Howdy, Stranger!

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

Sign In Register

Quick Links


Categories

In this Discussion

› › Experiencing slow disk performance, what is the interpretation of this yabs.
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.

Experiencing slow disk performance, what is the interpretation of this yabs.

k9bangerk9banger Member
in Help

I have been experiencing slow performance from a provider and I think I should move away.

They asked me to run a yabs when I complained. What do the disk speeds here indicate?

What does it say about the performance of the node it took 60 min 41 sec

# ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
#              Yet-Another-Bench-Script              #
#                     v2024-12-17                    #
# https://github.com/masonr/yet-another-bench-script #
# ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #

Wed 18 Dec 17:02:24 GMT 2024

Basic System Information:
---------------------------------
Uptime     : 0 days, 0 hours, 13 minutes
Processor  : AMD EPYC 7302 16-Core Processor
CPU cores  : 2 @ 2999.955 MHz
AES-NI     : ✔ Enabled
VM-x/AMD-V : ❌ Disabled
RAM        : 1.9 GiB
Swap       : 2.0 GiB
Disk       : 1.7 TiB
Distro     : Debian GNU/Linux 12 (bookworm)
Kernel     : 6.1.0-28-amd64
VM Type    : XEN
IPv4/IPv6  : ✔ Online / ❌ Offline

IPv4 Network Information:
---------------------------------
ISP        : <Redacted>
ASN        : <Redacted>
Host       : <Redacted>
Location   : <Redacted>
Country    : <Redacted>

fio Disk Speed Tests (Mixed R/W 50/50) (Partition /dev/xvdb1):
---------------------------------
Block Size | 4k            (IOPS) | 64k           (IOPS)
  ------   | ---            ----  | ----           ----
Read       | 46.35 MB/s   (11.5k) | 152.92 MB/s   (2.3k)
Write      | 46.41 MB/s   (11.6k) | 153.72 MB/s   (2.4k)
Total      | 92.76 MB/s   (23.1k) | 306.64 MB/s   (4.7k)
           |                      |
Block Size | 512k          (IOPS) | 1m            (IOPS)
  ------   | ---            ----  | ----           ----
Read       | 78.70 MB/s     (153) | 67.19 MB/s      (65)
Write      | 82.88 MB/s     (161) | 71.73 MB/s      (70)
Total      | 161.59 MB/s    (314) | 138.93 MB/s    (135)

iperf3 Network Speed Tests (IPv4):
---------------------------------
Provider        | Location (Link)           | Send Speed      | Recv Speed      | Ping
-----           | -----                     | ----            | ----            | ----
Clouvider       | London, UK (10G)          | busy            | busy            | 77.4 ms
Eranium         | Amsterdam, NL (100G)      | 2.13 Gbits/sec  | 1.14 Gbits/sec  | 79.8 ms
Uztelecom       | Tashkent, UZ (10G)        | busy            | 314 Mbits/sec   | 172 ms
Leaseweb        | Singapore, SG (10G)       | 689 Mbits/sec   | 359 Mbits/sec   | 239 ms
Clouvider       | Los Angeles, CA, US (10G) | 2.06 Gbits/sec  | 422 Mbits/sec   | 74.7 ms
Leaseweb        | NYC, NY, US (10G)         | 2.68 Gbits/sec  | 1.58 Gbits/sec  | 9.53 ms
Edgoo           | Sao Paulo, BR (1G)        | 1.45 Gbits/sec  | 394 Mbits/sec   | 132 ms

Geekbench 6 Benchmark Test:
---------------------------------
Test            | Value
                |
Single Core     | 1149
Multi Core      | 963
Full Test       | https://browser.geekbench.com/v6/cpu/<redacted>

YABS completed in 60 min 41 sec

Comments

  • danblazedanblaze Member

    It's basically HDD level speed, definitely not fast. More notable is the multi-core performance, you have two CPUs but the multi-core scores lower than the single core, which would seem to indicate something, such as a high load on this host of theirs.

    Thanked by 1k9banger
  • k9bangerk9banger Member

    @danblaze said:
    It's basically HDD level speed, definitely not fast.

    Is it basically acceptable for an HDD, and does it look like it is a locally hard disk rather than one on a SAN or a remote networked drive?

    More notable is the multi-core performance, you have two CPUs but the multi-core scores lower than the single core, which would seem to indicate something, such as a high load on this host of theirs.

    When I enter a command it takes too long to before I get a response, but the speed of the output looks normal. Is that due to the disk latency? Is there a test for that?

  • ralfralf Member

    @k9banger said:
    When I enter a command it takes too long to before I get a response, but the speed of the output looks normal. Is that due to the disk latency? Is there a test for that?

    That's pretty normal if you have a slow disk with high number of IOPS in flight. Partially due to latency, but also because there are a lot of requests to get serviced before yours.

    Could also be an indication that RAM is oversold and that your VM has been swapped out, perhaps even to HDD. If your VM has been idle for a long time, and you only experience this the first time you run a command, this might well happen even if RAM isn't over-provisioned, because the system as a whole (from the perspective of the host OS) can make better use of the RAM for disk buffer caches than storing your rarely used stuff. That said, the host can change its "swappiness" to prevent this behaviour.

    Maybe the easiest thing you can do is run top in another connection while doing your typical tasks and keep an eye on different percentages. If it's spending most of the time doing system IO, that's just because you're on a disk that's either slow or trashed by many other people or both.

    As you have a poor multi-core score, it's obvious that the node is quite busy. You might well be seeing a high steal in top as well in this case.

    Thanked by 1k9banger
  • FalzoFalzo Member
    edited December 2024

    disk speed are fine for HDD. Some caching is there to help with small blocksizes, so overall seems fine.

    At least that is probably not the issue you are experiencing unless you are hammering that disk with IO intensive stuff. Run top to see steal and iowait.
    could well be that the CPU is somewhat limited or it is just overcrowded.

    Is this XEN HV or PV?

  • k9bangerk9banger Member

    @ralf said:

    @k9banger said:
    When I enter a command it takes too long to before I get a response, but the speed of the output looks normal. Is that due to the disk latency? Is there a test for that?

    That's pretty normal if you have a slow disk with high number of IOPS in flight. Partially due to latency, but also because there are a lot of requests to get serviced before yours.

    Could also be an indication that RAM is oversold and that your VM has been swapped out, perhaps even to HDD. If your VM has been idle for a long time, and you only experience this the first time you run a command, this might well happen even if RAM isn't over-provisioned, because the system as a whole (from the perspective of the host OS) can make better use of the RAM for disk buffer caches than storing your rarely used stuff. That said, the host can change its "swappiness" to prevent this behaviour.

    It happens on every command and every account. I have 4 accounts on the server and it happens to all of them.

    I had a dd test as low as 1MB/s

    Maybe the easiest thing you can do is run top in another connection while doing your typical tasks and keep an eye on different percentages. If it's spending most of the time doing system IO, that's just because you're on a disk that's either slow or trashed by many other people or both.

    As you have a poor multi-core score, it's obvious that the node is quite busy. You might well be seeing a high steal in top as well in this case.

    How do you check CPU steal in top?

  • k9bangerk9banger Member

    @Falzo said:

    Is this XEN HV or PV?

    How can I check this? Do I have to ask the provider?

  • ralfralf Member

    @k9banger said:

    @ralf said:

    @k9banger said:
    When I enter a command it takes too long to before I get a response, but the speed of the output looks normal. Is that due to the disk latency? Is there a test for that?

    That's pretty normal if you have a slow disk with high number of IOPS in flight. Partially due to latency, but also because there are a lot of requests to get serviced before yours.

    Could also be an indication that RAM is oversold and that your VM has been swapped out, perhaps even to HDD. If your VM has been idle for a long time, and you only experience this the first time you run a command, this might well happen even if RAM isn't over-provisioned, because the system as a whole (from the perspective of the host OS) can make better use of the RAM for disk buffer caches than storing your rarely used stuff. That said, the host can change its "swappiness" to prevent this behaviour.

    It happens on every command and every account. I have 4 accounts on the server and it happens to all of them.

    Interesting. If you run the same command that doesn't involve IO is it also slow? There are different types of commands, so shell builtins like echo should be instant, you can force a simple from-disk command, e.g. /bin/id may need to load things from disk on the first run but should be instant on the second run, and other stuff that hits disk might be slow every time, e.g. ls -l /var/log

    You can also run these commands prefixed with time to see what the system is doing.

    If they're all slow, that suggests heavy CPU load on the host. If it's just the disk ones, then it's probably just someone else hammering the node.

    I'd also say their support should probably be investigating the underlying cause themselves if it's that bad.

    I had a dd test as low as 1MB/s

    :open_mouth:

    Maybe the easiest thing you can do is run top in another connection while doing your typical tasks and keep an eye on different percentages. If it's spending most of the time doing system IO, that's just because you're on a disk that's either slow or trashed by many other people or both.

    As you have a poor multi-core score, it's obvious that the node is quite busy. You might well be seeing a high steal in top as well in this case.

    How do you check CPU steal in top?

    It's on the right, 0.0% st (ideally).

  • FalzoFalzo Member

    @k9banger said:

    @Falzo said:

    Is this XEN HV or PV?

    How can I check this? Do I have to ask the provider?

    Should be part of the product or offer description.

  • k9bangerk9banger Member

    @ralf said:

    @k9banger said:

    @ralf said:

    @k9banger said:

    How do you check CPU steal in top?

    It's on the right, 0.0% st (ideally).

    It goes up to 0.2 every few seconds.

  • gbzret4dgbzret4d Member

    @k9banger said:

    @ralf said:

    @k9banger said:

    @ralf said:

    @k9banger said:

    How do you check CPU steal in top?

    It's on the right, 0.0% st (ideally).

    It goes up to 0.2 every few seconds.

    0.2% is no problem at all

  • ZhenmueZhenmue Member

    @k9banger said:

    @ralf said:

    @k9banger said:

    @ralf said:

    @k9banger said:

    How do you check CPU steal in top?

    It's on the right, 0.0% st (ideally).

    It goes up to 0.2 every few seconds.

    check that ona 2nd connection, while on the first one you run the yabs.

    Thanked by 1k9banger
  • k9bangerk9banger Member

    @JoseDieguez said:

    check that ona 2nd connection, while on the first one you run the yabs.

    Good idea

  • k9bangerk9banger Member

    @gbzret4d said:

    @k9banger said:

    @ralf said:

    @k9banger said:

    @ralf said:

    @k9banger said:

    How do you check CPU steal in top?

    It's on the right, 0.0% st (ideally).

    It goes up to 0.2 every few seconds.

    0.2% is no problem at all

    At what level is it a problem?

  • dev_vpsdev_vps Member
    edited December 2024

    This is disk io for 1 TB HDD storage vps and it is working pretty smoothly for my use cases.

    fio Disk Speed Tests (Mixed R/W 50/50) (Partition /dev/vda1):
---------------------------------
Block Size | 4k            (IOPS) | 64k           (IOPS)
  ------   | ---            ----  | ----           ----
Read       | 3.22 MB/s      (806) | 28.85 MB/s     (450)
Write      | 3.25 MB/s      (813) | 29.24 MB/s     (456)
Total      | 6.47 MB/s     (1.6k) | 58.10 MB/s     (906)
           |                      |
Block Size | 512k          (IOPS) | 1m            (IOPS)
  ------   | ---            ----  | ----           ----
Read       | 94.35 MB/s     (184) | 133.55 MB/s    (130)
Write      | 99.36 MB/s     (194) | 142.45 MB/s    (139)
Total      | 193.71 MB/s    (378) | 276.01 MB/s    (269)
    Thanked by 1k9banger
  • k9bangerk9banger Member

    Are these Geekbench values acceptable for the CPU types involved.

    ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ##

    Yet-Another-Bench-Script

    v2024-12-17

    https://github.com/masonr/yet-another-bench-script

    ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ##

    Wed 18 Dec 17:02:24 GMT 2024

    Basic System Information:

    Uptime : 0 days, 0 hours, 13 minutes
    Processor : AMD EPYC 7302 16-Core Processor
    CPU cores : 2 @ 2999.955 MHz
    AES-NI : ✔ Enabled
    VM-x/AMD-V : ❌ Disabled
    RAM : 1.9 GiB
    Swap : 2.0 GiB
    Disk : 1.7 TiB
    Distro : Debian GNU/Linux 12 (bookworm)
    Kernel : 6.1.0-28-amd64
    VM Type : XEN
    IPv4/IPv6 : ✔ Online / ❌ Offline

    Geekbench 6 Benchmark Test:

    ---------------------------------

    Test | Value

    Single Core | 1149

    Multi Core | 963

    Full Test | https://browser.geekbench.com/v6/cpu/

    YABS completed in 60 min 41 sec

  • DataRecoveryDataRecovery Member
    edited December 2024

    @k9banger said:
    Are these Geekbench values acceptable for the CPU types involved.
    VM Type : XEN

    Quite likely yes.

    Xen is known to be limiting performance in benchmarks, but distributing resources more evenly among users. So it's more likely that you won't get something blazing fast, but won't be affected by other clients at the same time.

    Better look at real performance of your app/script/website on Xen and KVM.

  • k9bangerk9banger Member
    edited December 2024

    @dev_vps said:
    This is disk io for 1 TB HDD storage vps and it is working pretty smoothly for my use cases.

    fio Disk Speed Tests (Mixed R/W 50/50) (Partition /dev/vda1):
---------------------------------
Block Size | 4k            (IOPS) | 64k           (IOPS)
  ------   | ---            ----  | ----           ----
Read       | 3.22 MB/s      (806) | 28.85 MB/s     (450)
Write      | 3.25 MB/s      (813) | 29.24 MB/s     (456)
Total      | 6.47 MB/s     (1.6k) | 58.10 MB/s     (906)
           |                      |
Block Size | 512k          (IOPS) | 1m            (IOPS)
  ------   | ---            ----  | ----           ----
Read       | 94.35 MB/s     (184) | 133.55 MB/s    (130)
Write      | 99.36 MB/s     (194) | 142.45 MB/s    (139)
Total      | 193.71 MB/s    (378) | 276.01 MB/s    (269)

    If that is the case then the VPS probably has a problem with loading commands fast enough and quickly and swaps them out immediately after they've executed.

    Is there a tool to test how quickly disks respond to commands?

  • dev_vpsdev_vps Member

    @k9banger said:

    @dev_vps said:
    This is disk io for 1 TB HDD storage vps and it is working pretty smoothly for my use cases.

    fio Disk Speed Tests (Mixed R/W 50/50) (Partition /dev/vda1):
---------------------------------
Block Size | 4k            (IOPS) | 64k           (IOPS)
  ------   | ---            ----  | ----           ----
Read       | 3.22 MB/s      (806) | 28.85 MB/s     (450)
Write      | 3.25 MB/s      (813) | 29.24 MB/s     (456)
Total      | 6.47 MB/s     (1.6k) | 58.10 MB/s     (906)
           |                      |
Block Size | 512k          (IOPS) | 1m            (IOPS)
  ------   | ---            ----  | ----           ----
Read       | 94.35 MB/s     (184) | 133.55 MB/s    (130)
Write      | 99.36 MB/s     (194) | 142.45 MB/s    (139)
Total      | 193.71 MB/s    (378) | 276.01 MB/s    (269)

    If that is the case then the VPS probably has a problem with loading commands fast enough and quickly and swaps them out immediately after they've executed.

    Is there a tool to test how quickly disks respond to commands.

    I am streaming 4k video to my iphone/ipad/tv with no issues.

    Debian 11 / webmin is my vps configuration.

  • k9bangerk9banger Member

    The problem I'm experience seems to be the problem @johngko is experiencing here - https://lowendtalk.com/discussion/comment/4139263/#Comment_4139263.

    It is a storage node but it should be able to respond to commands quickly. These are not commands to do complex stuff, just information retrievel commands such as docker ps -a or docker info.

  • dev_vpsdev_vps Member
    edited December 2024

    @k9banger said:
    Are these Geekbench values acceptable for the CPU types involved.

     Geekbench 6 Benchmark Test:
 ---------------------------------
 Test            | Value
                
 Single Core     | 1149
 Multi Core      |  963
 Full Test       | https://browser.geekbench.com/v6/cpu/

    Here is correct GB6 score link for your yabs
    https://browser.geekbench.com/v6/cpu/9494894

  • dev_vpsdev_vps Member
    edited December 2024

    @k9banger said:
    The problem I'm experience seems to be the problem @johngko is experiencing here - https://lowendtalk.com/discussion/comment/4139263/#Comment_4139263.

    It is a storage node but it should be able to respond to commands quickly.
    These are not commands to do complex stuff, just information retrievel commands such as docker ps -a or docker info.

    Do not take storage vps as normal compute vps

    Use hybrid vps if such cases
    My hybrid vps has 150GB NVMe + 500 GB HDD

  • k9bangerk9banger Member

    @dev_vps said:

    @k9banger said:
    The problem I'm experience seems to be the problem @johngko is experiencing here - https://lowendtalk.com/discussion/comment/4139263/#Comment_4139263.

    It is a storage node but it should be able to respond to commands quickly.
    These are not commands to do complex stuff, just information retrievel commands such as docker ps -a or docker info.

    Do not take storage vps as normal compute vps

    Use hybrid vps if such cases
    My hybrid vps has 150GB NVMe + 500 GB HDD

    Who is the provider of this VPS?

  • dev_vpsdev_vps Member

    @k9banger said:

    @dev_vps said:

    @k9banger said:
    The problem I'm experience seems to be the problem @johngko is experiencing here - https://lowendtalk.com/discussion/comment/4139263/#Comment_4139263.

    It is a storage node but it should be able to respond to commands quickly.
    These are not commands to do complex stuff, just information retrievel commands such as docker ps -a or docker info.

    Do not take storage vps as normal compute vps

    Use hybrid vps if such cases
    My hybrid vps has 150GB NVMe + 500 GB HDD

    Who is the provider of this VPS?

    https://lowendtalk.com/discussion/comment/4187199/#Comment_4187199

    @CharityHost_org

    Thanked by 2k9banger CharityHost_org
Sign In or Register to comment.