Zing - A JVM for virtualized x86 platforms with pauseless GC
azulsystems.com
azulsystems.com
I also think their use cases are either a joke or written by a fool. "Scary for Apps"? Seriously? http://www.azulsystems.com/products/zing/use-cases Also makes a lot of bold statements that it doesn't even try to back up with numbers like "Better application performance under hypervisors than if run natively".
My scam sense tells me to run for the hills just from that one page. I hope for their sake that they redo that page in a hurry.
Keep this in mind: most companies that sell to enterprise and service provider customers aren't looking to make money with their website. Their SALESPEOPLE are there to properly qualify the leads and demonstrate the value of the product.
The whole point of not listing prices is to get customers to self-qualify. Those that genuinely find value in the products as presented will make the effort to contact the company. Price is really only something that is mentioned towards the end of the sale and indicates that there is considerable room for discounting from list price.
THAT's how it's done.
Sun was selling to a particular customer that demanded rock solid stability.
If Linux had never happened, Dell servers wouldn't have made a dent in Sun's marketshare. Don't count out Windows either. They both ate away at the BOTTOM of the market for a while.
But neither of those have anything to do with a sales model.
If your hardware lists for $1M+ per box, you don't sell that with a webpage. You sell it after many many visits with the customerto be sure that this is really what they need and that they will be able to use it.
Go look for a price for the CRS-1/8 on Cisco's webpage and tell me if you see it. Same for Juniper.
The appearance of a price on a webpage is a clear indication that you will never speak to anyone at that company who understands your business or requirements.....EVER.
I'm not Boeing's target market. It doesn't surprise me that they aren't going out of their way to post prices of 777 engines for my perusal.
In the end, you got the product you wanted (cheap Dell) and a business got a customer they wanted (dude with small budget).
Sure they did. I have personal experience with a company that dumped Sun/Solaris to move to Linux servers in the last 5 or 6 years.
The OP made a comment about web pages with pricing. My point was that lots of companies target customers that are not THEM. In fact, they often choose a strategy that allows them to explicitly target those customers that they think are their most likely prospects. Sometimes they get it right; sometimes they don't. Like a lot of things that drive company success, picking the right way to reach customers is an important decision.
Don't forget that Sun was purchased by Oracle for a princely sum. A lot of those Sun customers didn't seem to feel that their dollars were wasted.
You do realize that Sun did have pricing for almost their entire line of servers and workstations on their website years ago, right? I distinctly remember playing with their configurator to put together a $1.6M+ many-processor system many years ago - if only my credit limit had been higher…
When we wanted a confit from Sun, we called our account manager and they did a config/quote for us. We also never paid list pricing.
Not just a page for pricing, they had an entire webapp for configuring and ordering systems. You could go in, select a server or workstation, configure it (CPUs, memory, disk, etc...), with the price updating as you went, and (IIRC) purchase it online. Not all at dissimilar to the Dell online purchasing process, actually.
This was around ten years ago, and they removed the capability not too long after that. That was clearly an attempt to diversify their sales process away from just the salesman, but for whatever reason it failed and they gave up.
List price or no, they obviously weren't content only having the process that you insist is just how it's done.
AAPL tried to license their platform for a little while. That died too.
My point (to get back to the original discussion) is that selling $1M+/unit equipment is fundamentally different from selling boxes that cost $2K. The OP shouldn't be surprised that the price isn't listed b/c that probably indicates that the whole pricing exercise isn't a neat process and involves more $$$ that they would be willing to part with.
For a lot of other organizations, price is (very nearly) no object since they are often the only suitable customer for whole classes of product. Since they're the market, THEY set the price. If you're a Telco or CableCo, you are squarely within the market for high-end $1M+ routers. If you're a financial exchange, you're in the market for ultra-high performance OLTP systems.
Lockheed doesn't sell fighter jets to school systems. I doubt a vice principal would find it insulting that they couldn't order one via a webpage. That's what "qualification" is: determining whether a customer has a use for your product and the budget (more or less) to purchase it. Qualification isn't about making snap judgments about a customer; it's about making sure you're allocating finite sales/engineering resources across the most likely purchasers.
Having said that, prices would be nice to see.
When I used to work with JVM's all day I was always frustrated about having to decide how big a maximum heap for a given app would be. We had some xml processing tasks that almost never used much memory but every now and then would need tons. It was one part I couldn't thin-provision.
Am I right in interpreting this as shipping the JVM work off to a dedicated appliance-like server? Does that mean the memory profile on the appliance is the one that can be thin-provisioned?
Heap size tuning is one of those things that made me hate the JVM.
My impression is that it's a copying collector that works on pages at a time. When a page is relocated, it's marked protected in the memory address mappings, allowing the runtime to trap and fix any stale pointers at the moment they are used. At the same time, the collector proper uses memory fences to keep track of what pointers have been fixed, eventually allowing the protected pages to be released from the memory map.
I don't know GC literature very well, but AFAIK this design is novell. The original implementation made heavy use of high performance memory primitives supplied by the Azul hardware. I gather they now have found a way of running with acceptable performance on commodity x86 processors under the virtualization.
In any case, you're free to mess with (virtualised) page tables when the parent is a hypervisor.
While I'm not a systems programmer, from what I've read (particularly from Poul-Henning Kamp) I think enriching the virtual memory abstraction is going to be a key area for OS innovation shortly. MMU's and the signaling mechanisms offer a unique ability to trap events happening dynamically within your program. I think by revising the userland/kernel API, I think there's likely many interesting ways for userland software to better utilize the MMU's in modern processors.