text stringlengths 0 1.8k |
|---|
[1321.86 --> 1322.08] Yeah. |
[1322.56 --> 1322.86] Yeah. |
[1323.32 --> 1325.84] Well, I don't think anybody's optimized for reading. |
[1326.18 --> 1327.66] Unless you're imager or something like that. |
[1327.82 --> 1329.80] Maybe you're optimizing for those things to be super fast. |
[1330.88 --> 1332.70] Large GIFs like that, for example. |
[1333.26 --> 1336.88] Well, if we had it on a CDN in Hong Kong, it would be much faster. |
[1336.88 --> 1337.32] Exactly. |
[1337.32 --> 1337.52] Okay. |
[1337.52 --> 1338.96] So that's the question I was thinking of asking. |
[1339.10 --> 1343.28] Like, okay, the observability lets us know this event happened, right? |
[1343.32 --> 1348.38] The event being this GIF was served from Newark to Hong Kong at this speed. |
[1348.60 --> 1349.96] It's this size, et cetera. |
[1349.96 --> 1352.58] The other question is, it was a miss. |
[1352.78 --> 1353.88] So why was it a miss? |
[1354.04 --> 1357.68] So these are questions we begin to answer ourselves as we dig into this. |
[1357.76 --> 1358.80] Okay, why was it a miss? |
[1358.88 --> 1359.68] Okay, now we know. |
[1360.16 --> 1361.94] And we figure, what was the answer to that? |
[1362.02 --> 1362.84] Why was it a miss? |
[1363.94 --> 1364.80] Or was it a cash miss? |
[1365.34 --> 1367.04] First Hong Kong visitor of the day. |
[1367.40 --> 1368.48] Or since it expired. |
[1368.60 --> 1369.04] Or who knows? |
[1369.60 --> 1369.82] Yeah. |
[1369.82 --> 1375.98] I mean, those are kept in cash right on Fastly and they can't cash like the entire internet. |
[1377.18 --> 1380.38] So even for us, they can't cash like all of our content. |
[1381.38 --> 1382.26] So the... |
[1382.26 --> 1387.54] They can probably cash all of our content at all of their pops and barely ever notice, don't you think? |
[1388.16 --> 1397.90] They could, but I think the reason why they're not is because they have to shed some of the extra content that is not accessed within, I don't know, X hours, days, whatever. |
[1398.46 --> 1406.02] So they don't guarantee that everything will be in the CDN all the time, even though our headers ask for it to be in the CDN for a few weeks, I believe. |
[1406.32 --> 1407.94] I'm not sure exactly this one. |
[1407.94 --> 1413.04] We can check to see how long it should be kept in the CDN for this specific request. |
[1413.76 --> 1416.70] But if it's fire remember, it's just meant to be a few weeks at most. |
[1416.70 --> 1423.06] So if that wasn't accessed in a few weeks and it may expire when it's requested again, which will be a miss. |
[1424.36 --> 1424.48] Right. |
[1425.50 --> 1426.10] And it'll be very slow. |
[1426.10 --> 1428.62] Why won't they just make people pay for that desire then? |
[1429.00 --> 1435.38] I guess if you're a larger site with much more assets than we have, maybe that becomes more and more expensive. |
[1435.60 --> 1439.66] But, you know, it's in our affordance right now to ask them to do that. |
[1439.90 --> 1440.06] Yeah. |
[1440.86 --> 1442.34] That's a great, great idea. |
[1442.34 --> 1443.52] So why won't they offer us a service? |
[1443.70 --> 1445.94] Like, hey, just cash the whole thing indefinitely. |
[1445.94 --> 1449.02] If, you know, and just, and I'll pay for it. |
[1449.50 --> 1451.20] I would love us to be able to do that. |
[1451.28 --> 1453.34] Like, all our stuff should be cached all over the world. |
[1454.00 --> 1455.08] Like, I agree. |
[1455.12 --> 1456.22] What's our total assets on stuff like that? |
[1456.26 --> 1459.02] What's our total, like, what would be the weight? |
[1459.20 --> 1459.88] In terabytes? |
[1460.44 --> 1460.88] No. |
[1461.00 --> 1461.32] No. |
[1461.34 --> 1461.92] Or in gigs? |
[1462.28 --> 1464.00] A hundred, 150 gigs? |
[1464.48 --> 1464.84] That's not that much. |
[1464.84 --> 1465.90] I mean, that's pretty reasonable. |
[1465.90 --> 1472.32] I mean, I can go buy a 14 terabyte hard drive for under 400 bucks. |
[1473.04 --> 1478.26] Yeah, but you need to multiply that times, you know, how many times you want, how many pops you want. |
[1479.06 --> 1480.80] So, but still, still you're right. |
[1480.96 --> 1481.10] That's true. |
[1481.10 --> 1482.52] It's not a lot of data. |
[1482.80 --> 1484.20] I wish it was cached. |
[1484.34 --> 1489.92] And I wish we had an e-tag implementation so that if the content doesn't change, it won't expire from the cache. |
[1490.80 --> 1493.02] And, I mean, we have it configured. |
[1493.14 --> 1503.14] We have cache shielding so that, or like pop shielding, which means that there should be at least one pop where this is always kept in cache. |
[1503.14 --> 1507.40] So, if another one doesn't have it, it should get it from that pop rather than come to us. |
[1507.84 --> 1507.98] Right. |
[1508.06 --> 1509.60] And their network's probably faster than ours. |
[1509.96 --> 1510.64] Of course, yes. |
[1510.64 --> 1510.82] Right. |
[1511.02 --> 1512.24] It should be at least by design. |
[1512.26 --> 1513.18] It's optimized, right? |
[1513.32 --> 1515.18] I mean, they should, yeah, they have all the optimization. |
[1515.32 --> 1518.42] They have the best routing between their pops, which is how it should be. |
[1518.98 --> 1519.74] So, you're right. |
[1519.82 --> 1522.04] I mean, but this we never had before. |
[1522.10 --> 1523.26] And this is the exciting thing. |
[1523.32 --> 1528.20] So, now we know why our 99th percentile, why we have such a bad tail latency. |
[1528.64 --> 1530.34] Because sometimes this stuff happens. |
[1530.34 --> 1533.64] We didn't have this visibility before, and that's the exciting stuff. |
[1535.62 --> 1537.82] Where does the law of diminishing returns come in? |
[1539.82 --> 1541.12] I didn't hear any of you. |
[1541.76 --> 1542.64] Do you want to try again? |
[1543.12 --> 1546.30] When does the law of diminishing returns come in? |
[1546.76 --> 1548.50] Because, you know, slow clients are slow. |
[1549.12 --> 1550.70] We can't make every request fast. |
[1551.38 --> 1558.48] Where do we know now we're just basically toiling away at something that's not worth our time anymore? |
[1558.48 --> 1560.88] Versus this is actually a valuable optimization. |
[1561.40 --> 1562.80] I'm really glad you brought this up. |
[1562.90 --> 1566.88] Because we have, this is something which we weren't able to see before. |
[1567.04 --> 1572.08] We have Apple Watches consuming MP3 files. |
[1572.62 --> 1573.90] And they are slow. |
[1574.22 --> 1576.68] So, they take, you know, many, many minutes. |
[1577.22 --> 1580.54] Like, our longest consumer was something like 40 minutes. |
[1580.54 --> 1584.90] Imagine someone being connected to our website and consuming MP3s for 40 minutes. |
[1585.42 --> 1586.40] It was on Apple Watch. |
[1586.40 --> 1588.90] And there's, like, a couple others like that. |
[1589.50 --> 1597.70] So, when it comes to content that is not in the cache, I don't think we should spend much more time on that. |
[1597.70 --> 1601.30] Except if we're talking about using an object store versus local store. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.