abadid

7 ms·

1.
An Analysis of Two Architectures for Agentic Data Analysis(starburst.io)
3 points by abadid 6d ago|1 comments
2.
by abadid 6d ago
AI agents are bad at joining large tables.
3.
Will GPUs Accelerate Your Traditional SQL Workload?(starburst.io)
2 points by abadid 2mo ago|1 comments
4.
by abadid 2mo ago
The hype doesn't always live up to reality.
5.
Benchmarking the JDBC Bottleneck in Trino(starburst.io)
1 points by abadid 8mo ago|1 comments
6.
by abadid 8mo ago
More proof that it is really slow to access data in a database over JDBC.
7.
The dangers of the JDBC bottleneck in Trino(starburst.io)
2 points by abadid 1y ago|1 comments
8.
by abadid 1y ago
Short summary: Trino uses JDBC in ways in which it was not designed and this can result in massive performance bottlenecks. Sometimes partitioning source data can mitigate the bottleneck.
9.
How to Avoid Data Access Bottlenecks When Using Trino(starburst.io)
2 points by abadid 1y ago|1 comments
10.
by abadid 1y ago
A lot of people don't realize how slow it is to access data in bulk from a database using JDBC. JDBC was never designed for the use case that Trino uses it for.
11.
Parquet and ORC's many shortfalls for machine learning, and what to do about it?(starburst.io)
6 points by abadid 2y ago|1 comments
12.
by abadid 2y ago
This article summarizes research from my lab in collaboration with ByteDance published in CIDR (a computer science conference held in Amsterdam two weeks from now) on a new columnar format designed for ML workloads.
13.
by abadid 2y ago
Why use PostgreSQL instead of columnar databases that are likely to perform way better for these types of analytical workloads?
14.
by abadid 5y ago
I agree that in theory they could both co-exist for the reasons you state, but in practice I think it's unlikely a company that invests in a data fabric (which is largely a technology cost) is going to simultaneously invest in the ince
15.
by abadid 5y ago
In general, whenever you need to perform a join (of multiple datasets), that ends the pipeline of local operations on a partition. Other operators as well that necessarily require data from other partitions end local pipelines. This is why
16.
by abadid 5y ago
I'm the author of this piece. I'm happy to respond to comments in this thread.
17.
by abadid 6y ago
IMO, it's hard to put a price on strong isolation and consistency. Being able to write an app that that uses atomic transactions, that are isolated from concurrently running transactions, and that see the correct data is something that
18.
by abadid 7y ago
See what I wrote below regarding Spanner. The same thing applies to the CockroachDB solution. If you run 2PC for multi-region transactions that is very slow (increases latency), and prevents conflicting transactions from running for longer
19.
by abadid 7y ago
Yes --- the post explicitly states that deterministic execution is a prerequisite. SLOG is CP from CAP, so indeed suffers from unavailability in the event of a network partition.
20.
by abadid 7y ago
Let's say you have a transaction that writes data located in different quorums atomically. This is called a "multi-home" transaction in SLOG. Most of the machinery in the paper is dealing with that case. SLOG can do that with
21.
by abadid 7y ago
Latency is measured from the client. In the example in the post (and more details in the paper), you see the latency tail from when clients access data that is far from them. The challenge is to make multi-home transactions no worse than re
22.
by abadid 7y ago
As mentioned in the post: "There are several ways to achieve [serializability] — such as via locking, validation, or multi-versioning." Deadlock happens under some, but not all implementations of serializability via locking. There
23.
by abadid 7y ago
Glad to see to see this post on HN. I'm the author and happy to respond to questions in this thread.
24.
by abadid 8y ago
Please see: http://www.cs.umd.edu/~abadi/papers/determinism-vldb10.pdf
25.
by abadid 8y ago
magicalhippo asked a similar question both with regard to complexity and garbage collection. I think it is best to combine these threads, so please see my response there ... https://news.ycombinator.com/item?id=19003212
26.
by abadid 8y ago
Every transaction has an numeric identifier. These identifiers are used to order versions of updates to a data item. Versions from higher transaction IDs are considered to be "after" lower ones. Reading versions as of a particular
27.
by abadid 8y ago
It is one coordinator per transaction. As far as coordinator robustness, I discussed this in the following paragraph from my post: "There are two categories of work-arounds to the blocking problem. The first category of work-around mod
28.
by abadid 8y ago
The advantage of removing 2PC is achieved in Calvin and Fauna. But I'm arguing in this post that it can also be achieved in nondeterministic systems (or really any system) and maintain the guarantees of that system.
29.
by abadid 8y ago
Actually the approach described by this post does much better than 2PC under contended workloads because it removes the cloggage problem (see the section entitled "The problems with 2PC").
30.
by abadid 8y ago
(1) I'm not sure I fully understand what you are referring to wrt the "acks" but either way, one major difference is that the acks don't have to be made durable in the alg. described by the post. Also, the other key thin

More ›