Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Going to leave Buyvm
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.

Going to leave Buyvm

caiappscaiapps Member
edited November 2011 in Reviews

Buyvm is hot, check this page, http://www.doesbuyvmhavestock.com/, their stock gone quickly. I think this's because BuyVM is the cheapest one in the top 3 here (I mean LowEndBox). Everyone here are looking for a cheap and good VPS, and so many VPSs are posted every day, and we don't which ones are good and what ones are not, so the ranking by LEB became a standard for a newbe like me to pick one.

Have being with BuyVm about two months, 512M OpenVZ on Node07, but the performance was very bad, so I posted "my review" at here, http://www.lowendtalk.com/discussion/111/no-sure-why-buyvm.net-ranks-high, Francisco is a nice guy, not sure if he did something, the node seemed to be better. But I am still experiencing slowness. First of all the ssh is slow, they said this is router issue and they could not control, that's ok. But I would say about half of time, php execution is slow, database connection is slow, so the website is slow even a static page. Could not image if I did not turn on the cache.

I was about to jump out. Last Thursday, the node was down, finally they told me,

Due to multiple complaints on your node we were forced to move everyone off the node in preparation of bringing it down to upgrade, the control panel has been updated with your new node location."

I was so glad to hear that and I found I was moved to node31 from node7. I thought everything would be changed. Unfortunately, I was wrong. The performance is still unstable, sometime is ok and sometime is terrible.

Another thing I would like to mention is Memory. I never see my memory usage is over 300M. Most of time it's around 200M.

[root@www:~]# free -m
             total       used       free     shared    buffers     cached
Mem:          1024        145        878          0          0          0
-/+ buffers/cache:        145        878
Swap:            0          0          0

So, I decided to move my sites out of BuyVM this weekend.
Note that this is just my experience.

Good luck!

«1

Comments

  • FranciscoFrancisco Top Host, Host Rep, Veteran

    The SSH slow down was because our main resolver went bonkers and resolver02 didn't pickup (i hate power-dns sometimes :()

    We've been tracking some performance issues due to EXT3 degrade. Alas, most of our nodes sit around 15,000,000 (yes, million) files on them due to some users having crazy caching methods that don't cleanup, etc.

    We've been rebuilding older nodes that were on EXT3 onto EXT4, though, the worst that happened was a user got a restart as they were migrated (90%+ migrated without a restart).

    I apologize that the ride was bumpy, we've been having to mass roll changes as some of it requires hardware work and our onsite-tech only has limited time to help us these days :(

    Could you let me know what felt 'slow' besides the initial SSH login? The IO looks pretty good on the box, but maybe there's some different requirements?

    Francisco

  • fanfan Veteran

    My $15 BuyVM still work fine for me, and I'm thinking about getting another one if I can make sure the route goes through nlayer or globalcrossing instead of he.net. I've also experienced the SSH slowdown you mentioned but it's not that often for me.

    In fact the $15/yr one is even more speedy than Burst.net's $5.95/mon 512mb box.

  • FranciscoFrancisco Top Host, Host Rep, Veteran

    We can't force things to go out GBLX 100% yet, but I can for sure confirm that there is a hell of a lot less HE in the routes :)

    Our resolver box is sitting at 100% of an entire core so I think it's time for us to get away from powerdns or get it a bigger CPU to eat (it's eating a full core of an L5520 so yea)

    Francisco

  • @caiapps said: Another thing I would like to mention is Memory. I never see my memory usage is over 300M. Most of time it's around 200M.

    What means that?

    Your memory usage is high? Or is low? lol

  • @fan said: My $15 BuyVM still work fine for me, and I'm thinking about getting another one if I can make sure the route goes through nlayer or globalcrossing instead of he.net. I've also experienced the SSH slowdown you mentioned but it's not that often for me.

    My is working fine too, but the new route didn't work well from my location :( But this is acceptable ;)

  • FranciscoFrancisco Top Host, Host Rep, Veteran

    @giang said: My is working fine too, but the new route didn't work well from my location :( But this is acceptable ;)

    Log a ticket, egi has been begging us to get traceroutes in to help with route adjustments :)

    Just link to this post in it and Anthony will forward it.

    Francisco

  • fanfan Veteran

    @Francisco said: Log a ticket, egi has been begging us to get traceroutes in to help with route adjustments :)

    Just link to this post in it and Anthony will forward it.

    Francisco

    Mine is going through gblx now and it's excellent! (GBLX inbound/nlayer outbound) And I'm trying to get rid of he.net as much as I can. :P

  • @yomero said: @caiapps said: Another thing I would like to mention is Memory. I never see my memory usage is over 300M. Most of time it's around 200M.

    What means that?

    Your memory usage is high? Or is low? lol

    I meant the memory is not an issue, and 512M seems to be a lot to me.

  • @Francisco said: Could you let me know what felt 'slow' besides the initial SSH login? The IO looks pretty good on the box, but maybe there's some different requirements?

    A page took 15 secs to show. I don't think it's page issue, because sometime it ran very fast (not from cache). You know page load time is king, and no one like to wait for that long.

  • FranciscoFrancisco Top Host, Host Rep, Veteran

    @caiapps said: A page took 15 secs to show. I don't think it's page issue, because sometime it ran very fast (not from cache). You know page load time is king, and no one like to wait for that long.

    True enough, but still, it can be a number of things, especially if the issue happened yesterday for a few hours. Our resolver took a dump for about an hour or two and we didn't notice it since most people were falling back to 4.2.2.1 or otherwise.

    Anyways, i'm not going to argue or claim you're wrong, but if the issue only happened then, then that would be why :) Especially if your code/site relies on outside sites (not just in javascripts, etc, but in CURL/WGET calls)

    Francisco

  • FranciscoFrancisco Top Host, Host Rep, Veteran

    @fan said: Mine is going through gblx now and it's excellent! (GBLX inbound/nlayer outbound) And I'm trying to get rid of he.net as much as I can. :P

    If you get an IP starting with 209.* then it's going for sure through GBLX the easiest (go figure). All of our ranges are showing GBLX routes, but some ISP's haven't updated to show it (comcast is a big one).

    Francisco

  • I also experience SSH slowdowns sometimes. I never bothered making a ticket as it seemed like a minor issue (it did/does get annoying though...)

  • @maxexcloo, @caiapps: You might want to try enabling compression for your ssh connections. Put "Compression yes" in your ~/.ssh/config file or use the -C switch.

  • FranciscoFrancisco Top Host, Host Rep, Veteran

    @maxexcloo - has it improved since GBLX came on? It came online on Monday.

  • Hmmm I got it. Fran, why to use your nameservers for the our VPS's? Public dns is fine IMHO. But, this isn't the issue on thiei thread :P

  • @Francisco said: Log a ticket, egi has been begging us to get traceroutes in to help with route adjustments :)

    Just link to this post in it and Anthony will forward it.

    Francisco

    Thank you. I've opened one :D

  • I have 2 VPS' with BuyVM, a $15 a year one and a BuyVM-256 and the 256 was so slow when I got it I could not use it, even with a fresh install. I opened a ticket but got a canned response about how the host seems fine and next time report it right away (obviously I did report it right away but whatever)

    I managed to get my box setup and all it does is a PBX for one or two phone calls a day so its no big deal if it is a bit slow.

    After a few days things got better and it works fine now.

  • FranciscoFrancisco Top Host, Host Rep, Veteran

    @luma - That may have been when we were noticing some degrade. There wasn't any high loads but other things were affecting it (namely EXT3, etc). For what it's worth, the majority of our older nodes have been moved to EXT4 :) There is still a dozen or so boxes left that need it but Anthony should have them all done this week.

    @yomero 4.2.2.1 actually starts to rate limit if you do too many requests :( They were starting to rate limit our whole ranges just because of how busy things can get. I'm assuming google is the same.

    Besides, if we started pushing everyone to google, they would see 20 - 30ms added to their lookups since 8.8.8.8 isn't located on the west coast when we did tests. We've had some extremely paranoid staffers & customers around that feared getting data mined by 'the big corporations' (i'm not joking) and really got on our case about having localized resolvers :)

    Francisco

  • I think if 20-30 ms is an issue for some people, they can install their own DNS cacher. Or, maybe you can suggest using your resolvers for the people that needs it (IMHO almost nobody). I like the google ones :P No issues until now

  • maxexcloomaxexcloo Member
    edited November 2011

    I have compression on but haven't done much with my BuyVM servers recently. I'll report any issues that occur though, don't worry :)

  • FranciscoFrancisco Top Host, Host Rep, Veteran

    @yomero said: I think if 20-30 ms is an issue for some people, they can install their own DNS cacher. Or, maybe you can suggest using your resolvers for the people that needs it (IMHO almost nobody). I like the google ones :P No issues until now

    Running a caching DNS server can eat up quite a bit of RAM & CPU. Ours is doing 600M RAM and a full core :P It's a case of damned if we do, damned if we don't. For a while we simply just used 4.2.2.1. Anthony then got a fair number of tickets from people complaining about resolving delays, slower SSH logins, etc.

    Francisco

  • ...
      8    43 ms    47 ms    44 ms  203.208.131.233
      9    42 ms    42 ms    43 ms  203.208.166.61
     10   221 ms   291 ms   219 ms  203.208.153.162
     11   259 ms   274 ms   249 ms  203.208.171.154
     12   228 ms   295 ms   234 ms  paix.he.net [198.32.176.20]
     13     *        *        *     Request timed out.
     14   236 ms   234 ms   246 ms  energy-group-networks-llc.10gigabitethernet1-3.core1.sjc1.he.net [64.71.150.22]
     15   246 ms   235 ms   234 ms  173.245.86.18
     16   231 ms   265 ms   238 ms  node12.buyvm.net [205.185.112.112]
    

    Sorry to hijack this thread.... But obviously still going through HE.net... Could I/we do something to improve the only BuyVM part I dislike? :-)

  • FranciscoFrancisco Top Host, Host Rep, Veteran

    send me a PM with the source IP and i'll forward the request :)

    Francisco

    Thanked by 1TigersWay
  • 600MB Ram for a DNS server? WOW, We are using one called Simple DNS Plus and its currently using 278MB... Maybe an upgrade is in order? :D

  • @Francisco said: Running a caching DNS server can eat up quite a bit of RAM & CPU. Ours is doing 600M RAM and a full core

    I was having a lot of slowness running DNS servers in OpenVZ as resolvers, so while those are still there, they are just doing authorative DNS now, and moved to caching resolvers in KVM and stuff is lightning fast now, with one in LA and one in SC with both set to forward to the authorative servers.

    Even if you need to toss in a couple of dedi D525's, I feel we have a responsibly to provide resolvers.

    [root@dns2 ~]# free -m
                 total       used       free     shared    buffers     cached
    Mem:           249        244          4          0         25         42
    -/+ buffers/cache:        177         72
    Swap:          255         41        214
    [root@dns2 ~]# ps aux | grep named
    named     1746  3.6 59.6 189192 152004 ?       Ssl  Oct30 561:30 /usr/sbin/named -u named
    [root@dns2 ~]# w
     04:09:26 up 10 days, 17:20,  1 user,  load average: 0.09, 0.29, 0.19
    
  • For a dns cacher I have used pdnsd and it uses a joke of ram.
    After some bugs, I moved to maradns. Still a cacher, but isn't persistent, if it crashes, all is gone. What do you use? Normal bind9?

    And btw, what are the benefits of routing via gblx instead the current routes? Better pings for my gameservers?? =P

  • FranciscoFrancisco Top Host, Host Rep, Veteran

    @miTgiB said: moved to caching resolvers in KVM

    Ours are on KVM as well but i'm going to likely have to move the resolvers off powerdns since it simply can't keep up. I know Shaw uses PDNS for their resolvers but it's quite possible that they have an anycast setup within their own network.

    @yomero said: For a dns cacher I have used pdnsd and it uses a joke of ram. After some bugs, I moved to maradns. Still a cacher, but isn't persistent, if it crashes, all is gone. What do you use? Normal bind9?

    And btw, what are the benefits of routing via gblx instead the current routes? Better pings for my gameservers?? =P

    I really don't know what to swap to if pdns can't keep up. I may very well just use bind/djbdns or something :(

    As for GBLX, a LOT of isp's have GBLX in their blend so it could benefit a lot. You can ping 209.141.33.1 to see if it makes a difference. The 209's for sure have GBLX with the 205's getting it as ISP's update.

    Francisco

  • @yomero said: What do you use? Normal bind9?

    Yes, I just use bind, I have little interest in reinventing the wheel. It has served me well since 1995, why change now?

  • @Francisco if one core can't keep up just put it on two cores? Bind atleast can use multiple cores/threads, can't tell about powerdns since i am not using it.

  • FranciscoFrancisco Top Host, Host Rep, Veteran

    @rds100 said: @Francisco if one core can't keep up just put it on two cores? Bind atleast can use multiple cores/threads, can't tell about powerdns since i am not using it.

    PowerDNS can't alas =\ PowerDNS held up well before we did the build in August but it seems it simply can't keep up with our work loads :(

    I'll bring resolver02 down tomorrow, put it on bind then bring down 01 and see how it goes.

    Francisco

Sign In or Register to comment.