• This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn more.

Linode vs Rackspace

Andy.N

Well-known member
#1
Considering Linode SSD VPS so I got a 8GB server located in New Jersey to test. Run @eva2000 centminmod benchmark script and I'm puzzled by the much slower bandwidth from Linode compared to the Rackspace server (currently located in Virginia, same east coast US).
Code:
-------------------------------------------
Running bandwidth benchmark...
-------------------------------------------

----------------------------------------------
Download from Cachefly (http://cachefly.cachefly.net/100mb.test)
Download Cachefly: 2.87MB/s

-------------------------------------------
USA bandwidth tests...
-------------------------------------------

----------------------------------------------
Download from Linode, Atlanta, GA, USA (http://speedtest.atlanta.linode.com/100MB-atlanta.bin)
Download Linode, Atlanta, GA, USA: 15.7MB/s
----------------------------------------------
Download from Linode, Dallas, TX, USA (http://speedtest.dallas.linode.com/100MB-dallas.bin)
Download Linode, Dallas, TX, USA: 1.55MB/s
----------------------------------------------
Download from Leaseweb, Manassas, VA, USA (http://mirror.us.leaseweb.net/speedtest/100mb.bin)
Download Leaseweb, Manassas, VA, USA: 9.44MB/s
----------------------------------------------
Download from Softlayer, Seattle, WA, USA (http://speedtest.sea01.softlayer.com/downloads/test100.zip)
Download Softlayer, Seattle, WA, USA: 3.15MB/s
----------------------------------------------
Download from Softlayer, San Jose, CA, USA (http://speedtest.sjc01.softlayer.com/downloads/test100.zip)
Download Softlayer, San Jose, CA, USA: 942KB/s
----------------------------------------------
Download from Softlayer, Washington, DC, USA (http://speedtest.wdc01.softlayer.com/downloads/test100.zip)
Download Softlayer, Washington, DC, USA: 8.36MB/s
----------------------------------------------
Download from VersaWeb, Las Vegas, Nevada (http://199.47.210.50/100mbtest.bin)
Download VersaWeb, Las Vegas, Nevada: 3.48MB/s
----------------------------------------------
Download from OVH, BHS, Canada (http://bhs.proof.ovh.net/files/100Mio.dat)
Download OVH, BHS, Canada: 92.2MB/s
----------------------------------------------
Download from Vultr, Los Angeles, California (http://lax-ca-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Los Angeles, California: 20.0MB/s
----------------------------------------------
Download from Vultr, Seattle, Washington (http://wa-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Seattle, Washington: 1008KB/s
----------------------------------------------
Download from Vultr, Dallas, Texas (http://tx-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Dallas, Texas: 27.8MB/s
----------------------------------------------
Download from Vultr, Chicago, Illinois (http://il-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Chicago, Illinois: 10.0MB/s
----------------------------------------------
Download from Vultr, Atlanta, Georgia (http://ga-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Atlanta, Georgia: 16.9MB/s
----------------------------------------------
Download from Vultr, Miami, Florida (http://fl-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Miami, Florida: 1.33MB/s
----------------------------------------------
Download from Vultr, New York / New Jersey (http://nj-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, New York / New Jersey: 8.03MB/s

-------------------------------------------
Asia bandwidth tests...
-------------------------------------------

----------------------------------------------
Download from Linode, Tokyo, JP (http://speedtest.tokyo.linode.com/100MB-tokyo.bin)
Download Linode, Tokyo, JP: 670KB/s
----------------------------------------------
Download from Softlayer, Singapore (http://speedtest.sng01.softlayer.com/downloads/test100.zip)
Download Softlayer, Singapore: 1.48MB/s
----------------------------------------------
Download from Vultr, Tokyo, Japan (http://hnd-jp-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Tokyo, Japan: 1.46MB/s

-------------------------------------------
Europe bandwidth tests...
-------------------------------------------

----------------------------------------------
Download from Linode, London, UK (http://speedtest.london.linode.com/100MB-london.bin)
Download Linode, London, UK: 1.75MB/s
----------------------------------------------
Download from OVH, Paris, France (http://proof.ovh.net/files/100Mio.dat)
Download OVH, Paris, France: 4.23MB/s
----------------------------------------------
Download from SmartDC, Rotterdam, Netherlands (http://mirror.i3d.net/100mb.bin)
Download SmartDC, Rotterdam, Netherlands: 2.63MB/s
----------------------------------------------
Download from Vultr, Amsterdam, Netherlands (http://ams-nl-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Amsterdam, Netherlands: 16.5MB/s
----------------------------------------------
Download from Vultr, London, UK (http://lon-gb-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, London, UK: 3.87MB/s
----------------------------------------------
Download from Vultr, Paris, France (http://par-fr-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Paris, France: 3.17MB/s

-------------------------------------------
Australia bandwidth tests...
-------------------------------------------

----------------------------------------------
Download from Vultr, Sydney, Australia (http://syd-au-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Sydney, Australia: 2.18MB/s
 

Andy.N

Well-known member
#2
Compared this to the result from Rackspace
Code:
-------------------------------------------
Running bandwidth benchmark...
-------------------------------------------

----------------------------------------------
Download from Cachefly (http://cachefly.cachefly.net/100mb.test)
Download Cachefly: 59.4MB/s

-------------------------------------------
USA bandwidth tests...
-------------------------------------------

----------------------------------------------
Download from Linode, Atlanta, GA, USA (http://speedtest.atlanta.linode.com/100MB-atlanta.bin)
Download Linode, Atlanta, GA, USA: 25.7MB/s
----------------------------------------------
Download from Linode, Dallas, TX, USA (http://speedtest.dallas.linode.com/100MB-dallas.bin)
Download Linode, Dallas, TX, USA: 15.2MB/s
----------------------------------------------
Download from Leaseweb, Manassas, VA, USA (http://mirror.us.leaseweb.net/speedtest/100mb.bin)
Download Leaseweb, Manassas, VA, USA: 5.63MB/s
----------------------------------------------
Download from Softlayer, Seattle, WA, USA (http://speedtest.sea01.softlayer.com/downloads/test100.zip)
Download Softlayer, Seattle, WA, USA: 15.8MB/s
----------------------------------------------
Download from Softlayer, San Jose, CA, USA (http://speedtest.sjc01.softlayer.com/downloads/test100.zip)
Download Softlayer, San Jose, CA, USA: 22.5MB/s
----------------------------------------------
Download from Softlayer, Washington, DC, USA (http://speedtest.wdc01.softlayer.com/downloads/test100.zip)
Download Softlayer, Washington, DC, USA: 52.1MB/s
----------------------------------------------
Download from VersaWeb, Las Vegas, Nevada (http://199.47.210.50/100mbtest.bin)
Download VersaWeb, Las Vegas, Nevada: 20.5MB/s
----------------------------------------------
Download from OVH, BHS, Canada (http://bhs.proof.ovh.net/files/100Mio.dat)
Download OVH, BHS, Canada: 5.54MB/s
----------------------------------------------
Download from Vultr, Los Angeles, California (http://lax-ca-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Los Angeles, California: 20.4MB/s
----------------------------------------------
Download from Vultr, Seattle, Washington (http://wa-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Seattle, Washington: 17.4MB/s
----------------------------------------------
Download from Vultr, Dallas, Texas (http://tx-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Dallas, Texas: 13.8MB/s
----------------------------------------------
Download from Vultr, Chicago, Illinois (http://il-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Chicago, Illinois: 39.7MB/s
----------------------------------------------
Download from Vultr, Atlanta, Georgia (http://ga-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Atlanta, Georgia: 42.8MB/s
----------------------------------------------
Download from Vultr, Miami, Florida (http://fl-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Miami, Florida: 38.3MB/s
----------------------------------------------
Download from Vultr, New York / New Jersey (http://nj-us-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, New York / New Jersey: 49.8MB/s

-------------------------------------------
Asia bandwidth tests...
-------------------------------------------

----------------------------------------------
Download from Linode, Tokyo, JP (http://speedtest.tokyo.linode.com/100MB-tokyo.bin)
Download Linode, Tokyo, JP: 6.10MB/s
----------------------------------------------
Download from Softlayer, Singapore (http://speedtest.sng01.softlayer.com/downloads/test100.zip)
Download Softlayer, Singapore: 3.10MB/s
----------------------------------------------
Download from Vultr, Tokyo, Japan (http://hnd-jp-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Tokyo, Japan: 7.38MB/s

-------------------------------------------
Europe bandwidth tests...
-------------------------------------------

----------------------------------------------
Download from Linode, London, UK (http://speedtest.london.linode.com/100MB-london.bin)
Download Linode, London, UK: 10.2MB/s
----------------------------------------------
Download from OVH, Paris, France (http://proof.ovh.net/files/100Mio.dat)
Download OVH, Paris, France: 3.12MB/s
----------------------------------------------
Download from SmartDC, Rotterdam, Netherlands (http://mirror.i3d.net/100mb.bin)
Download SmartDC, Rotterdam, Netherlands: 3.19MB/s
----------------------------------------------
Download from Vultr, Amsterdam, Netherlands (http://ams-nl-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Amsterdam, Netherlands: 16.8MB/s
----------------------------------------------
Download from Vultr, London, UK (http://lon-gb-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, London, UK: 14.5MB/s
----------------------------------------------
Download from Vultr, Paris, France (http://par-fr-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Paris, France: 15.0MB/s

-------------------------------------------
Australia bandwidth tests...
-------------------------------------------

----------------------------------------------
Download from Vultr, Sydney, Australia (http://syd-au-ping.vultr.com/vultr.com.100MB.bin)
Download Vultr, Sydney, Australia: 6.79MB/s
 

WSWD

Well-known member
#3
Oversold, overloaded server or network? Or perhaps Linode's bandwidth providers are just utilizing congested routes.

Download from OVH, BHS, Canada (http://bhs.proof.ovh.net/files/100Mio.dat)
Download OVH, BHS, Canada: 92.2MB/s
This shows that that the server/VPS/network is perfectly capable of achieving gbit speeds, but why it's not getting those all the time is anyone's guess.
 

Xon

Well-known member
#4
@WSWD, check the trace routes & ping latency. I'm betting Linode doesn't have direct connectivity to some of those locations where as rackspace may have.
 

WSWD

Well-known member
#5
@WSWD, check the trace routes & ping latency. I'm betting Linode doesn't have direct connectivity to some of those locations where as rackspace may have.
It's still horrendously slow, even taking that into consideration. This is especially true when you look at the Cachefly results. Basically ~22mbps? WAY too slow. There is something very wrong with the results that have been posted. There is absolutely no way in hell that the east coast to west coast (i.e. San Jose) should ever be <1 mbps. You would get better speeds from the west coast to Europe or Asia.

I don't know what the problem is, but there is most definitely an issue somewhere.
 

Jim Boy

Well-known member
#9
TBH, I'm not sure why you care. Linode and Rackspace are really aimed at different markets. If all you want is a server capable of running a forum about some special interest group with relatively low concurrent users at minimal cost, then you go Linode, which runs XF just fine. If you want a really robust system with bells and whistles to support a mission critical forum, but is going to cost you a bit more then you go Rackspace.
 

Marcus

Well-known member
#10
Linode was dead slow when i tested it. Their cpu speed is VERY volatile. It seems you do not get a dedicated cpu core there.
 

WSWD

Well-known member
#12
Are you referring to the 16CPU, 48GB RAM, 1152GB SSD space for almost $500/month that you tested?

ROFL!! You can hardly compare the performance of what is really a beefy dedicated server costing $500/mo. to a basic 1 or 2 core VPS through them. That's just an extremely silly comparison to make. I highly doubt @Marcus was running that configuration, or else yeah...he probably would be getting great performance.
 

Marcus

Well-known member
#13
I tested the 40 USD plan around two months ago. Their CPU cores were highly volatile. I tested with both my xenforo community and with the php test program that's included within the Zend package (as much as I remember).

On amazon I always got the same results everytime I run the test, I always got the same results on my former dedicated server. On linode I got highly mixed results. I showed the linode team my results they didn't care about that. You can also find several topics about huge differences in cpu speed at their official linode forums.

My community is rather small, so I focus on giving each member the same very fast experience. It really doesn't make any difference whether I would have 2 or 12 cpu cores. My community now runs on 4 cores at the same speed as with 12 cores before, the two cores now are faster and not all cores are needed at the same time.

My current Amazon setup scales indefinitely, meaning that when my server is maxed out at defined CPU level, more cpu core web servers are activated. I could run now like 1000 cpu cores, they are just not needed. If you really care about cpu speed you should have dedicated access to the server to set up cores to performance status and not on demand, but that's another topic.
 

eva2000

Well-known member
#14
strange then as that's exact plan I am on $40 Linode 4GBVPS for my forums https://community.centminmod.com/ (forums are 4 months old and been with Linode since day 1)

I've tested all the Linodes from 2GB all the way up to 48GB FYI https://blog.centminmod.com/346 :)

@Marcus what Amazon plan you on ? Would like to test it out for myself :)

I tested with both my xenforo community and with the php test program that's included within the Zend package (as much as I remember).
FYI, php Zend test package is single threaded so you won't scale with number of cpu cores ;)

here's php Zend/bench.php and Zend/micro_bench.php tested capturing the cpu utilisation too, it only ever uses 1 cpu thread = 99-100%

Code:
-------------------------------------------
Run PHP test Zend/bench.php
-------------------------------------------

simple             0.099
simplecall         0.109
simpleucall        0.146
simpleudcall       0.148
mandel             0.230
mandel2            0.268
ackermann(7)       0.115
ary(50000)         0.016
ary2(50000)        0.015
ary3(2000)         0.133
fibo(30)           0.305
hash1(50000)       0.027
hash2(500)         0.033
heapsort(20000)    0.068
matrix(20)         0.072
nestedloop(12)     0.164
sieve(30)          0.075
strcat(200000)     0.010
------------------------
Total              2.034
real: 2.05s user: 2.03s sys: 0.01s cpu: 99% maxmem: 124320 KB cswaits: 7

simple             0.099
simplecall         0.111
simpleucall        0.146
simpleudcall       0.147
mandel             0.230
mandel2            0.276
ackermann(7)       0.114
ary(50000)         0.016
ary2(50000)        0.015
ary3(2000)         0.133
fibo(30)           0.308
hash1(50000)       0.027
hash2(500)         0.034
heapsort(20000)    0.069
matrix(20)         0.080
nestedloop(12)     0.164
sieve(30)          0.075
strcat(200000)     0.010
------------------------
Total              2.054
real: 2.07s user: 2.05s sys: 0.01s cpu: 99% maxmem: 124320 KB cswaits: 1

simple             0.099
simplecall         0.109
simpleucall        0.146
simpleudcall       0.147
mandel             0.230
mandel2            0.268
ackermann(7)       0.114
ary(50000)         0.016
ary2(50000)        0.015
ary3(2000)         0.133
fibo(30)           0.306
hash1(50000)       0.028
hash2(500)         0.033
heapsort(20000)    0.067
matrix(20)         0.072
nestedloop(12)     0.164
sieve(30)          0.075
strcat(200000)     0.010
------------------------
Total              2.032
real: 2.05s user: 2.03s sys: 0.01s cpu: 99% maxmem: 124320 KB cswaits: 1

bench.php results from 3 runs
2.034
2.054
2.032

bench.php avg: 2.0400
Avg: real: 2.06s user: 2.04s sys: 0.01s cpu: 99.00% maxmem: 124320.00KB cswaits: 3.00
created results log at /home/phpbench_logs/bench_200914-040455.log
server PHP info log at /home/phpbench_logs/bench_phpinfo_200914-040455.log
Code:
-------------------------------------------
Run PHP test Zend/micro_bench.php
-------------------------------------------

empty_loop         0.099
func()             0.280    0.182
undef_func()       0.309    0.210
int_func()         0.257    0.159
$x = self::$x      0.213    0.114
self::$x = 0       0.207    0.108
isset(self::$x)    0.184    0.085
empty(self::$x)    0.199    0.100
$x = Foo::$x       0.193    0.094
Foo::$x = 0        0.189    0.090
isset(Foo::$x)     0.232    0.134
empty(Foo::$x)     0.177    0.078
self::f()          0.318    0.220
Foo::f()           0.273    0.175
$x = $this->x      0.194    0.095
$this->x = 0       0.261    0.162
$this->x += 2      0.193    0.095
++$this->x         0.174    0.075
--$this->x         0.166    0.067
$this->x++         0.198    0.099
$this->x--         0.188    0.090
isset($this->x)    0.178    0.079
empty($this->x)    0.189    0.090
$this->f()         0.313    0.214
$x = Foo::TEST     0.179    0.080
new Foo()          0.609    0.510
$x = TEST          0.139    0.040
$x = $_GET         0.299    0.201
$x = $GLOBALS['v'] 0.483    0.384
$x = $hash['v']    0.342    0.244
$x = $str[0]       0.280    0.181
$x = $a ?: null    0.189    0.090
$x = $f ?: tmp     0.261    0.162
$x = $f ? $f : $a  0.209    0.111
$x = $f ? $f : tmp 0.265    0.166
------------------------
Total              8.438
real: 8.45s user: 8.44s sys: 0.00s cpu: 100% maxmem: 76176 KB cswaits: 1

empty_loop         0.099
func()             0.280    0.182
undef_func()       0.303    0.204
int_func()         0.257    0.158
$x = self::$x      0.212    0.114
self::$x = 0       0.201    0.103
isset(self::$x)    0.186    0.088
empty(self::$x)    0.199    0.101
$x = Foo::$x       0.192    0.093
Foo::$x = 0        0.179    0.080
isset(Foo::$x)     0.239    0.140
empty(Foo::$x)     0.177    0.078
self::f()          0.310    0.212
Foo::f()           0.273    0.175
$x = $this->x      0.195    0.096
$this->x = 0       0.252    0.153
$this->x += 2      0.199    0.101
++$this->x         0.177    0.079
--$this->x         0.164    0.066
$this->x++         0.198    0.099
$this->x--         0.189    0.090
isset($this->x)    0.178    0.080
empty($this->x)    0.203    0.105
$this->f()         0.310    0.212
$x = Foo::TEST     0.178    0.080
new Foo()          0.610    0.512
$x = TEST          0.139    0.041
$x = $_GET         0.300    0.201
$x = $GLOBALS['v'] 0.482    0.383
$x = $hash['v']    0.330    0.231
$x = $str[0]       0.281    0.183
$x = $a ?: null    0.186    0.087
$x = $f ?: tmp     0.256    0.157
$x = $f ? $f : $a  0.206    0.107
$x = $f ? $f : tmp 0.260    0.161
------------------------
Total              8.400
real: 8.41s user: 8.41s sys: 0.00s cpu: 100% maxmem: 76176 KB cswaits: 1

empty_loop         0.098
func()             0.281    0.182
undef_func()       0.303    0.204
int_func()         0.257    0.158
$x = self::$x      0.214    0.116
self::$x = 0       0.203    0.105
isset(self::$x)    0.183    0.085
empty(self::$x)    0.199    0.100
$x = Foo::$x       0.193    0.095
Foo::$x = 0        0.180    0.081
isset(Foo::$x)     0.234    0.135
empty(Foo::$x)     0.177    0.079
self::f()          0.311    0.213
Foo::f()           0.273    0.174
$x = $this->x      0.196    0.097
$this->x = 0       0.251    0.153
$this->x += 2      0.192    0.094
++$this->x         0.173    0.075
--$this->x         0.165    0.067
$this->x++         0.198    0.100
$this->x--         0.189    0.090
isset($this->x)    0.178    0.079
empty($this->x)    0.188    0.090
$this->f()         0.310    0.211
$x = Foo::TEST     0.178    0.080
new Foo()          0.608    0.509
$x = TEST          0.139    0.041
$x = $_GET         0.300    0.201
$x = $GLOBALS['v'] 0.493    0.395
$x = $hash['v']    0.328    0.229
$x = $str[0]       0.279    0.180
$x = $a ?: null    0.185    0.087
$x = $f ?: tmp     0.254    0.156
$x = $f ? $f : $a  0.216    0.118
$x = $f ? $f : tmp 0.259    0.160
------------------------
Total              8.385
real: 8.40s user: 8.40s sys: 0.00s cpu: 100% maxmem: 76192 KB cswaits: 1

micro_bench.php results from 3 runs
8.438
8.400
8.385

micro_bench.php avg: 8.4077
Avg: real: 8.42s user: 8.42s sys: 0.00s cpu: 100.00% maxmem: 76181.33KB cswaits: 1.00
created results log at /home/phpbench_logs/bench_micro_200914-040455.log
server PHP info log at /home/phpbench_logs/bench_phpinfo_200914-040455.log
 
Last edited:

eva2000

Well-known member
#15
Are you referring to the 16CPU, 48GB RAM, 1152GB SSD space for almost $500/month that you tested?

ROFL!! You can hardly compare the performance of what is really a beefy dedicated server costing $500/mo. to a basic 1 or 2 core VPS through them. That's just an extremely silly comparison to make. I highly doubt @Marcus was running that configuration, or else yeah...he probably would be getting great performance.
I was speaking towards the volatility comment as I've tested Linode 2GB to 48GB and cpus do scale with the right multi threaded utilising tests
 

Marcus

Well-known member
#16
@eva2000 your results are showing really dedicated cpu cores and a stable performance. It looks good. Unfortunately I didn't have that experience.

Here are the results from my live server.
Code:
# php -v
PHP 5.5.14 (cli) (built: Jul  9 2014 21:40:15)
Copyright (c) 1997-2014 The PHP Group
Zend Engine v2.5.0, Copyright (c) 1998-2014 Zend Technologies
    with Zend OPcache v7.0.4-dev, Copyright (c) 1999-2014, by Zend Technologies

# php Zend/bench.php
simple             0.116
simplecall         0.118
simpleucall        0.117
simpleudcall       0.121
mandel             0.314
mandel2            0.373
ackermann(7)       0.147
ary(50000)         0.021
ary2(50000)        0.018
ary3(2000)         0.151
fibo(30)           0.377
hash1(50000)       0.034
hash2(500)         0.031
heapsort(20000)    0.085
matrix(20)         0.092
nestedloop(12)     0.196
sieve(30)          0.095
strcat(200000)     0.011
------------------------
Total              2.418

# php Zend/bench.php
simple             0.116
simplecall         0.115
simpleucall        0.116
simpleudcall       0.120
mandel             0.280
mandel2            0.371
ackermann(7)       0.157
ary(50000)         0.021
ary2(50000)        0.018
ary3(2000)         0.151
fibo(30)           0.378
hash1(50000)       0.033
hash2(500)         0.030
heapsort(20000)    0.084
matrix(20)         0.089
nestedloop(12)     0.195
sieve(30)          0.094
strcat(200000)     0.011
------------------------
Total              2.378


# php Zend/bench.php
simple             0.116
simplecall         0.115
simpleucall        0.116
simpleudcall       0.120
mandel             0.280
mandel2            0.371
ackermann(7)       0.157
ary(50000)         0.021
ary2(50000)        0.018
ary3(2000)         0.151
fibo(30)           0.378
hash1(50000)       0.033
hash2(500)         0.030
heapsort(20000)    0.084
matrix(20)         0.089
nestedloop(12)     0.195
sieve(30)          0.094
strcat(200000)     0.011
------------------------
Total              2.378

# php Zend/micro_bench.php
empty_loop         0.120
func()             0.386    0.266
undef_func()       0.391    0.271
int_func()         0.324    0.205
$x = self::$x      0.365    0.245
self::$x = 0       0.252    0.133
isset(self::$x)    0.220    0.100
empty(self::$x)    0.236    0.116
$x = Foo::$x       0.331    0.211
Foo::$x = 0        0.228    0.109
isset(Foo::$x)     0.289    0.169
empty(Foo::$x)     0.207    0.087
self::f()          0.415    0.295
Foo::f()           0.385    0.265
$x = $this->x      0.359    0.239
$this->x = 0       0.280    0.160
$this->x += 2      0.238    0.119
++$this->x         0.309    0.190
--$this->x         0.305    0.185
$this->x++         0.321    0.202
$this->x--         0.232    0.112
isset($this->x)    0.218    0.098
empty($this->x)    0.231    0.111
$this->f()         0.432    0.313
$x = Foo::TEST     0.217    0.097
new Foo()          0.734    0.614
$x = TEST          0.167    0.047
$x = $_GET         0.363    0.243
$x = $GLOBALS['v'] 0.430    0.310
$x = $hash['v']    0.343    0.224
$x = $str[0]       0.338    0.219
$x = $a ?: null    0.338    0.218
$x = $f ?: tmp     0.305    0.185
$x = $f ? $f : $a  0.333    0.213
$x = $f ? $f : tmp 0.402    0.283

Warning: gettimeofday(): 
------------------------
Total             11.043

# php Zend/micro_bench.php
empty_loop         0.119
func()             0.382    0.263
undef_func()       0.386    0.267
int_func()         0.309    0.190
$x = self::$x      0.364    0.245
self::$x = 0       0.242    0.123
isset(self::$x)    0.221    0.102
empty(self::$x)    0.235    0.115
$x = Foo::$x       0.332    0.212
Foo::$x = 0        0.219    0.099
isset(Foo::$x)     0.288    0.169
empty(Foo::$x)     0.206    0.087
self::f()          0.411    0.292
Foo::f()           0.383    0.264
$x = $this->x      0.359    0.239
$this->x = 0       0.278    0.159
$this->x += 2      0.237    0.118
++$this->x         0.311    0.192
--$this->x         0.293    0.174
$this->x++         0.322    0.202
$this->x--         0.232    0.112
isset($this->x)    0.216    0.097
empty($this->x)    0.227    0.107
$this->f()         0.431    0.312
$x = Foo::TEST     0.218    0.098
new Foo()          0.734    0.614
$x = TEST          0.167    0.048
$x = $_GET         0.355    0.235
$x = $GLOBALS['v'] 0.432    0.313
$x = $hash['v']    0.347    0.227
$x = $str[0]       0.337    0.218
$x = $a ?: null    0.342    0.222
$x = $f ?: tmp     0.311    0.191
$x = $f ? $f : $a  0.339    0.220
$x = $f ? $f : tmp 0.400    0.281

Warning: gettimeofday(): 
------------------------
Total             10.987
There are some more topics about that, this is the first I found now:
https://forum.linode.com/viewtopic.php?t=10815&p=63029
 

eva2000

Well-known member
#17
those numbers look well within a margin of variance between runs to me which would be normal - especially if it isn't an idle server but a live production server with active usage :)

and that post is for older linode plans with older cpus the Xeon L5520 is old and much slower clock for clock than the new linode Xeon E5-2680v2 used nowadays and Amazon E2 now uses E5-2670v2 or E5-2680v2 too

FYI, the PHP Zend bench test is single threaded and depends on cpu clock speed as well.

  • Xeon L5520 is base clocked at 2.26Ghz with Turbo boost at 2.48Ghz
  • Xeon E5-2680v2 is base clocked at 2.8Ghz with Turbo boost at 3.6Ghz which is almost 50% faster at single threaded Turbo boost speed.
  • Not to mention Xeon E5-2680v2 is like at least 4 generations newer than Xeon L5520 and performance clock for clocker is multiples faster to add on to that higher clock speed
I posted my Linode 2GB staging server benchmarks at https://community.centminmod.com/th...-for-centmin-mod-lemp-servers.1298/#post-5956 as well

for PHP Zend tests they look like yours which is close to expected give that it's single threaded test and Linode all VPS now use Xeon E5-2680v2 anyway

I wouldn't use PHP Zend test as the sole factor to determine performance anyway :)

Code:
-------------------------------------------
Run PHP test Zend/bench.php
-------------------------------------------

simple             0.122
simplecall         0.118
simpleucall        0.116
simpleudcall       0.119
mandel             0.253
mandel2            0.348
ackermann(7)       0.128
ary(50000)         0.027
ary2(50000)        0.024
ary3(2000)         0.160
fibo(30)           0.421
hash1(50000)       0.037
hash2(500)         0.040
heapsort(20000)    0.089
matrix(20)         0.090
nestedloop(12)     0.199
sieve(30)          0.097
strcat(200000)     0.012
------------------------
Total              2.400
real: 2.44s user: 2.39s sys: 0.04s cpu: 99% maxmem: 161408 KB cswaits: 6

simple             0.120
simplecall         0.116
simpleucall        0.113
simpleudcall       0.120
mandel             0.251
mandel2            0.338
ackermann(7)       0.129
ary(50000)         0.027
ary2(50000)        0.026
ary3(2000)         0.159
fibo(30)           0.407
hash1(50000)       0.038
hash2(500)         0.039
heapsort(20000)    0.091
matrix(20)         0.091
nestedloop(12)     0.200
sieve(30)          0.098
strcat(200000)     0.012
------------------------
Total              2.377
real: 2.41s user: 2.35s sys: 0.05s cpu: 99% maxmem: 161504 KB cswaits: 2

simple             0.123
simplecall         0.117
simpleucall        0.114
simpleudcall       0.118
mandel             0.249
mandel2            0.352
ackermann(7)       0.131
ary(50000)         0.027
ary2(50000)        0.024
ary3(2000)         0.165
fibo(30)           0.413
hash1(50000)       0.038
hash2(500)         0.041
heapsort(20000)    0.089
matrix(20)         0.090
nestedloop(12)     0.199
sieve(30)          0.097
strcat(200000)     0.012
------------------------
Total              2.398
real: 2.43s user: 2.39s sys: 0.03s cpu: 99% maxmem: 159312 KB cswaits: 2

bench.php results from 3 runs
2.400
2.377
2.398

bench.php avg: 2.3917
Avg: real: 2.43s user: 2.38s sys: 0.04s cpu: 99.00% maxmem: 160741.33KB cswaits: 3.33
created results log at /home/phpbench_logs/bench_100914-050403.log
server PHP info log at /home/phpbench_logs/bench_phpinfo_100914-050403.log

-------------------------------------------
Run PHP test Zend/micro_bench.php
-------------------------------------------

empty_loop         0.118
func()             0.363    0.245
undef_func()       0.479    0.361
int_func()         0.302    0.185
$x = self::$x      0.259    0.141
self::$x = 0       0.323    0.206
isset(self::$x)    0.367    0.249
empty(self::$x)    0.368    0.251
$x = Foo::$x       0.249    0.131
Foo::$x = 0        0.243    0.126
isset(Foo::$x)     0.316    0.198
empty(Foo::$x)     0.342    0.225
self::f()          0.397    0.279
Foo::f()           0.358    0.241
$x = $this->x      0.227    0.109
$this->x = 0       0.314    0.196
$this->x += 2      0.250    0.132
++$this->x         0.225    0.107
--$this->x         0.214    0.097
$this->x++         0.256    0.138
$this->x--         0.243    0.126
isset($this->x)    0.345    0.227
empty($this->x)    0.357    0.240
$this->f()         0.426    0.309
$x = Foo::TEST     0.331    0.213
new Foo()          0.737    0.619
$x = TEST          0.301    0.184
$x = $_GET         0.246    0.129
$x = $GLOBALS['v'] 0.314    0.196
$x = $hash['v']    0.235    0.118
$x = $str[0]       0.345    0.228
$x = $a ?: null    0.218    0.101
$x = $f ?: tmp     0.306    0.188
$x = $f ? $f : $a  0.347    0.230
$x = $f ? $f : tmp 0.303    0.186
------------------------
Total             11.025
real: 11.06s user: 11.03s sys: 0.01s cpu: 99% maxmem: 111776 KB cswaits: 3

empty_loop         0.118
func()             0.363    0.246
undef_func()       0.493    0.375
int_func()         0.357    0.239
$x = self::$x      0.270    0.152
self::$x = 0       0.324    0.206
isset(self::$x)    0.365    0.247
empty(self::$x)    0.352    0.234
$x = Foo::$x       0.243    0.125
Foo::$x = 0        0.229    0.111
isset(Foo::$x)     0.316    0.198
empty(Foo::$x)     0.345    0.228
self::f()          0.393    0.275
Foo::f()           0.359    0.241
$x = $this->x      0.228    0.110
$this->x = 0       0.333    0.215
$this->x += 2      0.247    0.129
++$this->x         0.218    0.100
--$this->x         0.212    0.094
$this->x++         0.258    0.140
$this->x--         0.237    0.119
isset($this->x)    0.333    0.215
empty($this->x)    0.366    0.248
$this->f()         0.427    0.310
$x = Foo::TEST     0.345    0.227
new Foo()          0.746    0.628
$x = TEST          0.280    0.162
$x = $_GET         0.232    0.114
$x = $GLOBALS['v'] 0.303    0.185
$x = $hash['v']    0.250    0.132
$x = $str[0]       0.356    0.238
$x = $a ?: null    0.233    0.115
$x = $f ?: tmp     0.304    0.186
$x = $f ? $f : $a  0.345    0.227
$x = $f ? $f : tmp 0.321    0.203
------------------------
Total             11.100
real: 11.13s user: 11.10s sys: 0.02s cpu: 99% maxmem: 111760 KB cswaits: 2

empty_loop         0.117
func()             0.382    0.265
undef_func()       0.480    0.363
int_func()         0.303    0.186
$x = self::$x      0.270    0.153
self::$x = 0       0.323    0.206
isset(self::$x)    0.346    0.228
empty(self::$x)    0.357    0.240
$x = Foo::$x       0.224    0.107
Foo::$x = 0        0.218    0.101
isset(Foo::$x)     0.315    0.198
empty(Foo::$x)     0.344    0.227
self::f()          0.397    0.279
Foo::f()           0.364    0.246
$x = $this->x      0.228    0.111
$this->x = 0       0.322    0.205
$this->x += 2      0.250    0.133
++$this->x         0.228    0.111
--$this->x         0.215    0.097
$this->x++         0.266    0.149
$this->x--         0.240    0.123
isset($this->x)    0.333    0.216
empty($this->x)    0.359    0.242
$this->f()         0.406    0.289
$x = Foo::TEST     0.346    0.229
new Foo()          0.721    0.604
$x = TEST          0.279    0.162
$x = $_GET         0.246    0.129
$x = $GLOBALS['v'] 0.292    0.175
$x = $hash['v']    0.239    0.122
$x = $str[0]       0.331    0.214
$x = $a ?: null    0.217    0.100
$x = $f ?: tmp     0.304    0.187
$x = $f ? $f : $a  0.340    0.223
$x = $f ? $f : tmp 0.313    0.196
------------------------
Total             10.915
real: 10.96s user: 10.94s sys: 0.01s cpu: 99% maxmem: 111776 KB cswaits: 1

micro_bench.php results from 3 runs
11.025
11.100
10.915

micro_bench.php avg: 11.0133
Avg: real: 11.05s user: 11.02s sys: 0.01s cpu: 99.00% maxmem: 111770.67KB cswaits: 2.00
created results log at /home/phpbench_logs/bench_micro_100914-050403.log
server PHP info log at /home/phpbench_logs/bench_phpinfo_100914-050403.log
 
Last edited:

Marcus

Well-known member
#18
My cores are part from Intel Xeon E5-2670 v2. Most likely with counsetting: on demand and not with cpu setting: on performance and not with turbo boost enabled.

Some more data from my memory.
Load time of my modded xf forum start page within debug mode in:
Current setup: 0.18
Linode setup: 0.20-0.38 (mostly at 0.34-0.38)
Former 12 core oone year old server: 0.12

Which php do you use? I use php 5.5 from the standard repository. Your setup is faster. If the php releases are the same then the ..80 xeon must be much faster.
 
Last edited:

rainmotorsports

Well-known member
#19
Linode was dead slow when i tested it. Their cpu speed is VERY volatile. It seems you do not get a dedicated cpu core there.
Not sure how Linode is doing it but one of my other providers for example is selling threads not cores and only deal with Xeon's. So an 8 core 16 thread Xeon VPS host is serving up 16 single core vps's or 8 dual core so on and so forth. The particular provider I have experience with under demand they always perform. But with 2 threads I expect at least the performance of a single core in comparison to dedicated hardware (leaving single versus multithread loads out of the equation). With 1, if other clients are loading the server pretty hard I expect it to be crap.

But running window's servers on the one provider I always benchmark against my i5 at home.
 
Last edited:

eva2000

Well-known member
#20
My cores are part from Intel Xeon E5-2670 v2. Most likely with counsetting: on demand and not with cpu setting: on performance and not with turbo boost enabled.

Some more data from my memory.
Load time of my modded xf forum start page within debug mode in:
Current setup: 0.18
Linode setup: 0.20-0.38 (mostly at 0.34-0.38)
Former 12 core oone year old server: 0.12

Which php do you use? I use php 5.5 from the standard repository. Your setup is faster. If the php releases are the same then the ..80 xeon must be much faster.
currently using PHP 5.6.0 but old numbers with PHP 5.5.16 at https://community.centminmod.com/threads/community-centminmod-com-journey-for-speed.3/#post-5256 (PHP-FPM via my Centmin Mod LEMP web stack - you can see my source compile options I use at http://centminmod.com/phpfpm.html)