I was gonna note how similar this looks to how Jensen operates - he also has something like 50 direct reports, in some ways removing the need for skip level interactions often.
I consider that ebb and flow a totally acceptable part of human reality - as long as you find yourself not self-destructively procrastinating in the lower energy times, it seems sustainable to go at 75% (making up a number) sometimes and 125% later when the time or energy is available to you! I like feeling like it's my choice to go at 125% though - personal agency has always been key to keeping myself invested too.
The behavior modeling you describe in your post most resonated with me at the present moment - as I become more of a senior employee, I feel pressure to perform from management, but also pressure to model sustainable behaviors. The Netflix culture book covered some of this with respect to time off. Top-down is important for culture setting.
I am holding up well, thank you! But ~6 months at the new company, I wasn't holding up great and needed to seriously rethink my time management (hence a lot of this being fresh in my brain). I'm fortunate to have been able to fix this to the best of my ability before my son arrived - but at a company with a pride for startupiness (read: at times a total lack of planning), the balance I have needs pretty regular maintenance.
This is great, and covers aspects of work balance/integration/whatever-you-want-to-call-it that I haven't seen talked about as much.
A big problem I see on an ongoing basis is the confusion between time spent working and work produced. After ~10 years in this field, I feel like a pitcher who loses their fastball sometimes, and have learned when to sense that and pull myself from the game (stop working). I feel for people who don't see or sense their performance deteriorating in a way that just calling it a day and coming back with fresh eyes could improve.
Should you just plow through sometimes? Definitely. But if I find myself needing to consider that for more than about a week a quarter, someone probably planned poorly or is taking advantage of the fact that I care about completing my work.
There are other comments about work vs personal/skills development - the latter I do not consider work, personally, even if they're inter-related at times. I have actually noticed a big difference between my previous employer and current employer. I had enough slack at my previous employer to pack in quite a bit of high value personal/skills development after I finished my work and on weekends. Ironically, its this slack that gave me the time to learn skills that I was eventually hired for at my current employer - however, I have none of this mental bandwidth or time at my current employer. I can't figure out how much of that is pandemic-related, new child related (had a baby while at the new employer), or new employer culture/balance issues with how I work best. I guess it's just hard to track the true cause when the world goes into a blender while one's life changes significantly too.
As someone who lives in the Houston metro area, I totally agree that there are places in the country better situated at the present moment. My broader point aligns with the other comment here - on the timeframe in which we'd think about these problems, virtually every area of the country will find itself wrestling with changing conditions, many of them with a net negative result. There was an interesting map I came across that tracked where climate change forecasts would move the "ideal zone" of habitation in North America - IIRC, quite a bit of the zone shifts into southern Canada and the northern midwest.
Are we proposing to relocate every resident and business along the US gulf coast? I don't think Texas is more or less susceptible to hurricanes than the rest of the gulf.
And from the looks of the northwest, there's a chance we'd be relocating folks into the path of worsening wildfires.
Climate change is serious and we should respond with corresponding urgency, but let's not pretend that there's anywhere in the US that is 100% guaranteed to be a safe haven.
I don't ever see the actual modeling work around AI becoming blue collar. It's complex and worth the price to the companies that can build a business around it. The data labeling, however, can be pretty simple.
It's true. I may have actually been higher for some loans until I got a couple of promotions and had enough of an income increase to be able to refinance to a 5 year 4.75% loan at SoFi. Which itself still feels high as you watch DAILY interest compounding happen.
I lean fairly liberal, but I think we can make significant strides with this issue in the US by simply changing what max interest rates are and how this interest compounds. The way things work right now truly feels like robbery when you're in the thick of it.
Some variation of this gain is implemented in various HPC systems as well. For PCIe response vs routing through the CPU/OS, you see all sorts of offload engines and CPU bypass technology being used in networking stacks. For small IO being stored efficiently to minimize the lookup->retrieval process, data-on-metadata in Lustre (I'm sure other scale-out storage systems have similar implementations).
The march is of course to have the devices on the most high performance bus that makes sense from a cost efficiency standpoint, hence their mention of GenZ and similar technologies. We're definitely getting closer in the industry, I was just using an FPGA connected to a GenZ bus that exposed an NVMe device this week! Early prototype, but it worked in Ubuntu 18.04.
It really depends on what the workload is and if performance or scale are priorities. Our team and the products we build come from the HPC side of the fence and we've been engaged with all sorts of large businesses that know their performance requirements (think 100s of GB/s read bandwidth), and it leads them down the path of either HPC filesystems or newer NVMe-oriented filesystems. Does 'AI' or 'data science' always mean 'high performance'? No, but it definitely can.
Everyone's data volume is different as well...working a proposal for several petabytes of NVMe storage right now. It's not everyone, but use cases are out there for large volumes of high performance storage.
Fair point on the ability to swap the hdfs interface trivially, I hope it's that simple everywhere. Another issue we run into is that these companies have invested heavily in these Hadoop clusters, and would prefer to continue to get some use out of the mountains of HDDs that are captive in these environments. So tiering/HSM functionality is another facet of the issue that these environments will anchor for quite some time.
We're seeing a lot of regret around sprawling Hadoop deployments, so this doesn't surprise me. Other Hadoop vendors (vendor?) pivoting to machine learning is a bandaid as the compute capabilities scale beyond HDFS's performance limitations. Look towards new-gen startups around NVME/NVMEOF (WekaIO, Excelero, E8, etc etc) to fill the void.
The question is going to be: will anyone provide an intelligent way to maintain compatibility with applications that expect to interface with HDFS rather than POSIX? It's a bit of a gap right now from what we see.
There are all sorts of synergies here that might not be obvious:
1. Slingshot interconnect enables HPE to offer their own high performance fabric. This is important as NVIDIA scoops up Mellanox.
2. Cray is clearly ahead of HPE in exascale systems. Cray's tech along with HPE's supply chain should make the exascale even more 'real'.
3. Cray picked up Xyratex a year or two ago in a deal to expand their Lustre solution capabilities. HPE has been growing their Lustre capabilities as well, but this launches both companies into a position to better compete with DDN in the future.
4. Cray + SGI reunites the IP from the original 'Cray' supercomputing company.
Moonshot turned into HPE's Edgeline product. You got me on memristor, but those were sins of previous company leadership. Antonio has proven to be a fairly realistic leader so far, IMO.
High performance networking on the whole, of which HPC GPU solutions are only one use case. And there's the continuing trend of sticking offload engines in the form of ARM ASICs alongside devices that connect over PCIe.