All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
VPS Benchmarks - What is considered acceptable?
I have several VPSs, and would like to consolidate them. It is very difficult because I feel like they are cheap and am getting a good deal if I ever want to use them. However they mostly sit idle, so my $7 here and $6 there every month adds up. And I want to keep the best service for what I want to do. Right now I want to learn about Linux, run a VPN, a WWW server, and a MySQL server to archive some historical weather data. Stability is also important. I hate it when some of the VPSs randomly reboot or even just shutting them down without a warning.
I ran a 'dd' benchmark command several times at random periods over a few days this past week and averaged the times and disk speed. Surprisingly my cheapest VPS was the fastest, and my most expensive was the slowest.
These benchmarks alone tell me that I should cancel the bottom two, but I also like the huge amount of storage and memory that are provided if I ever need them. My question is, if I ever run an application or set of services that consume a large amount of memory (1.5-2 GB) could this be a violation of some providers' acceptable use?
I am very new to the world of VPSs. And LowEndBox.com is like my Slickdeals.net of VPS. I see things that are good deals and buy them, even if I never know what to do with them!
Thank you for the advice.
dd if=/dev/zero of=~/2gb.dat bs=8k count=256k conv=fdatasync
Provider (Memory / Storage / Monthly Cost)
SemoWeb (256MB / 20GB / $1): 2147483648 bytes (2.1 GB) copied, 3.74051 s, 574 MB/s
RamNode (512MB / 50GB / $3): 2147483648 bytes (2.1 GB) copied, 4.89003 s, 439 MB/s
ChicagoVPS (3GB / 100GB / $6): 2147483648 bytes (2.1 GB) copied, 10.2702 s, 209 MB/s
EO Reality (2GB / 30GB / $7): 2147483648 bytes (2.1 GB) copied, 18.7394 s, 115 MB/s
Iniz (Storm) (3GB / 100GB / $7): 2147483648 bytes (2.1 GB) copied, 32.5445 s, 66.0 MB/s
Comments
So you're motivated by numbers and when stock is filled, your numbers don't look good and you're off to find greener pastures with higher numbers to ultimately repeat everything over again?
That is far from my intent. When I first discovered LowEndBox.com I was shocked at the great deals and purchased several different VPSs that I felt I might have a use for in the near future. I have actually held on to all of these accounts for over a year, some of them for 18 months. No plans to acquire any more. I realize there are costs associated to the providers for setup and decommissioning of accounts, so I do not believe in creating accounts for just a couple of months.
I would simply like to consolidate down to maybe 2 services, and I created this thread in hopes to ask what are some of the best benchmarks to use to gauge one provider against another.
I mentioned reliability, memory, storage, and throughout. Support is also important. I wasn't too pleased with the way ChicagoVPS handled its downtime a couple of months ago, however they transferred my account to a faster node after I noticed it was taking 20-30 minutes to do a simple 'apt-get upgrade'
If you think you may need 2GB VPS, you can always buy it WHEN you need it. Pointless having an idle server doing nothing. Decent deals come around every 1-2 months and generally the prices decrease over time so you will be able to get the same service cheaper in the future so no use idleing. Also it is greener for the planet to reduce idleing servers and good for IPv4 if we free some up.
I'm happy to learn more about correct benchmarking. I admit I am very new to this! No need for haterz
It will be your mission like most of the big number / cheap price fanboys on here after awhile as it's not your intent but subconsciously, this is how they work
Perhaps you could start from another direction and tell what are you using your VPSes for.
There are many types of benchmarks (UnixBench and GeekBench, general capacity, including calculating capacity; miscellany of disk I/O benchamrking - fio, iozone, ioping etc; task-specific tests - Phoronix suite, it comprises quite a number of typical usage, such as computational tasks, I/O-oriented tasks and so on).
Start from stating your goal. Benchmarks by themselves are nothing but numbers well suitable for you-know-what contests.
Example: I have a very modest (Micro) Amazon EC2 instance which serves me for years as reliable host for my Subversion repositories as well as backup controller.
There's a good rule of thumb - don't put all your eggs in one basket.
Thanks for the replies everyone.
Are there any good guidelines for how many resources a MySQL server should have? RAM vs disk I/O? I want to use a VPS to serve as a backup database for weather data. Right now it's about 1.4 million rows of data (350MB of data), and growing one row per minute. In the future I plan on creating a website to access this historical data, but expect no more than a few visits per day (probably me).
Is it better to have a massive multi-GB memory VPS with slow disk I/O for a small website with database, or would a 512 MB VPS suffice?
@C_Adam
Try ioping for (big) MySQL databases. And yes, you probably want some good amount of RAM for MySQL, not because it needs it, but because it can do magic if you feed it enough RAM to cache with :-)
Yes, any db server will stress the disk and CPU much less if can cache the most queries and tables, and will also be more responsive. Add a big raid (many disks in raid 10) SSD storage and you have the perfect DB server.
Run UnixBench, also check out serverbear.com , will give you a good idea of benchmarks from different providers.
66mb io is acceptable for me
MySQL can be tuned to virtually any amount of RAM; however, faster I/O and much RAM are always good.
Note that the actual task governs how much resources should be available for acceptable performance. Well-known scripts like tuning-primer.sh can be used to adjust configuration.
"It's better to be rich and healthy, than poor and ill". Both RAM amount and I/O speed donate to DB service performance.
oh damn, my raid-10 is so slow =(
2147483648 bytes (2.1 GB) copied, 77.1428 s, 27.8 MB/s