Page 3 of 11 FirstFirst 123456 ... LastLast
Results 51 to 75 of 254
  1. #51
    Join Date
    Dec 2002
    Location
    Novosibirsk, Russia
    Posts
    1,710
    Quote Originally Posted by Orien View Post
    Certainly a valid point.

    Worth noting that the $35/mo plan would return nearly the same numbers though.
    I made a footnote indicating that.

    That means that Alto's assigned CPU cores number is the same, otherwise UnixBench would give other results.

    Quote Originally Posted by doughnet View Post
    Can't wait to see the scores for serveraxis.com. I'm looking for a east coast VPS to run my game servers. It's a free service I provide so Im always looking for budget (hence why I went with virtualsrv & piesrv that neither worked out). As long as the numbers show decent for serveraxis.com I'll be trying them out next!
    Unfortunately, I have canceled a VPS from them prior to running all the tests.

    If they would allow me to have that VPS for a day, the results will appear really soon.

    If they won't (and they are not obliged to give it for free). well, I'll order one several weeks later and run tests.
    Last edited by Master Bo; 05-05-2012 at 12:04 AM. Reason: merging responses
    IT blog is under reconstruction, please standby
    "There are two kinds of people in the world, those who haven't lost important data yet, and those who do regular backups."
    Network Monitoring Tool - contact me for coupon code to get discount

  2. #52
    Join Date
    Dec 2009
    Location
    MILLAU
    Posts
    306
    Konstantin... can you just verify that what I read is correct about CleverKite CPU is 4x E31270? I just find that an unusual setup...

  3. #53
    Join Date
    Dec 2002
    Location
    Novosibirsk, Russia
    Posts
    1,710
    Quote Originally Posted by Henrik Holben View Post
    Konstantin... can you just verify that what I read is correct about CleverKite CPU is 4x E31270? I just find that an unusual setup...
    Those are 'virtual cores', reported by UnixBench.
    If I perform 'cat /proc/cpuinfo', I'll see this

    Code:
    ...
    vendor_id	: GenuineIntel
    cpu family	: 6
    model		: 42
    model name	: Intel(R) Xeon(R) CPU E31270 @ 3.40GHz
    stepping	: 7
    cpu MHz		: 3392.298
    cache size	: 8192 KB
    physical id	: 0
    siblings	: 8
    core id		: 3
    cpu cores	: 4
    ...
    I suppose this is where the number four is taken. I should make a remark on that in the article,
    IT blog is under reconstruction, please standby
    "There are two kinds of people in the world, those who haven't lost important data yet, and those who do regular backups."
    Network Monitoring Tool - contact me for coupon code to get discount

  4. #54
    Join Date
    Dec 2009
    Location
    MILLAU
    Posts
    306
    Quote Originally Posted by masterbo View Post
    Those are 'virtual cores', reported by UnixBench.

    I suppose this is where the number four is taken. I should make a remark on that in the article,
    Yes Konstantin... it must be all 4 cores avaliable in the E3 1270 (4 cores chip) which means all CleverKite offerings have access to all 4 cores? on a Single E3 1270 with RAID10 4x SSD's

  5. #55
    Join Date
    Dec 2002
    Location
    Novosibirsk, Russia
    Posts
    1,710
    Quote Originally Posted by Henrik Holben View Post
    Yes Konstantin... it must be all 4 cores avaliable in the E3 1270 (4 cores chip) which means all CleverKite offerings have access to all 4 cores? on a Single E3 1270 with RAID10 4x SSD's
    Well I think we should ask Orien to clarify that.
    IT blog is under reconstruction, please standby
    "There are two kinds of people in the world, those who haven't lost important data yet, and those who do regular backups."
    Network Monitoring Tool - contact me for coupon code to get discount

  6. #56
    Join Date
    May 2006
    Location
    San Francisco
    Posts
    7,325
    Quote Originally Posted by Henrik Holben View Post
    Yes Konstantin... it must be all 4 cores avaliable in the E3 1270 (4 cores chip) which means all CleverKite offerings have access to all 4 cores? on a Single E3 1270 with RAID10 4x SSD's
    Technically the E3-1270s report 8 virtual cores, but yes that's more or less our current builds.

  7. #57
    Join Date
    Jan 2010
    Location
    Earth
    Posts
    12
    nice performance ....

  8. #58
    Join Date
    Mar 2011
    Posts
    659
    Thanks for making this list. I think we're about to start seeing a faster rise in popularity for this as the VPS market continues growing and disk I/O continues to be on top of the most frequent bottlenecks for the lower price range.
    Ain't here to spam my signature, I'm not desperate for sales.

  9. #59
    Join Date
    Dec 2002
    Location
    Novosibirsk, Russia
    Posts
    1,710
    Update: I have also added
    Amazon EC2 benchmarks chart
    You can also download from that page a .ZIP containing all the benchmarks data.

    Quote Originally Posted by jarland View Post
    Thanks for making this list. I think we're about to start seeing a faster rise in popularity for this as the VPS market continues growing and disk I/O continues to be on top of the most frequent bottlenecks for the lower price range.
    Yes, there are more and more SSD hosting offers nowadays. I suppose this moment is the rise of this industry. Let's observe and participate.
    IT blog is under reconstruction, please standby
    "There are two kinds of people in the world, those who haven't lost important data yet, and those who do regular backups."
    Network Monitoring Tool - contact me for coupon code to get discount

  10. #60
    Join Date
    Jan 2007
    Posts
    184
    who paid to let you test the:
    m2.4xlarge

    package for ec2? haha!

  11. #61
    Join Date
    Dec 2002
    Location
    Novosibirsk, Russia
    Posts
    1,710
    Quote Originally Posted by doughnet View Post
    who paid to let you test the:
    m2.4xlarge

    package for ec2? haha!
    No one.
    However, since there are Spot instances, I didn't spend too much even on the largest types of EC2.
    IT blog is under reconstruction, please standby
    "There are two kinds of people in the world, those who haven't lost important data yet, and those who do regular backups."
    Network Monitoring Tool - contact me for coupon code to get discount

  12. #62
    Join Date
    Dec 2002
    Location
    Novosibirsk, Russia
    Posts
    1,710
    TWIMC, I also am doing the benchmarking of other cloud-based hosting facilities, such as GreenQloud, RackspaceCloud, GigenetCloud, NephoScale, VPS.NET, VPSFarm... name ye cloud, I'll test it.

    Stay tuned, there will be announces on VPSeer.
    Last edited by Master Bo; 05-09-2012 at 01:35 AM.
    IT blog is under reconstruction, please standby
    "There are two kinds of people in the world, those who haven't lost important data yet, and those who do regular backups."
    Network Monitoring Tool - contact me for coupon code to get discount

  13. #63
    Join Date
    Mar 2012
    Posts
    162
    I couldn't tell you where you've gone wrong but your IOPS values are off by more than an order of magnitude. If I look at your website EC2 data, you list IOPS at 900/1400 for a micro. The actual value is widely measured to peak around 100 IOPS.

  14. #64
    Join Date
    Dec 2002
    Location
    Novosibirsk, Russia
    Posts
    1,710
    Quote Originally Posted by jroc View Post
    I couldn't tell you where you've gone wrong but your IOPS values are off by more than an order of magnitude. If I look at your website EC2 data, you list IOPS at 900/1400 for a micro. The actual value is widely measured to peak around 100 IOPS.
    Let's start with a question how is widely measured the above value.
    The results were received by running the fio several times and picking an average.

    I'll re-run it, if necessary, to correct the results, no problem with that.
    IT blog is under reconstruction, please standby
    "There are two kinds of people in the world, those who haven't lost important data yet, and those who do regular backups."
    Network Monitoring Tool - contact me for coupon code to get discount

  15. #65
    Join Date
    Dec 2002
    Location
    Novosibirsk, Russia
    Posts
    1,710
    Quote Originally Posted by jroc View Post
    I couldn't tell you where you've gone wrong but your IOPS values are off by more than an order of magnitude. If I look at your website EC2 data, you list IOPS at 900/1400 for a micro. The actual value is widely measured to peak around 100 IOPS.
    There's nothing better than an experiment.
    I have built fio tool and ran the IOPS checks on my *busy* t1.micro instance (it means it's not idle as those I used to collect EC2 benchmarks).

    The results of several runs of fio (the same parameters I used in the benchmarking):

    IOPS random read:
    read : io=29656KB, bw=2953.6KB/s, iops=738 , runt= 10041msec
    read : io=52892KB, bw=5259.3KB/s, iops=1314 , runt= 10057msec
    read : io=76332KB, bw=7598.3KB/s, iops=1899 , runt= 10046msec
    read : io=94128KB, bw=9350.2KB/s, iops=2337 , runt= 10067msec
    The results then vary, but do not get below 738 or higher than 2500

    IOPS random write:
    write: io=74012KB, bw=7388.7KB/s, iops=1847 , runt= 10017msec
    write: io=72428KB, bw=7229.9KB/s, iops=1807 , runt= 10018msec
    write: io=73492KB, bw=7337.5KB/s, iops=1834 , runt= 10016msec
    write: io=72076KB, bw=7195.4KB/s, iops=1798 , runt= 10017msec

    and so on - they are around 1800

    I start to think my measurements were too pessimistic in the chart I posted. IOPS for Amazon EC2 might be significantly better than I obtained in the first tests.

    The t1.micro I used is being run for many months and I do not plan to stop it in the foreseeable future. The tests can be re-run whenever required.

    Any objections?
    Last edited by Master Bo; 05-09-2012 at 07:35 PM. Reason: typos
    IT blog is under reconstruction, please standby
    "There are two kinds of people in the world, those who haven't lost important data yet, and those who do regular backups."
    Network Monitoring Tool - contact me for coupon code to get discount

  16. #66
    Yes Konstantin... it must be all 4 cores avaliable in the E3 1270 (4 cores chip) which means all CleverKite offerings have access to all 4 cores? on a Single E3 1270 with RAID10 4x SSD's
    Access to all 4 but only able to burst use them in their is credits available. If they allow all users 4 cores it kind of balances things out but can increase overhead from context swap. Might be better to balance amount of vcpu and credits/weight with amount of Ram.

  17. #67
    Join Date
    Mar 2012
    Posts
    162
    What are the peak tps reported by "iostat 1 " in a separate window while your tool is running?

  18. #68
    Join Date
    Dec 2002
    Location
    Novosibirsk, Russia
    Posts
    1,710
    Since both fio and iozone perform massive I/O operations, this is what transfers per second (tps) indicate.

    That's expected.
    IT blog is under reconstruction, please standby
    "There are two kinds of people in the world, those who haven't lost important data yet, and those who do regular backups."
    Network Monitoring Tool - contact me for coupon code to get discount

  19. #69
    Join Date
    Mar 2012
    Posts
    162
    Quote Originally Posted by masterbo View Post
    Since both fio and iozone perform massive I/O operations, this is what transfers per second (tps) indicate.

    That's expected.
    I'm asking you, what iostat tells you are the tps while you are running fio/iozone. I cannot get a single EBS volume on a Micro to give me more than around 90 IOPS. So since you are reporting a number more than 10X higher than that, I'm trying to reconcile our methodologies. Mine is to saturate the system with IOPS which are reported in iostat and max at around 90. I'm skeptical of the values reported by your tool in this test, but iostat should be accurate.

  20. #70
    Join Date
    Dec 2002
    Location
    Novosibirsk, Russia
    Posts
    1,710
    Quote Originally Posted by jroc View Post
    I'm asking you, what iostat tells you are the tps while you are running fio/iozone. I cannot get a single EBS volume on a Micro to give me more than around 90 IOPS. So since you are reporting a number more than 10X higher than that, I'm trying to reconcile our methodologies. Mine is to saturate the system with IOPS which are reported in iostat and max at around 90. I'm skeptical of the values reported by your tool in this test, but iostat should be accurate.
    I'll check what iostat shows while fio/iozone are running.
    However, fio seems very popular tool and the command line used is quite typical.

    Ever tried to repeat the measurements by running exactly what I did?

    Update: OK, I ran the fio tests on my t1.micro, while keeping 'iostat 1' on the same server.
    When writing test files, tps was circa 450. When running tests, it burst up to 5700 tps for the first 2-3 seconds and after that gradually diminished to zero, as tests were over.
    Last edited by Master Bo; 05-10-2012 at 10:38 PM. Reason: tests update added/typos removed
    IT blog is under reconstruction, please standby
    "There are two kinds of people in the world, those who haven't lost important data yet, and those who do regular backups."
    Network Monitoring Tool - contact me for coupon code to get discount

  21. #71
    Join Date
    Mar 2012
    Posts
    162
    I made a new instance and with iozone I was able to generate 250-300 Iops on a single volume. Seems this has improved quite a lot!

  22. #72
    Join Date
    Dec 2002
    Location
    Novosibirsk, Russia
    Posts
    1,710
    Quote Originally Posted by jroc View Post
    I made a new instance and with iozone I was able to generate 250-300 Iops on a single volume. Seems this has improved quite a lot!
    Quite possibly.
    I remember vaguely, that in the beginning of EC2 everything was rather cumbersome and unhurried.

    I dareday they do invest into newer storage devices.
    IT blog is under reconstruction, please standby
    "There are two kinds of people in the world, those who haven't lost important data yet, and those who do regular backups."
    Network Monitoring Tool - contact me for coupon code to get discount

  23. #73
    Join Date
    Oct 2010
    Posts
    3,662
    Just to throw it out there, we've modified our offer, and I think we're going to really impress you. Do you sign up 'secretively' or do you use freebies that hosts gave you? It might not be fair if you use a freebie, as they could put you on your own server, and you would never know...
    SiFuQi.net - Affordable Dedicated Servers in Los Angeles, California
    24x7 Support • Enterprise Grade Hardware • Automated OS Reinstalls
    Check out our reseller program, with a unique two-tiered discount.

  24. #74
    Join Date
    Dec 2002
    Location
    Novosibirsk, Russia
    Posts
    1,710
    Quote Originally Posted by EndServer View Post
    Just to throw it out there, we've modified our offer, and I think we're going to really impress you. Do you sign up 'secretively' or do you use freebies that hosts gave you? It might not be fair if you use a freebie, as they could put you on your own server, and you would never know...
    I like being impressed. No, really.

    As for SSD VPSes... Three of them (CleverKite, Avante Hosting, VirtualSRV) offered their VPSes for tesing purposes. In all the other cases I buy the services. When there are trial offers, I use trials etc., if they are the same regular plans, just at different pricing.

    By the way, I use the same real name/address in all the cases, I see no sense to hide my identity. After all, anyone can run benchmarks, that's not against ToS of the hosters I met so far.

    We can ask the mentioned hosters whether their VPSes were regular, created on the same nodes the next paid VPSes would be hosted at.

    I see no real sense in offering me VPSes created on empty nodes. Anyone, as I have said, can re-run the benchmarks and if there's significant difference, it will be obvious something is wrong.

    I prefer to trust people unless there's a very solid reason to do otherwise.

    Sorry for typos.
    Last edited by Master Bo; 05-11-2012 at 10:52 AM. Reason: typos
    IT blog is under reconstruction, please standby
    "There are two kinds of people in the world, those who haven't lost important data yet, and those who do regular backups."
    Network Monitoring Tool - contact me for coupon code to get discount

  25. #75
    Join Date
    Dec 2002
    Location
    Novosibirsk, Russia
    Posts
    1,710
    TWIMC: commencing tests and getting the first impressions of HP Cloud Services (moved to public Beta state two days ago).

    Also, if someone wishes to have, sooner rather than later, benchmarks/reviews of certain cloud services I have experience of, please navigate to the lists at the bottom of "I recommend" page and let me know what one.

    Now let's see what HP hides in their cloud...
    IT blog is under reconstruction, please standby
    "There are two kinds of people in the world, those who haven't lost important data yet, and those who do regular backups."
    Network Monitoring Tool - contact me for coupon code to get discount

Page 3 of 11 FirstFirst 123456 ... LastLast

Similar Threads

  1. ramnode ssd powered vps
    By imthebest in forum VPS Hosting
    Replies: 16
    Last Post: 01-18-2013, 01:26 AM
  2. Replies: 7
    Last Post: 05-13-2012, 04:54 PM
  3. Replies: 0
    Last Post: 04-24-2012, 04:47 PM
  4. Premium cPanel Hosting + Litespeed Powered with SSD Drive
    By kunc86 in forum Shared Hosting Offers
    Replies: 0
    Last Post: 03-15-2010, 08:34 PM

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •