Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
nikita
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
31 ms
·
391.
▲
Through a Table, Sparsely
(blog.memsql.com)
32 points
by
nikita
13y ago
|
2 comments
392.
▲
MemSQL throws a curve, adds column store on flash
(gigaom.com)
4 points
by
nikita
13y ago
|
1 comments
393.
▲
The current top 5 tech companies in SV with the hardest engineering interviews
(quora.com)
2 points
by
nikita
13y ago
|
0 comments
394.
▲
The Story Behind MemSQL’s Skiplist Indexes
(blog.memsql.com)
5 points
by
nikita
13y ago
|
0 comments
395.
▲
by
nikita
13y ago
Congratulations Solomon and the team. We use docker every day. Thanks for all your hard work!
396.
▲
Unsexy Database Features That Matter: Part 1
(blog.memsql.com)
1 points
by
nikita
13y ago
|
0 comments
397.
▲
by
nikita
13y ago
http://www.storagepricing.org/fusion-io-pricing-at-ces-with-... And $1/Gb is coming soon
398.
▲
Oracle Announces an In-Memory Technology, At an Ungodly Speed And Cost
(blog.memsql.com)
57 points
by
nikita
13y ago
|
27 comments
399.
▲
MemSQL 2.1 Shipped
(developers.memsql.com)
2 points
by
nikita
13y ago
|
0 comments
400.
▲
ICPC World Finals
(myicpc7.icpcnews.com)
2 points
by
nikita
13y ago
|
0 comments
401.
▲
Startup Engineers at Google Code Jam
(blog.memsql.com)
2 points
by
nikita
13y ago
|
0 comments
402.
▲
Being a Startup Chef
(blog.memsql.com)
2 points
by
nikita
13y ago
|
0 comments
403.
▲
Announcing start[c]up
(developers.memsql.com)
8 points
by
nikita
13y ago
|
0 comments
404.
▲
by
nikita
14y ago
This is in the books. It's tricky to do it on the cluster and has perf implications, but it is doable and we have a design for that.
405.
▲
by
nikita
14y ago
there are a lot of new features! Beyond distributing data across multiple machines in a cluster, there's more SQL surface area, multiple levels of redundancy for HA, and a distributed query optimizer. Some cool stuff with bi-directional loc
406.
▲
MemSQL ships 2.0, scales across hundreds of nodes, thousands of cores
(developers.memsql.com)
127 points
by
nikita
14y ago
|
29 comments
407.
▲
by
nikita
14y ago
Right on!
408.
▲
by
nikita
14y ago
I definitely wouldn't argue with the fact that we have to think what we build, design carefully, etc. However imagine the world with massive test coverage that provides instant regression testing and the world without it. What would you cho
409.
▲
by
nikita
14y ago
We are not following any specific methodology such as agile. But we definitely pay a lot of attention to the varios engineering ideas and practices.
410.
▲
by
nikita
14y ago
We have thousands of C++ unit tests and hundreds of thousands of functional SQL tests
411.
▲
by
nikita
14y ago
I would say it was an initial push on 1.5 months of one engineer with 10% of time of one engineer for maintenance. As the load increases we have to invest a few days here and there to scale the system.
412.
▲
Practical Techniques to Achieve Quality in Large Software Projects
(developers.memsql.com)
76 points
by
nikita
14y ago
|
24 comments
413.
▲
Common Pitfalls in Writing Lock-Free Algorithms
(developers.memsql.com)
180 points
by
nikita
14y ago
|
33 comments
414.
▲
Competition programmers at work. One year history of working with top coders
(developers.memsql.com)
14 points
by
nikita
14y ago
|
1 comments
415.
▲
by
nikita
14y ago
Congratulations to the mailgun team!
416.
▲
by
nikita
14y ago
We stop accepting write queries. We also work on sharding. RAM gets you pretty far 12K buys a terabyte these days.
417.
▲
by
nikita
14y ago
- OLTP: lots of small high throughput transactions - Real time analytics (you ingest data fast and query it at the same time) More use cases are going to open up as we build more features.
418.
▲
by
nikita
14y ago
MemSQL uses multiversion concurrency control (MVCC), lockfree data structures and code generation. We are also building modern features like sharding. Times 10 uses row locks, doesn't compile queries, doesn't work on a cluster.
419.
▲
by
nikita
14y ago
> It seems as if it's designed for situations where you're doing a lot of table scans. This is not quite true. Lockfree data structure work exceptionally well for a lot of small concurrent reads and writes. If you look at the benchmarki
420.
▲
Reading Between the Benchmarks: How MemSQL Designs for Speed and Durability
(developers.memsql.com)
64 points
by
nikita
14y ago
|
36 comments
More ›