aphyr

25 ms·

241.
Jepsen: MongoDB 3.6.4(jepsen.io)
231 points by aphyr 8y ago|70 comments
242.
by aphyr 8y ago
Links to the code are throughout the "test design" sections of each report. :)
243.
by aphyr 8y ago
Aw, thank you! :)
244.
by aphyr 8y ago
it should not be an infeasible goal There's been a lot of work on both of these problems, but right now, proving concurrent algorithms correct, and proving equivalency of those algorithms to executable code, are very much open resear
245.
by aphyr 8y ago
Yeah, we did talk about Badger a bit, but I think Dgraph's internal testing was likely going to stress it more than I could over the network. We have occasionally found storage-engine level bugs with Jepsen (Tendermint, for instance),
246.
by aphyr 8y ago
Oh, yeah, that's something I could put together. In the meantime, there's a mailing list and twitter you can subscribe to! https://jepsen.io
247.
by aphyr 8y ago
Oh, I certainly took no shortage of flack for those early analyses, but I'm glad you appreciated them! Some of my clients have even expressed disappointment that my jepsen.io analyses for them weren't as snarky as the aphyr.com on
248.
Jepsen: Dgraph 1.0.2(jepsen.io)
183 points by aphyr 8y ago|124 comments
249.
by aphyr 9y ago
My understanding is that the folks who are responsible for this name were also former aerospace engineers. Name overloading is always tricky, and I hope you can forgive their choices. Best regards, another grumpy software & physics pers
250.
by aphyr 9y ago
> Jepsen thinks transactions can take minutes to complete, and it's ok Jepsen takes great pains to make no assumptions about the time it takes for transactions to execute. We test systems with millisecond-level timeouts, or multi-mi
251.
by aphyr 9y ago
The Redis test was quite early in Jepsen's history, and I totally rewrote the library after that--you can still find the original code in the `old` branch, if you'd like to experiment. The Mongo tests are reasonably current; you&#
252.
Jepsen: Aerospike 3.99.0.3(jepsen.io)
134 points by aphyr 9y ago|43 comments
253.
by aphyr 9y ago
You're welcome, and yes, I do design consulting as well: https://jepsen.io/services .
254.
by aphyr 9y ago
I assume that if you lose your ticket you can always reauthenticate to get a new one, so... that doesn't seem like it'd be too bad, right?
255.
by aphyr 9y ago
Oh yeah, I guess this is a startup website isn't it! And Jepsen is... sort of... a startup? For context, Jepsen started as a series of volunteer nights-and-weekends blog posts and conference talks. About three years in I bootstrapped t
256.
by aphyr 9y ago
Thank you. Unfortunately, it's just not a good fit for a MOOC, in terms of class format and cost.
257.
by aphyr 9y ago
Hahaha I suppose that's an entirely legitimate question, isn't it? Just to reassure you, Lisa, yes, this is me, and I'm the only person who works at Jepsen. I'm still exploring voices for the Jepsen brand--since at some
258.
Hazelcast 3.8.3(jepsen.io)
248 points by aphyr 9y ago|47 comments
259.
by aphyr 9y ago
To be specific, our "byzantine" faults are limited to very specific types of behavior: running multiple copies of the same validator on different nodes, and modifying on-disk state of the validator. We make no changes to the valid
260.
by aphyr 9y ago
The jepsen.io server is literally something I hacked together in a few weekends. I wouldn't really point to it as an exemplar of high quality production-facing engineering. :) That said, it does push 50,000 reqs/sec on my local
261.
by aphyr 9y ago
I had to migrate prod environments when Skyliner shut down; this is the first release I've done on Heroku. Clearly I should have provisioned more capacity. :) Should be back shortly.
262.
by aphyr 10y ago
The clock skew limit controls the maximum latency on reads--so yeah, if you're doing reads that contend with writes, improving clock resolution improves latency. Not entiiirely sure about the throughput impact though.
263.
by aphyr 10y ago
You're correct--large clock anomalies (or VM pauses, message delays, etc) can cause serializability anomalies. I don't exactly know what will happen, because our tests explicitly focused on staying within clock skew bounds. To spe
264.
by aphyr 10y ago
As I mentioned in the article's introductory paragraphs, where Spanner forces a minimum latency on writes to ensure consistency, CockroachDB pushes that latency to reads which contend with a write.
265.
by aphyr 10y ago
To be a tad more concrete: GunDB can run in the browser as well as on servers; CockroachDB is a traditional, client/server database. GunDB is aiming for eventual consistency, but when I looked a year ago, it seemed like the algorithm t
266.
by aphyr 10y ago
Bingo. I've been hoping Google (or, heck, any colo provider) would offer this for a few years. Semi-synchronous clocks enable significant performance optimizations!
267.
by aphyr 10y ago
I agree that it can and should be faster, but the targets will depend on what you're going for. 20k linearizable reads/sec on EBS-backed instances is, uh, maybe a tad optimistic.
268.
by aphyr 10y ago
Just as a headsup, I left Stripe in late 2015 to run Jepsen as an independent consultancy.
269.
by aphyr 10y ago
I just want to emphasize that: a.) this is a beta product where the team has been focusing on completeness and correctness before performance b.) I was testing Cockroach from August through October--Cockroach Labs has put a good deal of wor
270.
CockroachDB beta-20161013(jepsen.io)
309 points by aphyr 10y ago|203 comments

More ›