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.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Who is doing A + B ?
With the few recent event at psychz LA, I'm wondering who is currently running A + B ?
Personally interested in for a few provider : @AnthonySmith @Francisco @VirMach
Edit : This thread is not about what is hosted at psychz, but more on general purpose.
Comments
We're a+b, I believe @FR_Michael is as well from previous discussions.
I would highly recommend going a+b. Most of the time especially with rented dedis you'll get offered a single feed from a ups or pdu with redundant feeds, but what happens when the ups/pdu fails? Power outages are never fun and a little investment solves that.
Who isn't doing A+B is more like it!
While this might not be done for single smaller servers itself, almost every system we buy that have any decent cost have A+B it's just not worth the downtime.
Yep. Agreed with @PureVoltage who doesn’t do is the better question
By reading psychz thread, I saw few that got affected, so I just wondered how much lowend market is affected by that.
More than you'd think are single feed. It's really quite surprising.
correct, we are doing a+b as well.
Our cheaper, single CPU dedicated servers are on a single PDU, which connects to a circuit with it's own upstream redundancies in place, but more expensive dedicated servers, dual procs, and everything else in our infrastructure runs dual feeds from separate PDUs on separate circuits.
Edit:
I made my post a little more clear.
@IonSwitch is a+b except for some dedicated machines with single pdu’s.
How much redundancy does dual feed really give you? I look at cheap colo offers (single server) and it hadn't really occurred to me to pay for dual feed, but maybe there is something to it.
Well if a circuit goes out or a pdu goes out, or even a power supply in the server, you're covered.
We are fully A + B
actually we have couple of older switches that are single feed and we connect them to ATS that is connected to both A + B , it works perfectly we have tested it many times and we almost dont care about power outage any more
Guessing most of the low end market is single feed. Costs have to be cut for the deals some people hold out for. Nobody is going to comment and say either "all/most of our stuff is single feed" or "derp never thought to ask our cheapo dedi provider".
Also FWIW A+B can mean different things in different facilities. In a good facility, it should mean you can connect 2 PDUs in any rack to two separate circuit boards, feed from separate UPSs and baked up by separate generators. In cheap facilities, it might just mean separate circuit boards.
I'm curious to know how much the A + B power setup might add to the costs (just a general range, I know "it depends").
I can imagine in the long run it may "cost" much more to experience power outages - but I'd like to get a better sense of what's involved in the up-front part of that equation.
I know only x + y = ...
7
I just went back to the old price quotes I got few years ago and the A + B was 60% increase in power and considering the rack rental cost is negligible compared to power I would say A+B is maybe 50% to 60% more expensive than single feed
Having a list of providers doing a+b with ECC ram would be nice, since these would be more suitable for non-hobby production
so ... if that increased cost is passed along to the end-user ... then what would otherwise be a $7 service becomes more like $11 with A + B power? (I'm guessing my math may be a bit off somewhere, and is likely to be based on some incorrect and incomplete assumptions.)
Just to share my perspective from a cheapskate customer's point of view - yes, I can appreciate that professional-grade power redundancy and in some cases I may actually be inclined to pay a bit extra for it.
However ... for the bulk of my personal "DIY High Availability" seat-of-the-pants designs I tend to prefer having an entirely separate service (different provider in a different DC, possibly also even a far distant location altogether) ... so if I'm looking at 2 x $7 to do that vs. $11 to bulk up what is still going to be a single point of failure (as far as network or provider fail goes) ... well, in many scenarios I may be more comfortable with a "redundant array of inexpensive servers" rather than paying almost as much just for a single server with A + B power.
But my calculations might be entirely different If I were a provider supporting many clients on a single server.
Still - as a cheapskate customer, I'm happy to scoop up those "how do they do it" ultra-cheap storage deals such as the occasional HostHatch promotion (apparently without A + B power in Psychz LA) - and if it goes down a few times a year, well ... that's probably still going to be working well enough for me (until it doesn't).
¯\_(ツ)_/¯
A+B Power with UKServers
Take it from a man/potato whose name is uptime to know a little something something about uptime
mmmm yeah okay I can tell you this much at least - as much as I like to stay on top of things ... downtime is a fact of life - plan for it, deal with it.
I mean, I try to reboot my own damn meatbag self at least once every 36 hours lest my process tables get too irredeemably corrupted
all that said, power outage is a grim way for a box to go down so I dunno ... you pays your money and you takes your chances. Or not.
tl;dr: the finer points of A + B power seem to be just a bit above my pay grade ... but I'll take what I can get for $7
a+b has an higher initial cost (for hardware) and it uses a bit more power (I cannot confirm 50-60%, it's more like 5-10% for us) but you save a lot of headache, man hours, bad reputation once you lose power an a rack full of single feeded nodes .
We went to full a+b on infrastructure and nodes a while ago and we never looked back.
It can happen that a dying psu takes down the power feed it is connected to and it is a good feeling just bringing back the lost feed and go back to work instead of booting all nodes, answering to dozen tickets, handling fsck, talking to customers who loses one hundret million dollar cause of this downtime and so on.
I thought this was some driver license discussion or something related. Damn
I genuinely thought he was asking A/B tests...
To be clear, we have A+B already in almost all of our locations except LA. We've never had a power related outage in the past 8 years of doing business, other than the past few weeks with Psychz in LA. We've been protected from several power outages due to A+B in the past. We will also have A+B in all of the 5 new locations coming this year.
We're also changing to A+B in LA while adding network redundancies as well. This will happen on the 2nd of August. We planned this out as soon as the first power outage happened, but the reason it took long was because we wanted to be in LA ourselves for this, instead of putting all of our trust into remote hands.
The reason we do not have A+B in LA is because we had to move in LA with a 9 days notice, when Colo@/TotalServerSolutions (our previous upstream) decided to migrate their whole datacenter presence in LA with a 9 day notice, without apology or much detail, and their business dev manager decided that they wanted to (quite close to scamming -- extort money) and blackmail us to the last extent to try to get us to stay with them, breaking their own word, for a client that paid nearly 6 figures and was loyal for 5 years, but that's a story for another day (with screenshots and all that)
Psychz was very accommodating as they moved our deployment date from a couple months to about a week while they were running close to full on space in their DC. Things do go wrong from time to time - but I have my confidence in Psychz for many reasons.
Indeed - perhaps something for networks too, you might be surprised how many are running single-homed Cogent/HE
Maybe.
ax + b = 0
... = $7
...