donavanm

15 ms·

151.
by donavanm 2y ago
You might want to highlight chunking and how embeddings can/should represent subsections of your document as well. It seems relevant to me for cases like similarity or semantics search, getting the reader to the relevant portion of the
152.
by donavanm 2y ago
It looks like different versions of the ‘Pro’ based on core count and memory bandwidth. Im assuming the 12c Mini M4 Pro has the same memory bandwidth/channels enabled as the 14c MBP M4 Pro, enabling the 64GB. My guess would be related
153.
by donavanm 2y ago
Its reality. Its generally about site and infra access, including power and fiber paths. The bigger providers (eg AWS) simply dont have more feasible sites that are within a few ms of the existing region DCs. Expect to see more infrastructu
154.
by donavanm 2y ago
Shared resources and multitenancy are how you get efficiency and density. Those are at direct odds with strict security boundaries. IME you need hardware supported virtualization for consistent security boundary of arbitrary compute. Linux
155.
by donavanm 2y ago
Is that devices as in digital/alexa land? Never had too much overlap there. AWS and CDO were discrete for incident and problem management after ‘14 or soz
156.
by donavanm 2y ago
No. That was operational modification of system state using existing tools. The “miss” was an intended subset filter that was not interpreted correctly. > an authorized S3 team member using an established playbook executed a command whic
157.
by donavanm 2y ago
I worked at AWS for 13 years. I did “aws call leader” for 7 years, and worked in the reliability org when we rebuilt the coe tool. Ive personally blown up a service or two, and know other PEs whove done the same or larger. Ive never heard o
158.
by donavanm 2y ago
No. The majority of teams and individuals are using it as intended, to understand and prevent future issues from process and tool defects. The complaints Ive heard are usually correlated with other indicators of a “bad”/punitive team c
159.
by donavanm 2y ago
As I recall the coe tool “automated reviewer” checks cover this. It should flag any content that looks like a person (or customer name) before the author submits it.
160.
by donavanm 2y ago
> self-contained infrastructure/application stacks that provide fault boundaries There. That's it, more or less. Having 10% of my customers sad is better than 100% of my customers being sad. It's better to have a handful o
161.
by donavanm 2y ago
If anyone is interested in space & time efficient caching and working set tracking check out some of my favorites, gil einziger https://scholar.google.co.il/citations?user=kWivlnsAAAAJ and the ex coho data guys https:&
162.
by donavanm 2y ago
Process development, feature size, and ultimate yield are probably what theyre after. Yes, for the past 30+ years everyone has used a combination of disabling (“fusing”) unused/unreliable logic on the die. In addition everyone also “bi
163.
by donavanm 2y ago
No, this an incorrect analogy on most points. Amazon retail marketplaces are absolutely not correlated with distributed data centers. Most markets are served out of a single cluster, which is likely to be in a different country; eg historic
164.
by donavanm 2y ago
Im not familiar with this cases specifics, but AWS also has an approach of virtualizing actual hardware interfaces (like nvme/pcie) to the host through dedicated hardware/firmware. I wouldnt be surprised if their solution was to m
165.
by donavanm 2y ago
I literally worked for a “home town” newspaper in the 90s and knight ridder in the early 2000s. This is not in any way comparable to syndication. Syndicated stories have an accurate byline, the publishing paper paid for syndication/con
166.
by donavanm 2y ago
A mismatch in risk and term duration. In the worst case its short term cashflow, converted to profit ona regular basis. the huge tail risk is of a longer (likely) term. On a normal year reinsurance pays out expected losses, the primary insu
167.
by donavanm 2y ago
In addition to other comments see https://en.m.wikipedia.org/wiki/Ishikawa_diagram or the “new school” of safety engineering thats trying to get away from ideas like “RCA.” Sidney Dekker publishes an older version of h
168.
by donavanm 2y ago
Fyi there is a ton of good work in storage and cdn systems that is very similar. Check out some of my favorites, gil einziger https://scholar.google.co.il/citations?user=kWivlnsAAAAJ and the ex coho guys https://
169.
by donavanm 2y ago
Not specific for GPUs but I believe some of those giant and deeply discounted buys are at/below typical cost because of volume. They allow the vendor to increase their OEM/manufacturing commits, or shift bins theyre long on, to im
170.
by donavanm 2y ago
I dont believe your perception is accurate. I worked for Knight Ridder during this time, and print news was already a walking corpse. Cable/satellite news channels, and broadcast tv, and even radio before that had worn away the primacy
171.
by donavanm 2y ago
If youre interested in EC you might want to consider larger multi dimensional cases. Think of encoding not just across spindles, but another failure domain like rack, room, DC, or region. The goal being to tolerate common component failures
172.
by donavanm 2y ago
The americans are already on this path https://aws.amazon.com/blogs/aws/in-the-works-aws-european-s... .
173.
by donavanm 2y ago
Again, unless you have domain expertise I suspect you are drastically underestimating the complexity of reality. As I alluded to earlier metering, pricing, and billing are _at least_ three separate complicated problems in substantial system
174.
by donavanm 2y ago
Hes telling you the reality. Cloud providers dont want to _dissuade_ hobbyist users. But thats not their target market and they will make business decisions based on the revenue generating customer profile. Thats just being frank and honest
175.
by donavanm 2y ago
Ok, so you think unintended usage should be costed out by the provider. Sure, T&S and support definitely have a handle on the topic. Now what? Because _today_ it’s already baked in to the P&L and pricing. You want the provider to gi
176.
by donavanm 2y ago
I dont understand what youre on about. The post I replied to, and many others, use “caps” to refer to limits of service based on billing. This is an endless source of comments on every “cloud” topic. I provided a very brief overview of why
177.
by donavanm 2y ago
FWIW I believe your reasoning is exactly what I saw/learned at AWS. The potential cost of losing/disappointing “real” customers greatly outweighed the notional cost of lost “scared” customers. One interesting thought was trying to
178.
by donavanm 2y ago
Metering, pricing, and billing is way more complicated than you might assume. There are historical posts here with more details if you search. In short its all async, theres variable lag, theres multiple “types” or dimensions to metering, t
179.
by donavanm 2y ago
See other child reply upthread, lots of service-to-service style interactions that look more like distributed state than a CR. And my view was across an org scope where even “infrequent” quickly accumulated. AWS is on the order of 50,000 SD
180.
by donavanm 2y ago
No. Youre missing the UX forest for the pedantry trees here. Ive worked on a team that did similar change detection with little to no ML magic. It matters how its presented as a hint (“top five suggested”) and not THE ANSWER. In addition it

More ›