Answer the question
In order to leave comments, you need to log in
What determines the performance of MYSQL?
Hello! I decided to test the VPS servers that I currently have. That is VPSie, Digital Ocean and micro on EC2. I installed Sysbench, MariaDB 10.1 on them, they all have the same OS (Ubuntu 15.04), all the parameters are about the same (1 core, 1GB RAM, SSD drive). I won’t post irrelevant information (like download speed from different servers), I’ll write only a part. sysbench --test=cpu --cpu-max-prime=20000 run
DO did this test in 38 seconds, VPSie in 27, and EC2 in 30.
With this command
dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync && rm -f test
DO had 15.6 MB/s (very strange), VPSie had 504 MB/s, EC2 had 64.3 MB/s. sysbench --test=oltp --oltp-table-size=1000000 --mysql-db=test --mysql-user=root --mysql-password=1234 prepare
sysbench --test=oltp --oltp-table-size=1000000 --mysql-db=test --mysql-user=root --mysql-password=1234 --max-time=60 --oltp-read-only=on --max-requests=0 --num-threads=8 run
sysbench --test=oltp --mysql-db=test --mysql-user=root --mysql-password=1234 cleanup
OLTP test statistics:
queries performed:
read: 310884
write: 0
other: 44412
total: 355296
transactions: 22206 (370.04 per sec.)
deadlocks: 0 (0.00 per sec.)
read/write requests: 310884 (5180.56 per sec.)
other operations: 44412 (740.08 per sec.)
Test execution summary:
total time: 60.0098s
total number of events: 22206
total time taken by event execution: 479.9248
per-request statistics:
min: 4.85ms
avg: 21.61ms
max: 106.37ms
approx. 95 percentile: 33.49ms
Threads fairness:
events (avg/stddev): 2775.7500/12.39
execution time (avg/stddev): 59.9906/0.00
OLTP test statistics:
queries performed:
read: 349062
write: 0
other: 49866
total: 398928
transactions: 24933 (415.49 per sec.)
deadlocks: 0 (0.00 per sec.)
read/write requests: 349062 (5816.85 per sec.)
other operations: 49866 (830.98 per sec.)
Test execution summary:
total time: 60.0088s
total number of events: 24933
total time taken by event execution: 479.8953
per-request statistics:
min: 5.33ms
avg: 19.25ms
max: 114.93ms
approx. 95 percentile: 32.82ms
Threads fairness:
events (avg/stddev): 3116.6250/8.69
execution time (avg/stddev): 59.9869/0.00
OLTP test statistics:
queries performed:
read: 682780
write: 0
other: 97540
total: 780320
transactions: 48770 (812.76 per sec.)
deadlocks: 0 (0.00 per sec.)
read/write requests: 682780 (11378.62 per sec.)
other operations: 97540 (1625.52 per sec.)
Test execution summary:
total time: 60.0055s
total number of events: 48770
total time taken by event execution: 479.8626
per-request statistics:
min: 2.41ms
avg: 9.84ms
max: 77.80ms
approx. 95 percentile: 12.70ms
Threads fairness:
events (avg/stddev): 6096.2500/13.25
execution time (avg/stddev): 59.9828/0.00
Answer the question
In order to leave comments, you need to log in
I don’t know how it is with DO and VPSie, but I can speak for EC2. A micro instance is an instance with no guaranteed performance. It can give you high performance for a short time, and then take away your CPU time. This is how the entire t2 instance series behaves:
> T2 instances are Burstable Performance Instances ...
https://aws.amazon.com/ec2/instance-types/
> A CPU Credit provides the performance of a full CPU core for one minute.
Look at the "CPU Credits / hour"
setting of the micro instance to 6. Swipe more tests within an hour immediately after the start of the instance - you should see a performance drawdown. credits are accumulated until the end of the day.
Or run the test for more than one minute, since the duration of one credit is just 1 minute:
on the same link.
Offtopic: I always considered DO to be the fastest cloud platform... your tests discouraged me a little. Probably time to check it out yourself.
According to the database itself - the speed can very much depend on the settings, up to 1000% performance.
Provided that these same settings are different - the question is for you - did they differ on your instances?
Didn't find what you were looking for?
Ask your questionAsk a Question
731 491 924 answers to any question