jeffffff

14 ms·

91.
by jeffffff 6y ago
because things can and will fail inbetween updating the db and publishing the event leading to inconsistency between your database and kafka. a better approach would be to update the db and then use something like debezium or maxwell to pul
92.
by jeffffff 7y ago
materialize has to be able to keep all of its state in memory which only makes sense from a cost perspective for workloads which have a high value to space ratio. fine grained user behavior data typically just isn't that valuable.
93.
by jeffffff 7y ago
from what i've heard from reliable sources there are only 2 spanner clusters, one for ads and one for everything else. i'd be surprised if there isn't an isolated one for gcp but for internal google products there are only 2.
94.
by jeffffff 7y ago
beyond some fairly large size of company it's less that it makes data warehousing easier and more that it makes centralized data warehousing possible. fortunately this type of environment is available today as a managed service in a fe
95.
by jeffffff 7y ago
yes they are database engines, but in many if not most of these cases there is only a single instance that is shared across all products at the company. it is very different than the rds model. of course there are access controls and abstra
96.
by jeffffff 7y ago
yes, at scale this turns the database into the bottleneck. this is not strictly a downside though, as it means you can centralize ownership of your database to one team of experts who can handle optimization, capacity planning, sharding, mu
97.
by jeffffff 7y ago
you don't use cloud to avoid dealing with administrating systems, you use cloud to avoid dealing with systems administrators
98.
by jeffffff 8y ago
have you driven any recent bmw with an automatic? the zf 8 transmissions are really nice and faster than the manual versions. dct equipped cars have also been faster than the manual versions for years. the only reason to drive a manual anym
99.
by jeffffff 8y ago
the jvm lacks structs and more specifically arrays of structs as a way to allocate memory. this causes extreme bloat due to object overhead as well as a ton of indirections when using large collections. the indirections destroy any semblanc
100.
by jeffffff 9y ago
i think it's fine in principle but in this case the implementation was poor. mostly i don't understand why non-reusable paper bags were banned. they're recyclable, biodegradable, and made from a renewable resource. the regula
101.
by jeffffff 9y ago
the non-reusable plastic bag ban in austin, tx was a pretty big failure. the problem is that a lot of the big stores (heb in particular) sell heavier "reusable" bags for 25 cents that no one actually reuses. these bags have a sign
102.
by jeffffff 9y ago
if you want bit level granularity you can use either SIMD-BP128 or SIMD-FastPFOR from https://arxiv.org/pdf/1209.2137.pdf
103.
by jeffffff 9y ago
actually you are wrong, heat pumps (air conditioners) typically have a coefficient of performance that is greater than 1. this means that they move more heat energy from the cold side to the hot side than they consume in moving the energy.
104.
by jeffffff 9y ago
Every part of the cap theorem is commonly misunderstood. Availability is probably the most commonly misunderstood aspect as is outlined in this article. People commonly conflate consistency with durability when really CAP says nothing about
105.
by jeffffff 9y ago
with the exception of the log itself (and its in memory representation) it is a persistent data structure
106.
by jeffffff 10y ago
the reason for putting everything in json in one column is because alter table on a large database can take days. the only sql database i'm familiar with that doesn't have this problem is tokudb. > Serialisation time makes retu
107.
by jeffffff 13y ago
There are several issues in implementing gc as a library. The first thing a concurrent GC must do to make progress is identify the stack roots. This involves stopping all the threads and reading their stacks and register values. To stop all
108.
by jeffffff 13y ago
you can implement a rope as a 2-3 finger tree of string literals. i wrote a java implementation of 2-3 finger trees a few months ago and implemented both ropes of bytes and ropes of chars (rope backed by finger tree of java Strings here: h
109.
by jeffffff 13y ago
hash indexes tend to have really terrible write performance because the locations of the writes on disk are random. lsm trees have way better write performance.
110.
by jeffffff 13y ago
as a backend engineer the publish/subscribe part of this seems like a scalability nightmare. is anyone using the publish/subscribe features on a large site with lots of instances?
111.
by jeffffff 14y ago
wow i didn't realize madvise could actually modify the memory contents and return pages, but it makes sense that it can because that is a useful feature. very cool!
112.
by jeffffff 14y ago
i don't know of any malloc implementations that return memory to the system. the only way to do that is to pass a negative value to sbrk, which requires all the memory being returned to be at the end of the data segment. even if you free
113.
by jeffffff 14y ago
in practice instruction level parallelism and data access patterns make a big difference. if each memory access takes 100 ns but you can do 100 at a time, most of the time there is no practical difference from each memory access taking 1
114.
Indeed Logrepo: Enabling Data-Driven Decisions(engineering.indeed.com)
5 points by jeffffff 14y ago|0 comments
115.
by jeffffff 14y ago
i actually very much doubt that they can do better. writing any sort of large LRU cache in a machine with swap turned on is a bad idea because a lot of your cache will get swapped out and then swapped back in unnecessarily when you try to
116.
by jeffffff 14y ago
you can do this on heap with byte[] using Unsafe.arrayBaseOffset and getInt(Object, long), putInt(Object, long, int) and friends
117.
by jeffffff 14y ago
no
118.
by jeffffff 14y ago
what components were you having issues with the most?
119.
by jeffffff 15y ago
releasing without generics is a huge mistake
120.
by jeffffff 15y ago
propagation of an electromagnetic field in copper is actually only 75% of c, it's a significant factor in processor speed

More ›