New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
No SAN for sure, dunno about RAID tho.
How on earth was something working with 7 IOPS ? How was it feeling ?
I am not sure it is OK for the provider or the other customers. Some ppl were already suspended for this, I hope it will be tweaked somehow, I only dare to run those on VPSes I am sure will not be worth it.
M
Apparently their RAID had a rebuild, at least so they told me in a ticket. The problem is that it was 7 days ago and the IOPS still didn't go up.
This is as it is now. 7 days after the RAID rebuild had "started".
37 requests completed in 3112.2 ms, 12 iops, 0.0 mb/s
min/avg/max/mdev = 0.3/84.0/640.9/105.2 ms
root@mini:~#
No, I did shift + 1 because 7 IOPS needs an exclamation mark!
I didn't want to tell it but now that you've guessed it I cannot lie. I am on thier $10 a year France 128 MB OVZ plan.
Well, it did go up, from 7 to 12, and no, it was 7 IOPS :P
At that kind of IOPS i think the rebuild will still last a few months.
but you didnt tell me how was it feeling, I am really curious what can you do with something like that ?
M
I don't want to say anything till I have important data on it... Pretty much a simple
chmod -ron a simple directory took minutes to complete.Loading the simplest website took about 3-5 seconds.
The old FC StorageTek SAN are used on some regular openbsd nodes. They are very solid at handling load but cant compete with today DD values :P
Yes it is. this is the same setup I used for the kvm biz deploy since a few months now and all nodes are working fine so far.
@rds100 You should see this when you click on a plan on any of those pages. Since we aggegrate up the plan performance we don't show the individual scores on the filtered pages (unless there's only 1 submission of a bench for a particular plan i.e. Iperweb VZSSD2).
Clicking on the 1 Benchmark will take you straight to the benchmarks page for that plan (where you can browse through the individual reports):
Here's what happens when there's multiple benches for the same plan (you get an average):
We're doing more work around locations, so that we only show you the averages for benches in the locations you ask for (right now we show you that plan from around the world).
@zsero I wasn't aware Host Virtual was using SSD (mentions RAID10 on their plans page: http://www.vr.org/buy-vps/). I only added them last night, if someone is specifically using SSD Cache then I've created a different HDD type for that.
That was my bad, I didn't have the location information added to Iperweb. Should be fixed now!
Oh, @ServerBear - can you change my "United States" to Atlanta? :P No rush.
@Nick_A Already is (http://serverbear.com/9756/ramnode#view-locations) if you filter by country it'll show only the country, add the city filtering to see cities
@serverbear, I think the first part of your reply was for me, not for rds. My problem is that once I click on "1 benchmark" or "3 benchmarks" I shouldn't arrive to all the benchmarks and try to find the benchmark myself, but I should arrive on a page what only has the benchmarks for the selected package. And if there is only one, I should go directly to that page.
And there is one very important critical change I believe you should decide: you have to split the plans based on location. I believe there is no other option, as some of these multi location hosts are using totally different hardware at different locations and all of them are using obviously different networks at different locations. So what I mean is that taking average of something like this doesn't make sense. Obviously the Indian package has good bandwidth in India but we'll never be able to measure it, while the European system might have an amazing SSD while the Indian one only has SSD cache or a fast HDD array.
Something we're working on, easier said than done.
@ServerBear - Gotcha but I meant here: http://gyazo.com/1fdfe7302f2e354c5ca23e3f9832e821
What I mean is just the simplest idea: actually name them differently. So there would be VR256 - Netherland, VR256 - India, etc. like if they were totally independent packages. You just need to synchronize the prize, coupons, etc. I know it might not be easy, but you've already done a huge development in the last few months, so I'm sure this won't be a problem for your team.
@zsero We just pushed an update to address this issue, clicking on benchmarks now will take you to an individual plan page with graphs etc. Feedback welcomed.
Example: http://serverbear.com/536-openvz-128mb-buyvm
oooo
Graphs!
Yep, pretty cool
M
Very nice graphing and filtering now! The multi-location host problem still exists, and now with the graphs it actually looks bad, since it means that there is no correlation between continuous values in time. I know it's a hard thing to implement, I'm just saying that this graph as of now doesn't make sense (circled the 2 different locations):
BTW, the n benchmark and the plan info now brings to the same page, what is good it's just strange as in: "which one should I click, where would the other take me?"
All the benchmarks for the particular plan you're looking at are from the same location.
Here's one with multi-location: http://serverbear.com/887-ovz256-hostigation#benchmarks
Note the tabs above the graph.
One takes you to the top of the page & the other actually just brings you down to the benchmarks right away.
Oh, the tabs! (no, the one I looked at was from multiple location, it was just hidden by the rolling title).
OK, in this case the only thing what might be missing is the location data from the hoover tooltip (5 minutes to implement). And also IOPS seek values! (In my opinion DD speed still makes no sense).
But your progress is amazing! It's a totally different site then it was a few months ago!
Yeah, I was skeptic at first but the way he listens to us is great and it benefits everyone.
I just wish the tests were lighter, other than that... only good words
M
Yes, I was thinking that if the test didn't include unixbench i'd be happy to run it once every few months to better track performance over time. But a running unixbench on a dual core is just a pain in the ....
Skipping certain tests (and including others) is planned
Shouldn't be too far away.