Huge fan of eBPF but admittedly less informed than I could be; could someone point out some instances where this wouldn't be able to replace agent instrumentation eventually?
I understand the bits of the kernel we're able to hook this way are still being built out, but 5yrs down the road what would we still need other instrumentation to capture?
I'm curious to hear people's thoughts on eBPF generally, it seems likely that this is where observability companies are headed. It's non-trivial to implement but monitoring from the kernel layer makes so much sense that I expect the tooling will come along quickly.
"Open source was taking over enterprises long before k8s"
True, also not related to the point I was trying to make. Executives tend to be handwavey and vague sometimes but I'm pretty sure the "sea change" they're talking about in this case is about containerization and the rapid evolution of open source projects in that space, many of them related to observability. Kubecon had 500 attendees in 2015 and 12,000 in 2019. It's hard to deny that the popularity of that community will sway popular opinion about what technologies to embrace moving forward. I think (and again, I am guessing as I have exactly zero information that the public does not have) that New Relic and other companies are in part making decisions like this because it's becoming clear they will be left behind if they don't, and quickly.
I bet the first post-covid Kubecon will have 20k attendees and be taken over by the execs and VCs who've realized that's where the devs are, just in time for the devs to stop going and start the next thing á la reInvent. Wish I had gone before it blew up.
I also feel like "struggling" is a bit of an overstatement for a company with a $4B market cap, but maybe that's just me. I'm sure most FAANG4LYFers look at that like a lemonade stand.
I've mostly worked for companies that were all in on things like open source and cloud my entire career, but it's easy to forget that there are huge companies that just recently replaced their tape drives.
If I were to guess at the "sea change" discussed herein I'd say it's probably the explosion of tech around Kubernetes and the CNCF, as that created a very sudden and urgent need for better observability tooling.
- I typed these words all by myself and they are not official New Relic words, you can tell because they're not terribly well-formed and they aren't surrounded by logos and trademarks and buzzwords. Using the haphazardly spewed opinions of strangers on the internet as investment advice is a really, really terrible idea; like "open source is a virus" level of bad. Don't do it.
I've had some success convincing anti-cloud (cloud reticent?) companies by pointing to things like FedRAMP compliance: "You know how hard it is to get ______ compliance? Your data is probably safer there..." kind of conversations.
I'm sure your mileage will vary but in my not so humble opinion companies that are still trying to cling to their aging fleet of servers are making pretty poor business decisions. I think this internet stuff is catching on.
- I work at NR, these questionable opinions are mine and not those of my employer though I might have stolen them from my friend Vicky who yells about software a lot. Don't buy stock because I typed some words on a webpage that's ridiculous, also maybe don't buy stock period. Buy bitcoin it's the new pet rocks.
Stagnant seems like an overstatement given 500 commits on one of the agents alone in the last month. There is surely a way to criticize without devaluing the unread contributions of literally hundreds of developers.
I personally am much more motivated by seeing customers encouraged to add support for their favorite libraries and frameworks than scoring their sweet free codes, but I suppose there is potentially a sinister motive behind everything in life... like open source.
While I do not condone the use of the s-word here I understand the sentiment and I am also glad to see that they've open sourced them, however belated the decision.
With regard to opening the servers I've actually just now submitted a request to give all of hacker news direct access to the infrastructure via SSH. I'll let you know what they say.
- I work at NR, my words are terrible and unfit for any purpose but especially investment advice. If you read them and make bad decisions this paragraph is designed to prevent me from going to Martha-Stewart-jail. I hear they have tennis but I quite like choosing my own clothing.
I think this is probably less relevant today - after a certain size the "moat" for an observability company starts to be their existing customer base. If NR works with everything and everything works with NR it's much harder to make a case that you should continue adding alternatives to your observability stack.
- opinions expressed here are exactly those of a misguided employee with too much time on his hands and not those of said employee's corporate overlord, if you take any of my nonsensical rambling as investment advice you do so at your own peril. I regularly gamble all of my rupees away on coinflip games in Zelda.
Didip! Long time - I'm back at NR as of a month ago after 5yrs away.
We did have all the agent code out in the open but it was more "source available" than it was open source. You could see what was happening but not build on it and use it for your own nefarious purposes as you can now.
My understanding is that it has more to do with the general direction of observability. The agent code was always open, we frequently accepted PRs from the community when I was on their Ruby agent team, but they didn't use an open source license.
We're public so I probably shouldn't guess too much about the motivations beyond the explosion of interest in observability/kubernetes/cncf, but I do know there has been support for open sourcing these for a long time within the company and I'm glad to see it finally happen.
disclosure: I work at NR but my opinions are in fact cake and should not be used to make investment decisions (seriously that is a terrible idea I mostly invest in LEGOs and Magic cards)
Came here to share this - NR just redid their entire pricing structure, in my opinion it's now a much better approach for horizontal scaling.
Most everything is now part of a single product: New Relic One. You pay for data ingest beyond the first 100GB and you pay per seat for seats after the first. The AI pieces have some additional costs after a certain point as well, but from my understanding the free tier is very generous.
source: Worked at NR 5yrs ago for several years on an agent team (frequently lamented pricing structure), joined again about a month ago.
New Relic changed their pricing model significantly this morning, now charging for seats (first user is free) with 100GB data ingest, thereafter $0.25/GB.
I understand the bits of the kernel we're able to hook this way are still being built out, but 5yrs down the road what would we still need other instrumentation to capture?