text
stringlengths
0
1.82k
[2534.30 --> 2535.10] project which
[2535.10 --> 2535.54] is available
[2535.54 --> 2536.06] on GitHub
[2536.06 --> 2536.94] open source.
[2537.68 --> 2538.50] A very
[2538.50 --> 2539.20] nice and
[2539.20 --> 2539.94] technically
[2539.94 --> 2540.54] in-depth
[2540.54 --> 2541.06] article.
[2541.28 --> 2541.66] Just the
[2541.66 --> 2542.40] sort of
[2542.40 --> 2542.78] stuff that
[2542.78 --> 2543.38] I love to
[2543.38 --> 2543.84] wake up to
[2543.84 --> 2544.16] with my
[2544.16 --> 2544.50] coffee.
[2545.26 --> 2546.04] It even
[2546.04 --> 2546.40] has some
[2546.40 --> 2547.00] RocksDB in
[2547.00 --> 2547.28] there for
[2547.28 --> 2547.44] you,
[2547.50 --> 2547.64] Eric.
[2548.44 --> 2549.26] I love me
[2549.26 --> 2549.92] some RocksDB.
[2551.36 --> 2552.56] I read that
[2552.56 --> 2552.98] post too.
[2552.98 --> 2553.14] It's
[2553.14 --> 2553.34] actually
[2553.34 --> 2553.54] really
[2553.54 --> 2553.92] interesting.
[2554.30 --> 2554.76] REND
[2554.76 --> 2556.62] is kind
[2556.62 --> 2556.94] of wire
[2556.94 --> 2557.54] compatible
[2557.54 --> 2558.02] with
[2558.02 --> 2558.68] Memcache
[2558.68 --> 2558.98] D.
[2559.64 --> 2560.04] Basically,
[2560.40 --> 2560.80] what they
[2560.80 --> 2562.20] implemented
[2562.20 --> 2563.20] was this
[2563.20 --> 2564.22] proxy
[2564.22 --> 2565.04] almost in
[2565.04 --> 2565.64] between their
[2565.64 --> 2566.48] clients and
[2566.48 --> 2567.08] Memcache
[2567.08 --> 2567.42] D.
[2567.72 --> 2568.40] They implemented
[2568.40 --> 2569.72] an L1 and
[2569.72 --> 2570.74] L2 cache
[2570.74 --> 2571.58] so that
[2571.58 --> 2572.52] Memcache
[2572.52 --> 2573.32] D was the
[2573.32 --> 2574.66] L1, but
[2574.66 --> 2575.22] obviously they
[2575.22 --> 2575.74] could swap
[2575.74 --> 2576.30] that out.
[2576.90 --> 2577.30] Then they
[2577.30 --> 2577.72] were using
[2577.72 --> 2578.74] RocksDB to
[2578.74 --> 2580.08] communicate with
[2580.08 --> 2581.10] their SSDs
[2581.10 --> 2581.38] as kind
[2581.38 --> 2581.62] of like
[2581.62 --> 2582.16] an L2
[2582.16 --> 2582.64] cache.
[2583.28 --> 2583.84] All of
[2583.84 --> 2584.24] this was
[2584.24 --> 2585.20] to reduce
[2585.20 --> 2587.90] their financial
[2587.90 --> 2589.10] costs monthly
[2589.10 --> 2589.72] for their
[2589.72 --> 2590.94] Amazon instances
[2590.94 --> 2591.76] for high
[2591.76 --> 2592.36] memory because
[2592.36 --> 2592.56] they were
[2592.56 --> 2594.04] storing lots
[2594.04 --> 2594.62] of user
[2594.62 --> 2596.04] data in
[2596.04 --> 2596.44] memory.
[2596.84 --> 2597.50] They'd have
[2597.50 --> 2598.54] the hot
[2598.54 --> 2599.30] data set
[2599.30 --> 2600.14] in a given
[2600.14 --> 2600.80] region, but
[2600.80 --> 2601.18] they'd also
[2601.18 --> 2601.72] have a cold
[2601.72 --> 2602.62] data set so
[2602.62 --> 2603.14] in case
[2603.14 --> 2604.18] people
[2604.18 --> 2604.88] failed over
[2604.88 --> 2606.76] from another
[2606.76 --> 2608.00] region and
[2608.00 --> 2608.28] things like
[2608.28 --> 2608.46] that.
[2608.66 --> 2609.44] It's really
[2609.44 --> 2610.44] interesting how
[2610.44 --> 2610.96] much they
[2610.96 --> 2611.80] dropped off.
[2612.00 --> 2612.82] I love the