Exablox testing journal
-
@MattSpeller said:
To speed up testing I've installed a gigabit switch, this absolutely will change their performance.
Welcome to 2003.
-
-
Having some further errors on the oneblox units, apparently they're not talking to each other right now. While I find this slightly irritating I want to point out the following email chain, check out the times.
Matthew Speller
Feb 13, 11:27
Hello Support,
Could I please get a hand resolving this?Bob Gardiner (Exablox Support)
Feb 13, 11:30
I will take a look and let you know what I find.
Bob -
Hey @MattSpeller here's some interesting reading:
http://www.theregister.co.uk/2015/02/13/exablox_not_good_for_restore/
Ā
Have you tested the restore times?
Did you find them adequate? -
@nadnerB Not yet, though today may be the day.
-
Far from an exhaustive test, but I managed to toss around 100GB to get some BASIC performance numbers. Data used is my ISO cache, so really big files and not very many of them. Test setup was on gig lan, external ESATA 7200rpm drive connected to a Latitude 6410. Virtually no other traffic exists on this switch. I also double checked the drives in the exablox; Seagate 4TB 5900rpm consumer grade junk.
Results:
33MB/s up
20MB/s downObviously this is very primitive testing, the article linked by Brendan (and the test linked inside it) is much more exhaustive and includes queue depth >1. While these numbers are unimpressive, I feel it's important to point out they make zero claims about it's performance. It's also not currently capable of hosting VM's (lacks SMB3) so that is a demanding use case off the table.
Other points of interest while I'm on the topic:
- File system reclamation has been "paused" (deleting data does not free up space) for quite some time. Support tells me it's a software issue and will be patched in the next release. Having just updated to 2.3 Merlot I'm not hopeful that will be soon, however I'm unaware of their release schedule and may be overly pessimistic due to lack of coffee.
- Support continues to be exceptionally prompt and friendly.
- Unfortunately, I've had to contact support (or they've contacted me) several times to clear errors. Notably most errors have been with replication & reclamation.
- On the troublesome "Vancouver" unit, I just yanked a drive counted to ten and put it back in. Got an email alert promptly afterwards "The filesystem rebalancer is currently running to protect data health. Do not remove drives or OneBlox from the Ring while this is in progress. " - fair play. Will update later when it finishes if I remember.
-
Just to confirm the setup works I connected my laptop (SSD) as the target, promptly maxed out the LAN.
-
@MattSpeller thanks for the update.
We do support SMB3, but we haven't optimized for IOPS so aren't a good fit for virtual servers (yet).
regarding your other points/questions
-
yes, this will be fixed with the 2.3.1 release that will be available later this week or early next week
-
in the area of replication and capacity reclamation is where the bug fixes are coming in for the 2.3.1 release
-
-
@SeanExablox Awesome news, thank you!
-
Any last tests anyone wants run against them? I'm going to wrap up testing and start writing it out this week.
-
@MattSpeller it's not a performance test, but have you tried out the continuous data protection feature?
-
@SeanExablox said:
@MattSpeller it's not a performance test, but have you tried out the continuous data protection feature?
Yes I have and it works a treat! I'm a big fan of the continuous snapshots
-
@MattSpeller said:
@SeanExablox said:
@MattSpeller it's not a performance test, but have you tried out the continuous data protection feature?
Yes I have and it works a treat! I'm a big fan of the continuous snapshots
Cool. Glad to hear it!
-
I've spent more time poking around and I found some cool stuff to share. I'll let the screen shots speak for themselves!
Apparently they have an internal SSD!
-
One minor complaint I have is that searching it seems to take much longer than our NAS. I will have a tough time quantifying that so please take that as opinion only.
-
This post is deleted! -
This post is deleted! -
@MattSpeller said:
One minor complaint I have is that searching it seems to take much longer than our NAS. I will have a tough time quantifying that so please take that as opinion only.
yes, smartOne will be slow as this is a database that currently has more than 5.5 million data points from every disk drive used across the Exablox installed base. We are actively working to dramatically improve its responsiveness.
-
@MattSpeller said:
I've spent more time poking around and I found some cool stuff to share. I'll let the screen shots speak for themselves!
Apparently they have an internal SSD!
you found a OneSystem easter egg
-
Finally brought this project to an end! It's been a ton of fun and I've learned quite a bit! I've strongly recommended that we revisit Exablox next year for another round of testing. All of the features fit our environment really well, especially the replication and snapshots. I'd like to thank Sean* from Exablox for helping me along this road and I hope to revisit these units soon.
*Sean I would like to single out Bob Gardiner in support for a job very well done. Every time I contacted support it was him handling the case which made it very seamless. He was prompt and efficient with zero run around or hassle. Please continue this support model, it was a pleasure to deal with.