> Our multi-tenant, cloud-native platform was architected by big data and security experts and has been in operation continuously for nearly a decade. Our platform is built on a modern, microservices-based application and cloud architecture, leverages security-first principles, and incorporates artificial intelligence and machine learning, or AI/ML, algorithms to deliver real-time actionable insights.
Not a single technology buzzword went unused in this document.
> Notifications. Our alerts can directly notify users or teams via emails or other direct means, integrate with third party ticketing or workflow tools, such as OpsGenie, PagerDuty, ServiceNow, Slack, or trigger APIs, scripts, or serverless functions on AWS, Azure, or GCP.
Microservice or monolith has no bearing on whether something is serverless. You could deploy a microservice on a bare metal x86 server, or a raspberry pi, or an IBM Mainframe.
I am really surprised that AWS hasn't already built a competing product like Splunk, Sumo or ELK. Their CloudWatch offering is so obscure and in general not useful.
CloudWatch Insights and their security tools make me think that they’re exploring the idea but planning to grow up organically rather than try a direct feature comparison against established customers.
My previous employer used it, and I use it nowadays for my side project. I like it, but granted, its the only log ingestion software I've ever used, so cant compare it. The charts I made are nice and I can easily pinpoint errors, bottlenecks and similar in my platform/code.
Here are some views that we're using for our little startup:
From my experience, Sumo really touts its `logreduce` functionality, which is a process of filtering logs for anomalous messages and bringing interesting things to your attention.
It worked pretty well to be fair, but it got a bit expensive (as things usually do).
I've used it for some projects alongside Splunk and ELK/Kibana for others. They're all effective at the scales we've thrown at them, but I find Splunk more intuitive to use, and IMO the query language is better documented.
May I suggest you try out Scalyr? Our trials are easy to set up. We are cost effective at large scale ingest, our searches are fast and power queries are super nice!
Yes. I am the devops lead at a startup and it's an absolutely essential piece of our devops stack.
It's basically the Lite version of splunk, 90% of the features at 10% of the cost (which matters when we're talking 10s of thousands of dollars per year).
Do you know how much does it cost per month per GB?
I implemented SumoLogic about 5 years ago at another company and it was great. At some point we will need it in our current company (we have ELK, but it is clunky) and it does not compare with what sumo had to offer 5 years ago. But their pricing page has some crap "credit" metric that does not let you compare it.
Their pricing page has price per gb: https://www.sumologic.com/pricing/
Says the "essentials" is 2.50 per gb so if you're pumping 10gb a month it would be like $750 per month.
Have you tried Scalyr? I am an employee at Scalyr and been in the infra space for quite a while. I think our logging product is the best I've used personally. Searches are actually fast.
Speed hasn't really been an issue with any of the providers I've tried except Loggly which has a very slow UI if you run into big messages. Even with ingesting 200gb of logs a day at one job, per environment, Sumo wasn't slow that I recall.
Of course, we didn't try to negotiate, but it looks more expensive than Sumo. At one company we were spending around $600k a year on it, and it looks like you would cost closer to 1m.
Yes, for years in multiple companies. I greatly prefer it over Splunk or self rolled ELK crap.
If you reformat all your logs to output as json you can "| json auto nodrop" and suddenly you have fantastic statistics off every single field. It's also smart enough to figure out 'close enough' json, like if you're piping log output to syslog which adds it's own metadata.
Their Logreduce feature is also nice for "Ingestion suddenly shot up but eyeballing doesn't show a pattern, help me out here!" Hit the button and get logs grouped by 'these are pretty close except for x."
Amusingly they called me once. I apparently managed to craft a scheduled query that was causing their systems issues. I okayed them adjusting the query to something a little less aggressive.
The only feature that seems wonky is 'real time' (which they limit you on how many you can use). Never was a big issue, I generally would run 'real time' stuff every 15 minutes.
I will be very sad if they die or get bought out by Splunk. It's a great product.
Oh, one other great feature. There's no on-prem option besides the collectors. So someone can't get the bright idea to take it in-house (like with Splunk), wasting the time of a principle engineer for months on a project that doesn't make us any money.
I'll outsource SIEM every time and sleep very well doing it.
We're using CloudWatch, Stackdriver and SumoLogic.
SumoLogic is by far the best.
It still has some room for improvement. But if you're stuck with could providers log aggregators - it's miserable.
The original dotcom bubble had every possible company that stuck "on the internet" to their description, whether it made sense or not. This is a solution which is designed for what it does and has lots of existing customers. I'm not sure it's comparable at all.
Either way, you really need a better breakdown to understand how good or bad this is. Even something as basic as COGS or gross margin will tell you how sustainable (or not) the business is.
Just because the losses are scaling with revenue growth doesn’t mean the business is unsustainable. It could just mean they’re sinking a ton of money into sales/marketing/r&d because they view their market as a land rush.
I haven't been able to find a PDF version of the S-1 filing myself to compare. I will note that I'm also using Firefox and headers are aligned for me.
The irony is that everybody bemoans PDF yet formatting control is exactly why it has such staying power despite being such a hated de facto publishing format.
PDF fulfills a need that nothing else satisfies: a vector format that’s universally supported and guarantees that its layout will be the same everywhere, especially when printing.
61 comments
[ 3.0 ms ] story [ 127 ms ] threadhttps://news.ycombinator.com/item?id=24259821
Not a single technology buzzword went unused in this document.
> Notifications. Our alerts can directly notify users or teams via emails or other direct means, integrate with third party ticketing or workflow tools, such as OpsGenie, PagerDuty, ServiceNow, Slack, or trigger APIs, scripts, or serverless functions on AWS, Azure, or GCP.
Can't believe the cynicism on here. Selling bullshit IPOs is very profitable.
this flurry of IPOs is to take advantage of all the Liquidity the Fed is pumping into the markets
The IPO is the product
Tesla at 1,000 P/E
If you are running a tech company right now, you would have to be crazy to not IPO
you can raise half a billion or a few billion
that buys you 5 to 20 years of life as a company
Here are some views that we're using for our little startup:
https://i.imgur.com/Q1V6ArY.png
https://i.imgur.com/rCfg4nr.png
https://i.imgur.com/GgYHgPA.png
It worked pretty well to be fair, but it got a bit expensive (as things usually do).
I would probably use them again.
May I suggest you try out Scalyr? Our trials are easy to set up. We are cost effective at large scale ingest, our searches are fast and power queries are super nice!
No. Please stop.
It's basically the Lite version of splunk, 90% of the features at 10% of the cost (which matters when we're talking 10s of thousands of dollars per year).
I implemented SumoLogic about 5 years ago at another company and it was great. At some point we will need it in our current company (we have ELK, but it is clunky) and it does not compare with what sumo had to offer 5 years ago. But their pricing page has some crap "credit" metric that does not let you compare it.
How'd you swing that? We are fairly large Sumologic users, and it's more like 80% of the features at 80% of the cost.
It's a rock solid product, it's extremely reliable but it's not that cheap.
If I can't have grep, I'll probably always beg for SumoLogic. It's great, at least compared to what I've used so far.
Of course, we didn't try to negotiate, but it looks more expensive than Sumo. At one company we were spending around $600k a year on it, and it looks like you would cost closer to 1m.
If you reformat all your logs to output as json you can "| json auto nodrop" and suddenly you have fantastic statistics off every single field. It's also smart enough to figure out 'close enough' json, like if you're piping log output to syslog which adds it's own metadata.
Their Logreduce feature is also nice for "Ingestion suddenly shot up but eyeballing doesn't show a pattern, help me out here!" Hit the button and get logs grouped by 'these are pretty close except for x."
Amusingly they called me once. I apparently managed to craft a scheduled query that was causing their systems issues. I okayed them adjusting the query to something a little less aggressive.
The only feature that seems wonky is 'real time' (which they limit you on how many you can use). Never was a big issue, I generally would run 'real time' stuff every 15 minutes.
I will be very sad if they die or get bought out by Splunk. It's a great product.
I'll outsource SIEM every time and sleep very well doing it.
I miss Scalyr, but I think that’s mostly because I haven’t learned how to be very effective with Sumo Logic yet.
Dotcom 2.0
E: quick stats on IPO https://www.statista.com/statistics/270290/number-of-ipos-in...
>>> We generated net losses of $32.4 million, $47.8 million, $92.1 million
The revenues are good and increasing, but the losses, omg.
edit: OMG they got the numbers wrong? check pages 3 and page 13-14. Dates don't match, 2019 and 2020 are mixed up?
Just because the losses are scaling with revenue growth doesn’t mean the business is unsustainable. It could just mean they’re sinking a ton of money into sales/marketing/r&d because they view their market as a land rush.
The irony is that everybody bemoans PDF yet formatting control is exactly why it has such staying power despite being such a hated de facto publishing format.