donavanm

69 ms·

721.
by donavanm 14y ago
Serious? A huge amount of traffic sent over the Internet is multipath due to ECMP. Hashing on the full src/dst ip and port tuple makes for stable path selection. WRT to the second point, failure cases are not that simple. Edit: lets talk di
722.
by donavanm 14y ago
http://lmgtfy.com/?q=ecmp Secondly, no your bgp session won't automagically go do just because httpd dies. There's a whole mess of complexity hidden behind that assumption.
723.
by donavanm 14y ago
Interesting. The /30 ptp interface makes me suspect this is built on top of vpc. The other end of the connection would be a dedicated nat "router" instance. Efficiency would suck in that case. They could use a larger shared vpc (say /24). c
724.
by donavanm 14y ago
"OS X isn’t designed to behave differently while on different types of networks" Funny. This was exactly one of the justifications for writing launchd. "daemons" would be signaled on state change, like a network state transition. This behav
725.
by donavanm 14y ago
Sure. You can absolutely replicate chunks. But you start kicking the problem upstream. A rack down is a couple pb, so you start doing a ton of cross rack transfers to get your replica counts back up. Now you're gated on nic/TOR/agg switch t
726.
by donavanm 14y ago
Amazon had about 60 billion in revenue last year. 1/3 of 150M inst so much. Jeffy B wrote a personal check for ~30M to get in on über, for comparison. Edit: totally forgot, amazon already wrote most (all?) of the living social investment la
727.
by donavanm 14y ago
No. As you said port speed doesn't matter for data at rest. What matters is ingest/exfil of data due to "exceptional" conditions. Prime cases are cluster/mirror failure. Remirroring existing data to another pod is port limited, as is ingest
728.
by donavanm 14y ago
The CPU can change power states in 10s of microseconds. If your core is constantly bouncing between C states expect it to add a couple hundred microseconds of latency. And if you have a modern linux kernel try disabling the intel_idle drive
729.
by donavanm 14y ago
Yes, all things are possible. You'll also have to keep state and track rates. Otherwise a very low error rate could down all of your backend hosts. But you're now running a stateful l7 application proxy. That's waaaaaay more expensive than
730.
by donavanm 14y ago
No, lbs don't inspect established streams. The lb will periodically send requests to a known uri as a health heck instead. The problem is when the health check uri isn't indicative of availabilty. (Hint: it never is) Nope, don't do this eit
731.
by donavanm 14y ago
Or use a parameter that's effectively a nonce.
732.
by donavanm 14y ago
> I've never understood using round-robin or random algorithms for load-balancers... LeastConns/FastestConn selection is very dangerous when a backend host fails. Imagine a host has a partial failure, allowing health checks to pass. Thi
733.
by donavanm 14y ago
No template DSL/system in the console. But there is an API with quite a few client libraries. Shouldn't be much work to whip one out. Or it may exist already, actually.
734.
by donavanm 14y ago
Azure, in general, is just burning money with no significant uptake. Friends who work there say there aren't even significant internal uses. Just based on that I'd be dubious of any new Azure service dominating their segment. Maybe they'll
735.
by donavanm 14y ago
Nah, porn companies are generally too cheap to use services like aws.
736.
by donavanm 14y ago
Nope, video only. Audio is a logical step though. Go bug your friendly neighborhood rep/support person with a feature request.
737.
by donavanm 14y ago
Irony.
738.
by donavanm 14y ago
Approximately 1 hojillion[1] dollars? But I think we both know that's not the real question here. Not to dissect the frog too much, but the joke is that XML == Enterprise === Grown Up. Therefore more XML is mo betta. [1] www.penny-arcade.c
739.
by donavanm 14y ago
You guys do understand this is a joke, right? It's a (cool) example of stupid things that you can get up to in ruby. See the ruby con talk for other examples. Enterprise Ruby is my personal favorite.
740.
by donavanm 14y ago
Cloudfront can vary objects by using a different uri stem, query params, or cookie contents. As noted in the article you can also route uri prefixs to your image sizer origin. So yes, I'd expect most image manipulation to be cachable using
741.
by donavanm 14y ago
Nope. Route 53 has to ALIAS to another RR in your own hosted zone. There's no way for Route 53 to return the A records/IP address RDATA that CloudFront uses to direct clients to the fastest site. Other "ALIAS" providers can't do real CloudF
742.
by donavanm 14y ago
Sad face indeed. I wish route 53 had generic "alias" support like other providers. IIRC Route 53 can ALIAS to 1) your own hosted zone 2) an ELB instance 3) an S3 website bucket.
743.
by donavanm 14y ago
Actually, yeah. How about worldwide nxos crashes due to the leap second bug? Or the various poison bgp updates that've made the rounds? Or overrunning an ospf domain? Anyways my point, if you read the sentance after my quote, was there's a
744.
by donavanm 14y ago
Totally. I know I create lots of business value when I spend a day dicking around with mysqldump and rsync and inotify and scp and hfds. Who would want to use this kind janitorial service when the could do it themselves?
745.
by donavanm 14y ago
"A website/service cannot by definition be HA if it's reliant on one service or infrastructure provider." you seem to be conflating highly available with a diverse supply chain. A lot of highly available systems are "locked" in to one pro
746.
by donavanm 14y ago
This discussion is all about the agent that sits on the clients premises, correct? First define your server side RESTy JSON API. Then write poll based client agent, extract data, transform, PUT to API. Don't bother with embedded http server
747.
by donavanm 14y ago
Can you elaborate on CloudFronts "poor cache hit ratio"? Wouldn't this just be up to the origins use of cache headers (cache-control, expires, etc)? I suppose cold content could be LRUd out, but the CloudFront docs advertise advertise 1 day
748.
by donavanm 14y ago
Talk to legal representation with experience in employment law. But yes, you could very possibly get a (fractional) payout. Could be a cash "bonus" or salary for 6 months of an "advisory" position. Be prepared for real talk about non compet
749.
by donavanm 14y ago
Explain? ~600k tps on 1gb nics is very achievable. 10-15M tps on 10gb interfaces is also reasonable. In both cases I'm think of udp packets in/out measured on the wire. With modern hardware I'd buy similar numbers for tcp.
750.
by donavanm 14y ago
This is t a systems failure, it's a process failure. Why was the data unavailable? Hardware failure Why wasn't the most recent version available? The operator didn't manually do a "full" backup Why did the operator need to manually start a

More ›