I could shutdown whole nas and go to shop for other disk) UNLESS performance proves drastically.
That would mean raidz, not raidz2 I guess.
So you wouldn t make a raidz of ssd s then and use a mirror for zil and 1 disk for cache. Then I have 1 spare ssd 120 gb. Could use that in my laptop.
Note: microserver has already 16 gb ram which is max it can hold.
also note, that a single low power, low performing, AMD Neo N40L 1,5GHz dual core processor, is the bottleneck! also onboard nic does not support jumbo frames
my new Nexus phone performs faster than a ProLiant Microserver!
before you waste anymore time benchmark and test existing NAS versus Microserver, you may find your existing NAS has a faster processor.
my new Nexus phone performs faster than a ProLiant Microserver!
before you waste anymore time benchmark and test existing NAS versus Microserver, you may find your existing NAS has a faster processor.
Hanccocka, you always mentioned it s the fastest nas you ll find around? That s why I started with it in the first place ...
I have stated ZFS is the fastest NAS technology around, now being employed by other NAS vendors eg Thecus etc, even better when using SSD for ZIL and Cache.
without any published benchmarks with your current NAS, its difficult to compare, if you look back through my many posts you will see many references to the Microservers performance, you cannot expect the performance of the Microserver to be superior to a tower server because of its low spec.
if you also check forums, google, ProLiant Microservers performance bottleneck is limited by its design. again its a compromise of price, size and power versus performance.
and for its price, power, performance it packs a punch.
BUT, We do not have ANY tested performance metrics for your current setup, not that Ive seen.
We use these in OUR Research and Development, Training and Demo Labs for ALL VMware Products, performance is acceptable for us, we have no issues with ZFS, when using ESXi with NFS and iSCSI.
These SAN/ZFS based MicroServers replaced 1GB Fibre Channel SAN, using 1GB FC, and 10k and 15k rpm U320 SCSI disks. NFS and iSCSI performance was tested to be far superior to existing shared storage. Considering the cost of electricity, and performance, size of current storage in (old FC SAN), it was a no brainer to switch. (based on metrics ancd testing).
without any published benchmarks with your current NAS, its difficult to compare, if you look back through my many posts you will see many references to the Microservers performance, you cannot expect the performance of the Microserver to be superior to a tower server because of its low spec.
if you also check forums, google, ProLiant Microservers performance bottleneck is limited by its design. again its a compromise of price, size and power versus performance.
and for its price, power, performance it packs a punch.
BUT, We do not have ANY tested performance metrics for your current setup, not that Ive seen.
We use these in OUR Research and Development, Training and Demo Labs for ALL VMware Products, performance is acceptable for us, we have no issues with ZFS, when using ESXi with NFS and iSCSI.
These SAN/ZFS based MicroServers replaced 1GB Fibre Channel SAN, using 1GB FC, and 10k and 15k rpm U320 SCSI disks. NFS and iSCSI performance was tested to be far superior to existing shared storage. Considering the cost of electricity, and performance, size of current storage in (old FC SAN), it was a no brainer to switch. (based on metrics ancd testing).
What is your I/O model? Be specific, ratio of reads vs writes vs sequential vs random I/O? Will you use deduplication or compression? Are these small data files?
If you are going all SSD, then just stripe 2 mirrors together. No need for ZIL or cache if you are all SSD, you need them if you go RAIDZ1 and those cheap slow SATA disks.
If you are going all SSD, then just stripe 2 mirrors together. No need for ZIL or cache if you are all SSD, you need them if you go RAIDZ1 and those cheap slow SATA disks.
You can do both, it just depends on the I/O load. At some point CPU utilization may suffer, but if I/O utilization hits a ceiling first, then CPU load wont be a significant bottleneck.
I'll make a new ticket asap but nappit doesn't install ... just shows this:
admin@Nexenta:~$ wget -O - www.napp-it.org/nappit | perl
--14:33:49-- http://www.napp-it.org/nappit
=> `-'
Resolving www.napp-it.org... 80.67.28.3, 2a00:1158:0:300:9a63::1
Connecting to www.napp-it.org|80.67.28.3|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 23,126 (23K) [text/plain]
100%[=====================
14:33:49 (235.53 KB/s) - `-' saved [23126/23126]
Logging on with root gives "wrong command".
admin@Nexenta:~$ wget -O - www.napp-it.org/nappit | perl
--14:33:49-- http://www.napp-it.org/nappit
=> `-'
Resolving www.napp-it.org... 80.67.28.3, 2a00:1158:0:300:9a63::1
Connecting to www.napp-it.org|80.67.28.3|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 23,126 (23K) [text/plain]
100%[=====================
14:33:49 (235.53 KB/s) - `-' saved [23126/23126]
Logging on with root gives "wrong command".
Premium Content
You need an Expert Office subscription to comment.Start Free Trial