2,215 karma · joined October 29, 2008
C/C++, baremetal/cloud, nginx/openresty/nchan, LuaJIT, PHP, Redis, statsd/rrd
Cycling is awesome. Let's fix how they coexist on roads.
It's hard to imagine FB's stranglehold letting up. Some startup, prove me wrong.
They are... they are literally the cloud(s) and their biggest customer. In fact, they are big enough to offer you their ops and expertise for a small fee over baremetal. Many cloud consumers are literally using up the overhead existing in these businesses' hardware that would otherwise be sitting around.
It can be done cheaper, sure but not without giving up opportunity time and work you may just not want to deal with. I certainly am over having to manage supplier relationships or having to wake someone I don't know or find difficult to converse with to bounce a router or kick a generator.
The issue is embargoed, but that's, as usual, not keeping everyone from speculating. That's fine too, but also realize the layman (even people in this thread) is getting pummeled with ideas like "my new laptop is going to be 30% slower tomorrow; WTF!!!"
Workload dependence... average user... not significant:
A lot of users probably see "30%" and will assume the worst. There really isn't a lot of evidence of how much an average consumer will be affected. Pretty sure your average consumer isn't running pgbench all day. Hell, I'm willing to bargain 90% of servers (I really want to say more) are over-provisioned by 30% or more.
Mitigated over time:
Kernels are getting out safe patches ASAP, performance be damned. With more time to be careful, performance can be clawed back. And/or CPUs are upgrade and features such as PCID mitigate perf losses.
You will find many clients asking how to disable, for example, the Linux patch. Linux is releasing with a flag to disable it, so there is some merit. Why would you want that? There are a lot of times you trust everything running on your box and don't need to take the perf hit.
Intel (and possibly other archs/families) found a perf win that ends up having security implications. A perf nonetheless. If you're willing to bet on your userspace not reading kernel pages, does Intel have a feature for you!
This is wrong. Most cloud providers, AWS included, will _require_ a heads up for even small surges. There are also caps on capacity; in fact, you need to contact them to lift said caps.
It's solving the market of oversaturated shared-VPS that run Node, Python, or Go? Call when you have PHP.
I'm losing respect for Apples mobile hardware line. So what if their SoC is amazing, if it is downclocked every month. Review benchmarks are dishonest if not taking this into account.
The IEEE has been making a few small steps to ease this problem, but they’re hyper-timid incrementalist bullshit. There’s an IEEE program called GET that allows researchers to access certain standards (including 802.11) for free, but only after they’ve been public for six months — coincidentally, about the same time it takes for vendors to bake them irrevocably into their hardware and software.
In other words, it's getting better, but vendors are burning silicon by the time researchers have the time to even crack open whitepapers -- too late.
We are running game servers on WSL, where you can imagine we are using fairly to-the-metal socket work over UDP and no issues with libcurl or libopenssl for TCP sockets.
If I put a competitor (eg. yelp, angie[slist], etc) into my search query, I get only results for those competitors. Leave the qualifying term out, and I get a SERP of all non-Google review sites. But again, I enter "yelp pizza" and get all Yelp results. Hell, I enter "pizza reviews" and get Yelp as the top 3 links of my SERP.
The bigger win, for me, is the AZ flex, as it is a pain to manage multi-AZ setups.