Alex Miller

If you're an RSS user and a South Bay Systems attendee, I've added an RSS feed for the events at southbaysystems.xyz/atom.xml

Our next South Bay Systems meetup will be on March 31. We've got two awesome deep-dive talks: - @ssougou will present deconstructing consensus and its application to Multigres - @stuhood will talk about how modern full-text search is evolving toward columnar systems (and the tricky optimization challenges) Food and beverages will be provided, courtesy of our hosts, @Snowflake Come hang out with the systems crowd 👇

There's a few papers which argue that DBMSs do page eviction wrong, and they always feel like incredibly compelling arguments. As a bonus, "Writeback-Aware Caching" pdl.cmu.edu/PDL-FTP/Storag… points out that evicting a dirty page is more expensive than evicting a clean page.

[CIDR '25] Linear Elastic Caching via Ski Rental vldb.org/cidrdb/papers/… You should consider that holding a page in cache costs you, because RAM itself is expensive, and existing page replacement algorithms look at sizing cache independently (via miss-ratio curves).

I continue to have a weakness for papers that show B-Trees and LSM-Trees as just two opposite tradeoffs along a continuum. x.com/AlexMillerDB/s…

[arXiv] Dynamic read & write optimization with TurtleKV arxiv.org/pdf/2509.10714 TurtleKV shows a way to elastically move around the RUM conjecture space depending on what is important at the moment.

@sunbains Sort of the point of accord though is to avoid having to establish leadership. The key part is that the leader is necessary for good performance to ordering (conflicting) proposals, and accord does that by timestamping the proposals on the client instead.

Some type of dynamic “leadership “ for hot ranges will probably prevent the worst case being worse than leadership protocols I suppose. Need to study the Casandra implementation I think. This seems obvious.

I’ve been intrigued by Accord, trying to understand it a little deeper. The “fast path” use case is very compelling, in a uniform distribution I can see it working really well. However, real world workloads often exhibit Zipfian distributions. The LSM use case relies on this reality. For such use cases Accord seems to have a worst case that can be slower than leadership based protocols because it’s “slow path” requires renegotiation of the timestamps (and dependencies). I’m wondering if this intuition is correct?

[VLDB '26] Garnet: A Next-Generation Cache-Store for Accelerating Applications and Services vldb.org/pvldb/vol19/p2… It's fast, durable redis, brought to you by Badrish Chandramouli (et. al), known for other 🔥 work like FASTER and Bf-tree.

[CIDR '25] Adaptive Factorization Using Linear-Chained Hash Tables vldb.org/cidrdb/papers/… Adaptive execution + factorization + WCOJ = great paper. The best intro to factorized databases I know of is youtube.com/watch?v=-XmJ7a….

If you make an account you can seed your lists of interest off of my own: @linearizable/interests" target="_blank" rel="nofollow noopener">scour.ing/@linearizable/…

scour.ing has gotten pretty good at surfacing what new stuff I actually want to read on the internet, better than following subreddits. You can see my feed of mostly database things at @linearizable" target="_blank" rel="nofollow noopener">scour.ing/@linearizable. It surfaces small personal blogs particularly well.

[VLDB '25] MD-MVCC: Multi-version Concurrency Control for Schema Changes in Azure SQL Database vldb.org/pvldb/vol18/p4… A great discussion of the end-to-end impact of allowing multiple versions of schema metadata information to be live concurrently, in a real, production system.

@samlafer Hadn't seen! I'll take a look and add it to the list of pending changes if it pushes the frontier of any of the reconfiguration efficiencies. Thanks!

@AlexMillerDB just stumbled upon news.ycombinator.com/item?id=231306…, which talks about the LCR protocol, which seems to be a ring-topology variant of chain replication. Curious why you didn't include it in your transactional.blog/blog/2024-data…. Any specific reason or you just didn't know about it?

The recording from our last South Bay Systems meetup is up! youtu.be/8j1Dqi2M_pI

@sunbains Only Alibaba has exposed (proper) RDMA on public instances, because they chose iWARP. Seems like you have to thoughtfully plan topology and rely on well behaving clients for RoCE to work well. Putting a full TCP stack in your NIC is expensive, but makes things easier to run?

@sunbains Microsoft 2016 SIGCOMM paper dl.acm.org/doi/pdf/10.114… and newer 2023 NSDI paper usenix.org/system/files/n… are great reads. It's a lot of practical issues around multi-vendor and different version interop, routing issues, flow control and PFC storm mitigation, etc.

RoCE works reasonably well, yet I don’t see many people using it. Why? I came across this paper explaining some of the why. This explains why hyperscalers are moving away from RoCE but not everyone is building data centers. arxiv.org/pdf/2302.03337 Seems RoCE is widely used in: - Storage (NVMe‑oF, Ceph, BeeGFS) - Medium‑scale AI clusters - Datacenters with controlled topologies - Environments where IB is not an option

Our next South Bay Systems event will be on January 21st, featuring speakers from CIDR! Come to Databricks to hear about: * xNVMe by @pinartozun of ITU * Spilling in QP by Maximilian Kuschewski of TUM * NPUs in DBs by Alexander Baumstark of TU-Ilmenau luma.com/8a54z94d