
SVP of engineering @harperdbio, developer of lmdb-js, MsgPack/CBOR libs. Original content tweets. Toots at kriszyp@fosstodon.org
Cottonwood Heights, UT
Joined April 2008
- Tweets1.3K
- Following166
- Followers1.4K
- Likes100
Kris Zyp retweeted
Vector search just got a major upgrade in Harper 5.3. 🚀
⚡ Up to 6.5× higher throughput vs. Harper 5.2
⚡ 1.6–2.3× higher throughput vs. PostgreSQL + pgvector in our 200K-vector benchmark
Native Rust HNSW. No separate vector DB required.
harper.fast/resources/native…
I'm excited to share about some performance improvements we have made in Harper 5.2: harper.fast/resources/how-ha… particular, I think the combination of hash-based verification table with weak LRFU object caches is a database integration innovation that I am probably most proud of.
62
Its been exciting to see HarperDB's growth since I joined 2 yrs ago & we have recently attracted a significant growth investment from Serent capital (serentcapital.com/news/harpe…). This means we are hiring! If you'd like to work with me, check out our opening: harperdb-inc.breezy.hr/p/ba7…
1
212
Kris Zyp retweeted
Choosing a database is a critical decision in building any project, and it's one that you should make early on. Today we have @kriszyp from @harperdbio on the show to talk about how to choose the right database for your project!
Show notes: buff.ly/3sRBgkM
A blog post on how we (and you can) do socket routing for distributing HTTP load across multiple NodeJS worker threads without cluster module/processes, for better resource efficiency and control of request delegation (to achieve session affinity):
harperdb.io/post/worker-thre…
Heading to Lisbon for Web Summit this week. Anybody else going to the conference, want to connect?
1
A blog post on my experience working with bun, my first blog post for @harperdbio!
harperdb.io/blog/is-bun-the-…
Kris Zyp retweeted
since 2015 I keep a benchmark of how fast a MsgPack implementation can be - comparing them using a small, medium, and large input.
Today I've included "msgpackr" (made by @kriszyp), and this one is really hard to beat!
Check full results here: github.com/endel/msgpack-ben…
If someone makes an outstanding research contribution to a bug fix (but wrote no code), recognize it by:
* Add them as a co-author in the commit (need to ask their email address?)?
* Thank/at them in the commit?
* Thank them in release notes?
* Something else?
I have news! I am joining the engineering team at @harperdbio, and excited to be a part of this team building fast scalable databases to help drive high-performance sites and apps!
A wrote a little overview/response of a few different ways that you can model child/related data structures, like posts with comments, in lmdb-js (and perhaps applicable to generally modeling data with lower-level KV stores): github.com/DoctorEvidence/lm…
2022: The year when complex JS + native C capabilities can successfully be run across multiple JS platforms and engines.
Also 2022: Can't publish packages with a shebang script from Windows & install them on linux because carriage returns are just too hard.
2
Has anyone started suddenly seeing npm installs taking 13 min and 30secs in the last few days? My installs on CI suddenly take this long with strange consistency of 13.5min, even when I rerun older jobs that ran fast before. I think stuck on node-gyp, but no errors or warnings.
Trying out the new NAPI support in Bun this morning. This really opens up a lot of opportunities on Bun/JSC! Really cool to see all the advances in JS runtimes lately!
Been testing/playing with Bun's performance. I was a bit skeptical of Bun, but it has some impressive performance capabilities. @jarredsumner has created something with a lot of potential, and thought I would share some observations.
And hopefully looking at JSC vs V8 could offer some nudges to @v8js . In particular TextDecoder is about 4-5x slower (per call overhead) than JSC's version! And I wonder if performance of assigning dynamic properties (where they fit existing hidden classes) could be improved.
2
3







