New on LowEndTalk? Please Register and read our Community Rules.
VPS Reliability/Availability vs Performance?
edited February 2013 in General
For a VPS, if you had to pick between the two - what would you prefer? (assuming you can't pick both that fits into decent LEB pricing)
Would you go for:
higher uptime, say via each server having A+B power/network redundancy at the server level and other measures, including higher uptime SLA for the DC
faster network and/or server performance
This got me thinking seeing all the complaints over planned maint / outages, downtime, etc etc... Take this as a bit of market research for new products / locations etc and also just a general discussion
High uptime = stable number of customers
Great performance = fast increasing number of customers
Low uptime = unstable number of customers
Poor performance = slow decreasing number of customers
Might be inaccurate.
I'd prefer more uptime/reliability over greater performance
Depends entirely on what I'm using it for. If I'm setting up a monitoring node, obviously I want high uptime. If I'm doing batch image processing or something else like that, or if it's a node in a round robin cluster, then I want performance.
I've just had some thoughts... Is there an easy comparison of hosts by uptime?
I know ServerBear tracks some hosts nodes, but there is no way to compare them
I don't care about availability/performance once my vps gets past certain thresholds. If it can finish a task within a time constraint, doing it quicker might not mean much to me. If my existing vps has reasonable uptime, and I want my service to have even higher availability, I will get another cheap vps, which might not be more reliable than the first, to take up the service when the first is down. I will not get an expensive but more reliable vps to replace my existing vps.
for me for LEB type budget, i'd want performance first as I tend to load balance my LEB VPSes across multiple web hosts so I take care of availability side of things
i.e. During buyvm LV move, my site on buyvm only had 2-4 min intervals of downtime, longest being around 15mins as my secondary load balancer was on buyvm buffalo VPS (won't happen anymore as have tertiary failover load balancer on a different LEB VPS now.
But my LEB aren't for anything critical, so reliability isn't a high priority
Reliability over Performance. Because if you have a server that has an amazing CPU, amazing SSD disk, but goes down every few days then that won't exactly be useful would it?
I do not need amazing performance, but I need reliable performance. Somethings like a guaranteed Disk IO, Network IO and CPU load.
It is more than "do not oversell" it is all about finding the right mix of users per node.
I do not care much about uptime guarantee. The node can be restarted once a week if it is not down when I need it. Everything is ok as long as it is has a schedule.
I just don't want following things:
+1 @GIANT_CRAB +1 @Ishaq
A super fast server is useless if it is always down.
Reliability/Availability over performance, as long as a basic level of performance can be met.
That is what the most ppl will want.
My needs are different as I dont need anything critical 100% up on LEB. I dont make money with VPSes so it doesnt matter if it crashes once a week for a few minutes as long as data loss does not occur. If it crashes every day for tens of minutes, then there is something more serious going on and will have to switch unless the traffic is great and big in which case I will keep for tor if allowed.
Uptime tends to matter when there is an outage. Rest of the time, uptime is fine and no one really care (yeah obvious).
Performance is something that can make every day a ruined one and can contribute to downtime.
There is a balance.
Start with an offering best one can. Then (due to cost) add on higher performance pieces and finally the redundancy (A+B power being the biggest cost and needed one).
I'd never prioritize uptime over performance or visa versa.
Indeed, this is what we sell for... we found there was less churn when selling to achieve a known baseline for performance, than to sell shiny numbers.
With all of my VPSs the biggest issue is usually network performance. The issue isn't that the server has crashed or rebooted, but rather there's some network outage.
Definitely reliability / uptime over performance.
Baseline performance is always the goal, but this is more of, e.g. faster RAID card/cpu vs more redundancy.
I'm not just talking about A+B power here, but also network. Networks do go down - e.g. when the router/switches - which happened recently here on LET.
Looking at the costs, an extra 1U switch + B power feed + dual PSU on all servers isn't going to be too much to price things out of budget.
From the results of the replies here, it looks like it might be something that people want. Can't provide 100% uptime, but I think we can go with a minimum of 99.99% uptime SLA this way. We'll see.
That is the lack of an optimizing network with different network providers and enough carriers in the mix to deal with it. This is the importance of being multi-homed. Amazed at how many are in the mix in our Texas location too - haven't seen it go down or slow down.
Thanks for all the answers. Maybe other providers that are reading and/or have replied can make use of this.
I would definitely pick higher uptime.. hate it when I see pingdom send me those alerts..
The uptime issue is why folks went cloud crazy.
Performance only matters to me in so far as what I am allocated. Nice to have beefy machines but if CPU share is tiny and IOPs severely limited then not really relative to me whatsoever.
Uptime often is indicative of cheapness of provider or bad management practices.
I'd say you sell people lousy performance on CPU all the time. Disk, only a bit of the time. Bandwidth, never.
But uptime, selling services that offline due to power loss or network failure, can't ever do that. People notice that real quick.
I think that there we have a big problem when most of the comment I see are price based or quantity based. I think that a good HA virtual machine can be sold at leb prices, when I approched this site I only had that (ha vm) and tried but then realized that people was comparing my 256mb of ram with others single node deploy without considering my n+1 setup, dual powered servers, double switches, etc. Some were complaining about the san speed without considering double hba, fc switches, double controllers, etc. I taken this as a personal offense so i started the single node deployment most here know. I had to do some cut here and there and had to force myself to go ahead :-)
In the next month(s) I hope to be here to discuss about a high quality 512mb vps rather than a 8gb leb vps
It's hard @prometeus.
The redundancy like SAN often means lower end user high end numbers in this model.
People are buying here for bang for the buck or euro.
Me, I like to see good IOPs, throughput, etc. However not as fascinated as most are with things. I care about latency and being able to move files over bandwidth as needed at a high speed.
Can't wait to see what you cook up for a 512MB VPS on the high quality side. Happy with the LEB I have with you folks now though