|
|||
|
http://web-hosting.domaintools.com/a/vps-hosting
A new performance monitoring service from the excellent folks at Domaintools. Check it out. They have taken out a VPSLink4 account and used it to compile various benchmarks. They've done the same thing on a variety of other VPS and shared hosting sites. Their tests show a consistent pattern of sub-par performance from VPSLink compared to other VPS hosts over the last month on two variables: CPU speed, and in particular disk speed. However peak uptime, uptime, web speed and support response rate in the range of good to very good. The (automated) statistical review page for VPSLink is at: http://web-hosting.domaintools.com/a...slink-4/review (Flash required) VERY interesting. This data confirms some of the issues raised in this forum about disk and CPU sluggishness. I'll certainly be keeping an eye on their stats in future. The stats seem credible, and the presentation of the data is very nice. FYI - and any comments from VPSLink... |
|
|||
|
It amazes me everytime when I try to do the 'tar zxf' command in my VPSLink4 box.. It's like taking forever to execute the task and at least 3-5 times longer than the time needed if executed in my own 7200rpm HDD.
Not only that, I'd like to ask if OpenVZ allows 'ram overselling' by making use of the Swap in their server end? Because I've noticed that whenever you enable caching(which uses RAM) of PHP scripts with eAccelerator or APC, it sometimes takes around 16 seconds++ to run scripts like vpsinfo.php(around 1 secone needed without the caching turned on)... Isn't it amazing? It runs faster after 5-10 times of refreshes at around 1 second and then becomes slow again after maybe one hour.. Has anyone noticed it? It seems to me that those not so active ram usages will be buried into swap space.. But I don't know if my assumption is true or not. Other than those 2 issues, others work fine to me. |
|
|||
|
Last night an emerge sync took nearly 2 hours. Granted my last sync was a week ago, I'd expect it to be around 30mins, but not two hours? The disk IO is definitely slow.
Quote:
|
|
|||
|
My node is showing currently from top:
Mem: 6221376k total, 6159532k used, 61844k free, 162900k buffers Swap: 8192504k total, 2677152k used, 5515352k free, 2073212k cached And if around 2.67GB of Swap is being used actively(keep in mind that not every VPS is using the whole portion of the ram available to them yet, I can't imagine what if everyone is using the whole portion of the ram 'guaranteed' to them...), then no doubt the disk IO is so slow. I'd rather they increase the total amount of ram to sell more than to use the swap as ram to sell more! |
|
|||
|
So judging from your node, I would think that Link1 to Link3(single processor) servers may be actually a whole lot faster than the Link4(two processors) or higher plans... Faster processors with real slow disk IO, what's the use?
Or maybe it's just because of your node is still new? Sigh... |
|
|||
|
I don't think you can correlate VPS performance with the amount of swapfile in use on the host machine. I've read that Virtuozzo/OpenVZ makes more vigorious use of the swapfile than a dedicated machine, and I've experienced VPS's on host nodes showing lots of swap usage that still provided excellent disk i/o.
|
|
|||
|
I'm interested to hear what VPSLink has to say about this...
|
|
|||
|
We're working on a few projects right now to make sure we get the load on these boxes under better control. For what it's worth for the month of October we were the most reliable in terms of uptime and speed according to realmetrics.com (where domaintools.com gets the data).
|
|
|||
|
Sounds good
![]() I shall be moving my sites back onto my box soon after having them in a very small hostgator account for a few months
|
![]() |
| Bookmarks |
| Thread Tools | |
| Display Modes | |
|
|