The SPARCstation 20 was one heck of a great machine back in the day, especially for its size (a low profile pizzabox). The design was a lot like it's older brother (the SPARCstation 10 from 1992)... that is, two MBUS slots (for up to 4 CPUs) and 4 SBUS slots (Sun Expansion cards, 25 MHz x 64 bit = ~ 200 MB/sec each, but 400 MB/sec bus total).
I remember using a Sun evaulation model at Rice many years ago... the machine had two 150 MHz HyperSPARC processors (though 4 were avilable for more $$), a wide SCSI + fast ethernet card, two gfx cards for two monitors, and some sort of strange high speed serial card (for some oddball scanner, I think). Not to mention 512 MB of ram, in 1994! The machine was a pretty decent powerhouse and sooo small! I sort of wish the concept would have caught on, given how large modern workstations are in comparison. Heck, back then an SBUS card was about 1/3 the size of a modern 7" PCI card.
Then there's the other end of the spectrum... one department bought a Silicon Graphics Indigo2 Extrme in 1993. The gfx cardset was three full size GIO-64 cards (64 bit @ 100 MHz = about 800 MB/sec), one of which had 8 dedicated ASICs for doing geometry alone. 384 MB of RAM on that beast. Pretty wild stuff for the desktop.
Is it just me, or do most folks confuse these two. If a popular website only has a 9 Mbps pipe to the Internet, it doesn't matter how many Crays they have running their webserver farm, they're only going to be able to churn out 9 Mbps (minus overhead). Granted that the converse is possible... gobs of bandwidth, but a slow server... but I would imagine that bandwidth is the limiting factor of at least 99% of websites.
1. There is no such thing as Solaris 2.8, and 2.7 only exists in some places. I assume you meant 7 and 8.
Bah, "Solaris" is the marketing name. It's SunOS 5.8. Do a 'uname -a' if you don't belive me. Besides, it fits perfectly with SunOS history... versions prior to 5.0 were BSD based, version 5.0 and newer are heavily SysV based. The whole "Solaris" name just confused everyone.
I've been pouring over the details of this thing and something just doesn't sit right with me... the photomultiplier tubes. Why on earth would such a sensitive and crucial component be of an oldschool vacuum tube design? Did we learn nothing in the 1960s? I agree that it needs to rebuilt, but golly, use COTS (commercial, off the shelf) CCDs or similar.
It seems a lot of folks are misunderstanding the new AOL support. You see, the AirPort also has an internal modem for those without broadband. Previously, the modem could only be used to dial into a PPP account. With version 2.0 of the AirPort software, it can dial into an AOL account for Internet access (i.e., it now has a version of AOL's proprietary software in its firmware). This is not a feature that most geeks will notice, as we'll be using the AirPort's ethernet interface with our broadband connections and *maybe* the modem for PPP backup. But for those that get their Internet access via an AOL dialup, and would like to share it with their 802.11 equiped machines via NAT/DHCP, this is godsend.
I fully realize that this is talking about Covalent's Apache-based software, but I'm still wondering how ready the Apache 2.0 codebase is... I've been playing with 2.0.16 beta for awhile now on one of my test servers without and problems, but that doesn't mean diddly. I'm looking foward to verison 2.0, but not without extensive testing. Version 1.3.22 works way too well for me to make a switch anytime soon.
I can't believe the number of people implying that this guy is an idiot for needing 64GB RAM.... There are some scientific applications out there, that are extremely important to people with lots of money, where having 64GB worth of fast-access matrices/hash table/etc. will let you do experiments that are simply not possible if you are limited to 1GB
Very well put. I do a great deal of work with my university's high performance computing lab and have been exposed to this area for almost four years and continue to be amazed and blown away. By far our most demanding users are from the meteorology research group -- an organization that has been on the forefront of many new weather modeling and weather analysis applications. Their batch jobs range from crazy to insane when it comes to CPU and RAM usage and have allowed for many interesting case studies with single machines, clusters, and different platforms.
Some random observations:
Many of the jobs the meteorology folks run consume 8, 16, or even 32 GB of RAM per instance. Most jobs keep the disk arrays clattering away like a drive stress test. They've tried our cluster and tried our SunFire, but prefer the Origin 3000 (and for lesser jobs, the Origin 2000). Unlike "simple" partitionable tasks such as rendering (or SETI@home), their batch jobs don't take to clustering too well.
Many of our users put preference to total system size and maturity/support of the compilers moreso that the speed of each CPU or the newness of the software. We still have groups that prefer the Origin 2000 over anything else. One user wants us to invest in commercial fortran compilers for the Linux cluster. Many users simply prefer a single large system such as the Origins and the SunFire.
A few users continue to push the envelope. One that comes to mind has been asking for some adjustment to job quotas so that his app can utilize 128 GB of RAM. We're still not sure *what* he wants to do with that ram and are still waiting to hear back from him.
Not everyone uses their machines to surf the web, compile software, or even manipulate 256 megabyte files in Photoshop... for those that handle several terabytes of data every half hour and need to have half a terabyte or more in ram at any given time.... for those people there is truly big iron and it's available today (and yesterday). I belive we had an SGI Challenge XL with 36 R10K/195 CPUs in 1995. Two years ago we got a 128 CPU Origin 2000 and a 64 CPU SunFire. This summer we finally recieved a 256 CPU Origin 3000 with 512 GB of RAM. Big Iron costs a fortune and isn't for your highschool, your ISP, or even for your bank. But for those that *need* it, it's worth every penny.
At my university, our college of engineering bought a StorageTek Powderhorn for interdepartmental backup. The model we have currently has 100 TB of storage capacity and can be expanded to 300 TB. Its host is a massive Sun server connected to the core network switch via two gigE links and and one ATM link. At the server level in various departments and groups we are mostly doing RAID as disks have become so darn cheap. A simple script dumps data onto the Powderhorn across the street once a week in the event of a major malfunction (RAID recoveries don't always go smoothly), theft, or fire.
You'll probably want something with greater than 36 bit addressing if you're actually today going to buy a machine with 64 GB of RAM? Why? Because you'll want to do some upgrades in time. Even if you plan on replacing the beast with a cheaper, better, faster one down the road, you'll at least want to add some ram to the old one and use it as a backup or secondary machine.
A good rule of thumb is to never buy an already maxed out system. If you can afford and *need* a machine with 64 GB of RAM today, you very much need one that can handle at least 128 GB for future upgrades.
Typical big iron uses very custom components. For example, the Origin 3000 uses DDR-SDRAM chips, but on a memory module that also includes "directory ram" and some other oddball components. The O3K has also been using this ram since about 6 months before desktop PCs began using DDR, so they didn't really have a chance to embrace now-popular standards.
Sun's ram is similar. Sun uses slightly slower chips, but has many in parallel with some ungodly wide bit paths to and from the memory modules. Again, it's much different than what we have in our desktop PCs.
Keep in mind that these monster machines were designed with some insane requirements and low tolerance for error. As such, they often require much different components to keep everything working. Recall that most DDR-SDRAM based PC mobos up until the past few months shipped with two or three DIMM slots rather than four or eight because of timing and signal issues that hadn't been resolved yet. Big iron systems from Sun, IBM, and Compaq support over 100 CPUs, over 200 GB RAM, and do so without (many) problems. SGI's Origin 3000 supports up to 1 TB of RAM and 512 CPUs per single system (and several such huge machines can be clusterd via multiple 8 gigabit GSN networking connections if a task requires insane compute power).
You don't have too many options if you need 64 GB of ram in a single image (not a cluster)...
SunFire 3800
According to the techpub, you'll need both CPU trays (but perhaps with each only half populated with CPUs -- 2 + 2 = 4 CPUs) to house enough memory modules for 64 GB. Base price for a 3800 is $160,000 for two boards. Plus unique RAM (currently $1700 per GB). Keep in mind that the 3800 cannot be upgraded beyond 64 GB or 8 CPUs.
AlphaServer GS80
Looks like you'll need at least 4 CPUs to handle the 64 GB. Base price seems to be around $140,000. Plus unique RAM. Keep in mind that the GS80 cannot be upgraded beyond 64 GB or 8 CPUs.
IBM P660
2 - 8 CPUs. Up to 64 GB RAM. Starts at $66,000. Plus unique RAM. Runs the Linux-Friendly IBM AIX 5L. Keep in mind that the P660 cannot be upgraded beyond 64 GB or 8 CPUs.
SGI Origin 3000
Looking at about $220,000 for one that can accomodate 64 GB. Plus unique RAM (about $900 per GB). The 3000 can be upgraded to 1024 GB (1TB) of RAM and 512 CPUs as a supported configuration. 1 TB and 1024 CPUs unsupported and requiring unsupported OS patches.
Origin 300 would be *much* cheaper, but it only supports 32 GB right now (it will support 64 GB when SGI ships their high density memory modules) and it's nowhere near as expandable or upgradable as the 3000. Origin 300 cannot be upgraded beyond 32 CPUs and will most likely never support more than 64 GB of RAM.
I can't find an RPM of 2.4.14 in any of the usual places. Anyone know where one is available? I had problems compiling 2.4.13 last time and had to wait for an RPM.
I'd like to get this installed on the workstations and maybe some of the servers tomorrow morning.
Thanks in advance!
I love that UNIX timeline, I probably visit that page once a month. I did, however, notice that unlike most other unices on the page, the Cray UNICOS OS entry hasn't been updated to reflect recent versions. Cray has always been conservative with their numbering scheme, often heavily padding the numbers with zeros (current release of UNICOS is 10.0.1.0 with 11.0 coming soon). Would be nice to see minor updates such as with UNICOS releases reflected on the timeline as well. (UNICOS updates are no more frequent than linux kernel updates and are generally just as significant).
Even though I like OpenWindows, I almost always have my default sessions set to use CDE, since that way I can easily have my window manager the same on all the UNIX platforms I use (AIX, HP-UX, Solaris, OSF/1), with the exception of IRIX.:)
If you really want CDE for IRIX, call up SGI telesales and request "CDE 5.0" (part number SC4-CDE-5.0). It's free of charge, though you will have to pay for shipping and handling. Great for someone that wants or needs to have the exact same desktop environment on all of his machines.
I actually prefer many of the older, commercial desktop environments and window managers. These include not just CDE but vanilla Motif (mwm, the Motif Window Manager) and OpenWindows.
Why on earth would I say this, especially when GNOME is free *and* certainly looks better??
Because OpenWindows and Motif/CDE have worked. They still work. They'll most likely continue to work well into the future. And they work well with the software I use. As much as I like to fiddle and futz with GNOME (and KDE) on my Linux box at home, I'm actually glad I don't use them for work. Openwindows was fast and extremely configurable back when I used a SPARCstation 2 as my desktop workstation. Because it was running atop the X Window System, I could do far more with it than my buddies on Windows 3.1 boxes down in accounting. Several years later I moved to a SPARCstation 10 with Motif which was quite a change, but by that time most pointy-clicky gui-based software for SunOS/Solaris was becoming Motif based so the move made sense. A year later I tried CDE and found it to be a pleasant yet simple extension to the minimalist Motif/mwm desktop I had been running. Most of the time I didn't notice the difference as I had the CDE Front Panel minimized and was busy working in my own apps anyway. By early 1996 I moved to a Ultra 2, a machine that stayed with me (albeit with two new cpu modules in 1999) up until this past summer when I upgraded to a Blade 1000.
The experience has been great. Never once has the desktop gotten in the way or clashed aesthetically with Matlab, Xilinx, or any of the other tools I work with. I was never working with a piece of artwork, mind you, but it looked good enough and wasn't bothersome. OpenWindows was lean and fast back in an era when it had to be. Motif brought about some unix-wide standards (even SGI uses slightly modified Motif). In fact, I would be willing to bet that Motif has been documented moreso than even Microsoft's GUI libraries. CDE gave us a few extra goodies based on Motif. While I don't use CDE's mailer, text editor, or calculator, I do find myself relying heavily on the simple but quite useful calendar manager. These days it'll even sync with my Palm. A few other utilities are great for the little things I don't do too often (such as change the color scheme, screen saver, or fiddle with the print queue). Even the login manager is quite nice, especially for its age. Like many of the newer freeware options (gdm, kdm, etc) it allows the user to select a desktop environment at login. Quite handy when trying out a few other the up and coming environments (I tried Sun's version of GNOME 1.4 several months ago).
I say all of this as a combination hardware and software engineer, mostly working in the embedded RISC world. I'm not a sysadmin with every script, utility, flag, and manpage memorized. Nor am I a graphic artist with a need for a PowerMac and its Postscript, ColorSync, and FontSync. I'm just a guy that needs to get real work done on a platform that's both flexible and not going to give me problems. Thankfully the past 11 years have be extremely nice to me. Of the 5 workstations I've used, each has had it's OS installed only once (with the exception of the Ultra 2 which I upgraded from Solaris 2.5.1 to Solaris 7 [2.7]). I only had to install my software packages once. Of all this, the gui toolkits and windowmanagers played a very small part. But they played that part with exemplary performance. They weren't wiz-bang, but they weren't a moving target either. They did their job - well. As for me, I am going to continue working with my current setup. I don't need the toys and whistles while I'm at work and thus will continue using CDE. I'll let my two UltraSPARC III CPUs spend their time working on my code.
That said, I'm glad Sun has an open mind and is working with GNOME. I personally don't think GNOME (or KDE) is the long term answer, but at least they're looking in other directions. Motif and CDE are old, but well used and well documented. For many, it's time to move on. Lets do so with some common sense and a historical perspective.
Calculus and alcohol don't mix. Never drink and derive.
Hey now, OxyClean works pretty well (too good to be As Seen On TV). I've been using it to clean up carpet stains after parties, it works great, but if you mix it too thick, you'll end up with undisolved particles sticking to everything. So, I wouldn't recommended "making a paste, make it 10 times more parrrful" unless you *really* need to.
I too bought an Internal Antenna. I think it's helping by about 1 bar on average (may be a placebo effect). Not quite the miracles that the chiropractor woman on the TV commercial claimed, but good enough.
HD resolution is nothing new to SGI. "HDTV" support is. We had 1280x1024 and 1920x1080 demo reels driven by an SGI Onyx RE1 as far back as 1993 (back when our station's most powerful PC was a 66 MHz 486). We had a rackmount gizmo called a RaptorXL that could convert the high resolution RGB signal into a variety of (at that time) currently proposed HD standards. With no usable HD transmission equipment, nor consumer HD television sets, we didn't use it for much other than openhouses and tradeshows. Our Onyx2 from early 1997 came configured with an HD I/O card that worked well with an early Sony HDcam deck. It required an adapter, but other than that we had no conflicts. O2 is a *wonderful* standard resolution video box (especially for it's age, it came out in 1996) but I doubt O2 will ever support HD, at least for editing or effects. While it could easily record and play HD video, it lacks the I/O to handle multiple internal and external 200+ MB streams of data. The nonblocking crossbar switch in the Octane/Octane2 handles it without flinching.
Octane2 (and Octane with an Xbow 1.4 backplane) supports HD video via the "snowball" DM2 card and breakout box:
http://www.sgi.com/workstations/octane2/dig_medi a. html
Like most broadcast products, it only supports noncompressed video, so you'll also need at least two channels of Ultra160 SCSI or three channels of 1gbit fibrechannel. You *do not* want to work with compressed HD until you're finished with editing and compositing, otherwise the CPU and daughtercards spend 75% of their power compressing and decompressing every time you make a slight change.
We use an Octane2 daily for HD editing and mixing. Source is a Sony HDcam deck controlled by and uplinked to the DM2 breakout box. Video is stored on 8 rackmount Ciprico 7000 fibre disk arrays (RAID 3). It's certainly not consumer, but it works without a hitch. And it's fast.
What sort of compression are they using?? Even the most optimized Divx codecs aren't that good yet, are they? Say the stream is vanilla 1080i, that's 1920 x 1080 x 16 bit color x 30fps = about 124 MB/sec uncompressed or about 995 Mbps uncompressed. Not including 5.1 or 6.1 channel audio. Compressing that down to a 20 Mbps stream would require a 50:1 codec for the video alone. If it's 24 bit color, the resultant size would be 50% larger and thus require a 75:1 codec.
So what sorta compression codecs are they using? Please excuse my ignorance as I have never worked with compressed HDTV. My only experience has been on the production end where we used SGI Octane2 and Onyx2 systems with the Snowball DM2/DM3 I/O cards fo uncompressed work. Of course, that required a full rack of Ciprico fibre disk arrays to store the data, but the quality was awesome. On the software end, we used IFX's 'Piranha HD' and Discreet's 'Inferno'.
Emulation.net rocks! Not only has John Stiles put together a great website, he has also done many of the ports to Mac OS and has even written a few unique games (such as Candy Crisis - http://www.candycrisis.com).
Rackmy.com seems to have some pretty good prices. Be sure to calculate your bandwidth costs (in GB/month) and the rack space costs (they charge per rack unit... 1U = 1.75" tall). Could be pretty costly unless you have a slimline server with low network usage.
From the looks of this thread, colocation has been on the minds of quite a few slashdoters. For those that haven't dealt with the wonders (and horrors) of colocation, here are some pointers from my personal experience:
1) "Mom and Pop" vs "Brass and Glass"...
The huge datacenter colo facilities make any geek drool and may even be a requirement for the hosting of a site requiring tremendous bandwidth and security, but are they what you need? For my business's servers we initially went with a mega-huge provider and were very dissapointed. "24-hour tech support" and "24-hour staffed datacenter" often required a daytime call back to reach a tech with any sort of skill, or more often, authority to make changes. Sales departments were often a nightmare to deal with. Change orders would take weeks for the paperwork to reach the proper people. We eventually tried another provider before finally switching to a small local internet services group that offered colocation at their office just 90 miles from here. Their rates are a good deal cheaper, their bandwidth is sufficent, and the ability to call their real techs at any hour of the day has been a godsend. Never again will I do business with the big boys when I can do business with a real person.
2) Bandwidth Usage...
Some providers have fixed rates for your monthly usage (often expressed in GB/month), other have a bandwidth cap (so many Kbps/sec). Often you can pay for more bandwidth or arrange to be billed depending on the month's usage. Understand what you're getting into. 10 GB/month might sound like a lot, but that's just 2500 MB/week. More than enough for most sites, but if you plan on shuffling a lot of data to your server every few weeks it may not be enough. Obviously you're not going to have a solid sustained rate over the course of a month, but if you were, that 10 GB/month would be just 32 kbps over a month's time. Again, this has no meaning for the average bursty website, it's just that a 33.6 kbps modem can transfer about 10 GB per month if it were dialed up 24-7. A major advantage to many "Mom and Pop" colo providers is that they have a very lax bandwidth policy. Typically if you stay away from kiddy porn, warez, and spamming, they won't care how much bandwidth you use just as long as it's not a tremendous amount for a tremendous period of time. My provider and I agreed to discuss any major changes and thus far things have worked out well. I have no bandwidth cap or monthly limit. Some months we use 25 GB other months it's 75. My usage is very bursty (new product releases stir up a lot of hits for a few days) but overall it's a general few-hits-a-second "noise" traffic. However, I'm quite certain that if I were to set up a huge FTP site on my server that would generate massive traffic around the clock, my ISP would call at the end of the month to request that we make changes for the following month.
3) The server itself...
Be upfont with potential providers with the size of your server. It does make a difference. A 1U rackmount server takes up much less space than a full tower ATX case. The size often has a factor on the price. Some providers don't care. It's best to ask right away. Unless your provider is nearby, you'll probably be shipping your server to him. For your benefit and his, enclose information with the server, including your name and email address along with any important setup instructions. **Test your server before shipping it** Be sure that it'll work properly without the usual I/O accessories hanging off it. Most providers won't plug every single server into a massive KVM switch. If your server does not work without a keyboard attached, then let the provider know and enclose something that can be plugged in and hang behind the server. Most providers will provide you with network config data (IP address, etc). You can attempt to configure this before you ship the server, but it may be best to leave it in the config that works in your home/office and include simple instructions for the provider to change the IP address for you. This prevents nasty config surprises. Again, ask the provider for more info on this.
From my experience and having talked with others, I would highly recommend a qualified Mom & Pop provider over the brass and glass type... unless you do need the level of security offered by the hardcore pros. If there is one thing I can stress, it's to ASK QUESTIONS! And take note of how many people you need to talk to for an answer. You should be able to get a solid price quote from no more than two emails. You should be able to talk with a real tech (someone that maintains the network and works with the hardware) without having to talk to several people inbetween. Basicly, if you can't get ahold of these important people now, as a business prospect, then why expect to get ahold of them at 4am when everything's going wrong? Again, ask questions and be somewhat patient. The big boys often have untrained people answering the phones and Mom & Pop are often busy during key hours of the day. Don't be afraid to send some emails, fax some questions, or leave a voice mail.
Still, would the comparison change drastically when OS X is ready for primetime?
MS recently completed the Mac OS X version of Office (Office v.X) and it should be shipping soon. If that isn't a sign of OS X being ready for primetime, then I don't know what is.
In related news, Apple is gearing up to release Mac OS X 10.1.1, a 0.0.1 point release to address a few minor issues. OS X is looking better all the time.
Slashdot Mirror
The SPARCstation 20 was one heck of a great machine back in the day, especially for its size (a low profile pizzabox). The design was a lot like it's older brother (the SPARCstation 10 from 1992)... that is, two MBUS slots (for up to 4 CPUs) and 4 SBUS slots (Sun Expansion cards, 25 MHz x 64 bit = ~ 200 MB/sec each, but 400 MB/sec bus total).
I remember using a Sun evaulation model at Rice many years ago... the machine had two 150 MHz HyperSPARC processors (though 4 were avilable for more $$), a wide SCSI + fast ethernet card, two gfx cards for two monitors, and some sort of strange high speed serial card (for some oddball scanner, I think). Not to mention 512 MB of ram, in 1994! The machine was a pretty decent powerhouse and sooo small! I sort of wish the concept would have caught on, given how large modern workstations are in comparison. Heck, back then an SBUS card was about 1/3 the size of a modern 7" PCI card.
Then there's the other end of the spectrum... one department bought a Silicon Graphics Indigo2 Extrme in 1993. The gfx cardset was three full size GIO-64 cards (64 bit @ 100 MHz = about 800 MB/sec), one of which had 8 dedicated ASICs for doing geometry alone. 384 MB of RAM on that beast. Pretty wild stuff for the desktop.
Ahh, technology. I love you!
Is it just me, or do most folks confuse these two. If a popular website only has a 9 Mbps pipe to the Internet, it doesn't matter how many Crays they have running their webserver farm, they're only going to be able to churn out 9 Mbps (minus overhead). Granted that the converse is possible... gobs of bandwidth, but a slow server... but I would imagine that bandwidth is the limiting factor of at least 99% of websites.
1. There is no such thing as Solaris 2.8, and 2.7 only exists in some places. I assume you meant 7 and 8.
Bah, "Solaris" is the marketing name. It's SunOS 5.8. Do a 'uname -a' if you don't belive me. Besides, it fits perfectly with SunOS history... versions prior to 5.0 were BSD based, version 5.0 and newer are heavily SysV based. The whole "Solaris" name just confused everyone.
I've been pouring over the details of this thing and something just doesn't sit right with me... the photomultiplier tubes. Why on earth would such a sensitive and crucial component be of an oldschool vacuum tube design? Did we learn nothing in the 1960s? I agree that it needs to rebuilt, but golly, use COTS (commercial, off the shelf) CCDs or similar.
It seems a lot of folks are misunderstanding the new AOL support. You see, the AirPort also has an internal modem for those without broadband. Previously, the modem could only be used to dial into a PPP account. With version 2.0 of the AirPort software, it can dial into an AOL account for Internet access (i.e., it now has a version of AOL's proprietary software in its firmware). This is not a feature that most geeks will notice, as we'll be using the AirPort's ethernet interface with our broadband connections and *maybe* the modem for PPP backup. But for those that get their Internet access via an AOL dialup, and would like to share it with their 802.11 equiped machines via NAT/DHCP, this is godsend.
I fully realize that this is talking about Covalent's Apache-based software, but I'm still wondering how ready the Apache 2.0 codebase is... I've been playing with 2.0.16 beta for awhile now on one of my test servers without and problems, but that doesn't mean diddly. I'm looking foward to verison 2.0, but not without extensive testing. Version 1.3.22 works way too well for me to make a switch anytime soon.
I can't believe the number of people implying that this guy is an idiot for needing 64GB RAM.... There are some scientific applications out there, that are extremely important to people with lots of money, where having 64GB worth of fast-access matrices/hash table/etc. will let you do experiments that are simply not possible if you are limited to 1GB
Very well put. I do a great deal of work with my university's high performance computing lab and have been exposed to this area for almost four years and continue to be amazed and blown away. By far our most demanding users are from the meteorology research group -- an organization that has been on the forefront of many new weather modeling and weather analysis applications. Their batch jobs range from crazy to insane when it comes to CPU and RAM usage and have allowed for many interesting case studies with single machines, clusters, and different platforms.
Some random observations:
Many of the jobs the meteorology folks run consume 8, 16, or even 32 GB of RAM per instance. Most jobs keep the disk arrays clattering away like a drive stress test. They've tried our cluster and tried our SunFire, but prefer the Origin 3000 (and for lesser jobs, the Origin 2000). Unlike "simple" partitionable tasks such as rendering (or SETI@home), their batch jobs don't take to clustering too well.
Many of our users put preference to total system size and maturity/support of the compilers moreso that the speed of each CPU or the newness of the software. We still have groups that prefer the Origin 2000 over anything else. One user wants us to invest in commercial fortran compilers for the Linux cluster. Many users simply prefer a single large system such as the Origins and the SunFire.
A few users continue to push the envelope. One that comes to mind has been asking for some adjustment to job quotas so that his app can utilize 128 GB of RAM. We're still not sure *what* he wants to do with that ram and are still waiting to hear back from him.
Not everyone uses their machines to surf the web, compile software, or even manipulate 256 megabyte files in Photoshop... for those that handle several terabytes of data every half hour and need to have half a terabyte or more in ram at any given time.... for those people there is truly big iron and it's available today (and yesterday). I belive we had an SGI Challenge XL with 36 R10K/195 CPUs in 1995. Two years ago we got a 128 CPU Origin 2000 and a 64 CPU SunFire. This summer we finally recieved a 256 CPU Origin 3000 with 512 GB of RAM. Big Iron costs a fortune and isn't for your highschool, your ISP, or even for your bank. But for those that *need* it, it's worth every penny.
At my university, our college of engineering bought a StorageTek Powderhorn for interdepartmental backup. The model we have currently has 100 TB of storage capacity and can be expanded to 300 TB. Its host is a massive Sun server connected to the core network switch via two gigE links and and one ATM link. At the server level in various departments and groups we are mostly doing RAID as disks have become so darn cheap. A simple script dumps data onto the Powderhorn across the street once a week in the event of a major malfunction (RAID recoveries don't always go smoothly), theft, or fire.
You'll probably want something with greater than 36 bit addressing if you're actually today going to buy a machine with 64 GB of RAM? Why? Because you'll want to do some upgrades in time. Even if you plan on replacing the beast with a cheaper, better, faster one down the road, you'll at least want to add some ram to the old one and use it as a backup or secondary machine.
A good rule of thumb is to never buy an already maxed out system. If you can afford and *need* a machine with 64 GB of RAM today, you very much need one that can handle at least 128 GB for future upgrades.
Typical big iron uses very custom components. For example, the Origin 3000 uses DDR-SDRAM chips, but on a memory module that also includes "directory ram" and some other oddball components. The O3K has also been using this ram since about 6 months before desktop PCs began using DDR, so they didn't really have a chance to embrace now-popular standards.
Sun's ram is similar. Sun uses slightly slower chips, but has many in parallel with some ungodly wide bit paths to and from the memory modules. Again, it's much different than what we have in our desktop PCs.
Keep in mind that these monster machines were designed with some insane requirements and low tolerance for error. As such, they often require much different components to keep everything working. Recall that most DDR-SDRAM based PC mobos up until the past few months shipped with two or three DIMM slots rather than four or eight because of timing and signal issues that hadn't been resolved yet. Big iron systems from Sun, IBM, and Compaq support over 100 CPUs, over 200 GB RAM, and do so without (many) problems. SGI's Origin 3000 supports up to 1 TB of RAM and 512 CPUs per single system (and several such huge machines can be clusterd via multiple 8 gigabit GSN networking connections if a task requires insane compute power).
You don't have too many options if you need 64 GB of ram in a single image (not a cluster)...
SunFire 3800
According to the techpub, you'll need both CPU trays (but perhaps with each only half populated with CPUs -- 2 + 2 = 4 CPUs) to house enough memory modules for 64 GB. Base price for a 3800 is $160,000 for two boards. Plus unique RAM (currently $1700 per GB). Keep in mind that the 3800 cannot be upgraded beyond 64 GB or 8 CPUs.
AlphaServer GS80
Looks like you'll need at least 4 CPUs to handle the 64 GB. Base price seems to be around $140,000. Plus unique RAM. Keep in mind that the GS80 cannot be upgraded beyond 64 GB or 8 CPUs.
IBM P660
2 - 8 CPUs. Up to 64 GB RAM. Starts at $66,000. Plus unique RAM. Runs the Linux-Friendly IBM AIX 5L. Keep in mind that the P660 cannot be upgraded beyond 64 GB or 8 CPUs.
SGI Origin 3000
Looking at about $220,000 for one that can accomodate 64 GB. Plus unique RAM (about $900 per GB). The 3000 can be upgraded to 1024 GB (1TB) of RAM and 512 CPUs as a supported configuration. 1 TB and 1024 CPUs unsupported and requiring unsupported OS patches.
Origin 300 would be *much* cheaper, but it only supports 32 GB right now (it will support 64 GB when SGI ships their high density memory modules) and it's nowhere near as expandable or upgradable as the 3000. Origin 300 cannot be upgraded beyond 32 CPUs and will most likely never support more than 64 GB of RAM.
IBM is your best best.
I can't find an RPM of 2.4.14 in any of the usual places. Anyone know where one is available? I had problems compiling 2.4.13 last time and had to wait for an RPM.
I'd like to get this installed on the workstations and maybe some of the servers tomorrow morning.
Thanks in advance!
I love that UNIX timeline, I probably visit that page once a month. I did, however, notice that unlike most other unices on the page, the Cray UNICOS OS entry hasn't been updated to reflect recent versions. Cray has always been conservative with their numbering scheme, often heavily padding the numbers with zeros (current release of UNICOS is 10.0.1.0 with 11.0 coming soon). Would be nice to see minor updates such as with UNICOS releases reflected on the timeline as well. (UNICOS updates are no more frequent than linux kernel updates and are generally just as significant).
How long has 2.2.20 been in -pre state? Almost 8 months?
At any rate, it's a welcome sight. Several of our servers are still running 2.2, though most get a good dose of kernel.org and apt-get every few days.
... about running 2.2.19 (on RedHat 6.2) on my dual PII server. Guess it's about time to upgrade, though. RedHat 7.2 is looking nice.
How is bigmem support coming along? Is 2.4 still having problems with (32-bit) systems sporting more than 2 GB of ram?
Even though I like OpenWindows, I almost always have my default sessions set to use CDE, since that way I can easily have my window manager the same on all the UNIX platforms I use (AIX, HP-UX, Solaris, OSF/1), with the exception of IRIX. :)
If you really want CDE for IRIX, call up SGI telesales and request "CDE 5.0" (part number SC4-CDE-5.0). It's free of charge, though you will have to pay for shipping and handling. Great for someone that wants or needs to have the exact same desktop environment on all of his machines.
Prepare to lose all karma.
I actually prefer many of the older, commercial desktop environments and window managers. These include not just CDE but vanilla Motif (mwm, the Motif Window Manager) and OpenWindows.
Why on earth would I say this, especially when GNOME is free *and* certainly looks better??
Because OpenWindows and Motif/CDE have worked. They still work. They'll most likely continue to work well into the future. And they work well with the software I use. As much as I like to fiddle and futz with GNOME (and KDE) on my Linux box at home, I'm actually glad I don't use them for work. Openwindows was fast and extremely configurable back when I used a SPARCstation 2 as my desktop workstation. Because it was running atop the X Window System, I could do far more with it than my buddies on Windows 3.1 boxes down in accounting. Several years later I moved to a SPARCstation 10 with Motif which was quite a change, but by that time most pointy-clicky gui-based software for SunOS/Solaris was becoming Motif based so the move made sense. A year later I tried CDE and found it to be a pleasant yet simple extension to the minimalist Motif/mwm desktop I had been running. Most of the time I didn't notice the difference as I had the CDE Front Panel minimized and was busy working in my own apps anyway. By early 1996 I moved to a Ultra 2, a machine that stayed with me (albeit with two new cpu modules in 1999) up until this past summer when I upgraded to a Blade 1000.
The experience has been great. Never once has the desktop gotten in the way or clashed aesthetically with Matlab, Xilinx, or any of the other tools I work with. I was never working with a piece of artwork, mind you, but it looked good enough and wasn't bothersome. OpenWindows was lean and fast back in an era when it had to be. Motif brought about some unix-wide standards (even SGI uses slightly modified Motif). In fact, I would be willing to bet that Motif has been documented moreso than even Microsoft's GUI libraries. CDE gave us a few extra goodies based on Motif. While I don't use CDE's mailer, text editor, or calculator, I do find myself relying heavily on the simple but quite useful calendar manager. These days it'll even sync with my Palm. A few other utilities are great for the little things I don't do too often (such as change the color scheme, screen saver, or fiddle with the print queue). Even the login manager is quite nice, especially for its age. Like many of the newer freeware options (gdm, kdm, etc) it allows the user to select a desktop environment at login. Quite handy when trying out a few other the up and coming environments (I tried Sun's version of GNOME 1.4 several months ago).
I say all of this as a combination hardware and software engineer, mostly working in the embedded RISC world. I'm not a sysadmin with every script, utility, flag, and manpage memorized. Nor am I a graphic artist with a need for a PowerMac and its Postscript, ColorSync, and FontSync. I'm just a guy that needs to get real work done on a platform that's both flexible and not going to give me problems. Thankfully the past 11 years have be extremely nice to me. Of the 5 workstations I've used, each has had it's OS installed only once (with the exception of the Ultra 2 which I upgraded from Solaris 2.5.1 to Solaris 7 [2.7]). I only had to install my software packages once. Of all this, the gui toolkits and windowmanagers played a very small part. But they played that part with exemplary performance. They weren't wiz-bang, but they weren't a moving target either. They did their job - well. As for me, I am going to continue working with my current setup. I don't need the toys and whistles while I'm at work and thus will continue using CDE. I'll let my two UltraSPARC III CPUs spend their time working on my code.
That said, I'm glad Sun has an open mind and is working with GNOME. I personally don't think GNOME (or KDE) is the long term answer, but at least they're looking in other directions. Motif and CDE are old, but well used and well documented. For many, it's time to move on. Lets do so with some common sense and a historical perspective.
Calculus and alcohol don't mix. Never drink and derive.
Hey now, OxyClean works pretty well (too good to be As Seen On TV). I've been using it to clean up carpet stains after parties, it works great, but if you mix it too thick, you'll end up with undisolved particles sticking to everything. So, I wouldn't recommended "making a paste, make it 10 times more parrrful" unless you *really* need to.
I too bought an Internal Antenna. I think it's helping by about 1 bar on average (may be a placebo effect). Not quite the miracles that the chiropractor woman on the TV commercial claimed, but good enough.
HD resolution is nothing new to SGI. "HDTV" support is. We had 1280x1024 and 1920x1080 demo reels driven by an SGI Onyx RE1 as far back as 1993 (back when our station's most powerful PC was a 66 MHz 486). We had a rackmount gizmo called a RaptorXL that could convert the high resolution RGB signal into a variety of (at that time) currently proposed HD standards. With no usable HD transmission equipment, nor consumer HD television sets, we didn't use it for much other than openhouses and tradeshows. Our Onyx2 from early 1997 came configured with an HD I/O card that worked well with an early Sony HDcam deck. It required an adapter, but other than that we had no conflicts. O2 is a *wonderful* standard resolution video box (especially for it's age, it came out in 1996) but I doubt O2 will ever support HD, at least for editing or effects. While it could easily record and play HD video, it lacks the I/O to handle multiple internal and external 200+ MB streams of data. The nonblocking crossbar switch in the Octane/Octane2 handles it without flinching.
i a. html
Octane2 (and Octane with an Xbow 1.4 backplane) supports HD video via the "snowball" DM2 card and breakout box:
http://www.sgi.com/workstations/octane2/dig_med
Like most broadcast products, it only supports noncompressed video, so you'll also need at least two channels of Ultra160 SCSI or three channels of 1gbit fibrechannel. You *do not* want to work with compressed HD until you're finished with editing and compositing, otherwise the CPU and daughtercards spend 75% of their power compressing and decompressing every time you make a slight change.
We use an Octane2 daily for HD editing and mixing. Source is a Sony HDcam deck controlled by and uplinked to the DM2 breakout box. Video is stored on 8 rackmount Ciprico 7000 fibre disk arrays (RAID 3). It's certainly not consumer, but it works without a hitch. And it's fast.
What sort of compression are they using?? Even the most optimized Divx codecs aren't that good yet, are they? Say the stream is vanilla 1080i, that's 1920 x 1080 x 16 bit color x 30fps = about 124 MB/sec uncompressed or about 995 Mbps uncompressed. Not including 5.1 or 6.1 channel audio. Compressing that down to a 20 Mbps stream would require a 50:1 codec for the video alone. If it's 24 bit color, the resultant size would be 50% larger and thus require a 75:1 codec.
So what sorta compression codecs are they using? Please excuse my ignorance as I have never worked with compressed HDTV. My only experience has been on the production end where we used SGI Octane2 and Onyx2 systems with the Snowball DM2/DM3 I/O cards fo uncompressed work. Of course, that required a full rack of Ciprico fibre disk arrays to store the data, but the quality was awesome. On the software end, we used IFX's 'Piranha HD' and Discreet's 'Inferno'.
Emulation.net rocks! Not only has John Stiles put together a great website, he has also done many of the ports to Mac OS and has even written a few unique games (such as Candy Crisis - http://www.candycrisis.com).
Rackmy.com seems to have some pretty good prices. Be sure to calculate your bandwidth costs (in GB/month) and the rack space costs (they charge per rack unit... 1U = 1.75" tall). Could be pretty costly unless you have a slimline server with low network usage.
From the looks of this thread, colocation has been on the minds of quite a few slashdoters. For those that haven't dealt with the wonders (and horrors) of colocation, here are some pointers from my personal experience:
1) "Mom and Pop" vs "Brass and Glass"...
The huge datacenter colo facilities make any geek drool and may even be a requirement for the hosting of a site requiring tremendous bandwidth and security, but are they what you need? For my business's servers we initially went with a mega-huge provider and were very dissapointed. "24-hour tech support" and "24-hour staffed datacenter" often required a daytime call back to reach a tech with any sort of skill, or more often, authority to make changes. Sales departments were often a nightmare to deal with. Change orders would take weeks for the paperwork to reach the proper people. We eventually tried another provider before finally switching to a small local internet services group that offered colocation at their office just 90 miles from here. Their rates are a good deal cheaper, their bandwidth is sufficent, and the ability to call their real techs at any hour of the day has been a godsend. Never again will I do business with the big boys when I can do business with a real person.
2) Bandwidth Usage...
Some providers have fixed rates for your monthly usage (often expressed in GB/month), other have a bandwidth cap (so many Kbps/sec). Often you can pay for more bandwidth or arrange to be billed depending on the month's usage. Understand what you're getting into. 10 GB/month might sound like a lot, but that's just 2500 MB/week. More than enough for most sites, but if you plan on shuffling a lot of data to your server every few weeks it may not be enough. Obviously you're not going to have a solid sustained rate over the course of a month, but if you were, that 10 GB/month would be just 32 kbps over a month's time. Again, this has no meaning for the average bursty website, it's just that a 33.6 kbps modem can transfer about 10 GB per month if it were dialed up 24-7. A major advantage to many "Mom and Pop" colo providers is that they have a very lax bandwidth policy. Typically if you stay away from kiddy porn, warez, and spamming, they won't care how much bandwidth you use just as long as it's not a tremendous amount for a tremendous period of time. My provider and I agreed to discuss any major changes and thus far things have worked out well. I have no bandwidth cap or monthly limit. Some months we use 25 GB other months it's 75. My usage is very bursty (new product releases stir up a lot of hits for a few days) but overall it's a general few-hits-a-second "noise" traffic. However, I'm quite certain that if I were to set up a huge FTP site on my server that would generate massive traffic around the clock, my ISP would call at the end of the month to request that we make changes for the following month.
3) The server itself...
Be upfont with potential providers with the size of your server. It does make a difference. A 1U rackmount server takes up much less space than a full tower ATX case. The size often has a factor on the price. Some providers don't care. It's best to ask right away. Unless your provider is nearby, you'll probably be shipping your server to him. For your benefit and his, enclose information with the server, including your name and email address along with any important setup instructions. **Test your server before shipping it** Be sure that it'll work properly without the usual I/O accessories hanging off it. Most providers won't plug every single server into a massive KVM switch. If your server does not work without a keyboard attached, then let the provider know and enclose something that can be plugged in and hang behind the server. Most providers will provide you with network config data (IP address, etc). You can attempt to configure this before you ship the server, but it may be best to leave it in the config that works in your home/office and include simple instructions for the provider to change the IP address for you. This prevents nasty config surprises. Again, ask the provider for more info on this.
From my experience and having talked with others, I would highly recommend a qualified Mom & Pop provider over the brass and glass type... unless you do need the level of security offered by the hardcore pros. If there is one thing I can stress, it's to ASK QUESTIONS! And take note of how many people you need to talk to for an answer. You should be able to get a solid price quote from no more than two emails. You should be able to talk with a real tech (someone that maintains the network and works with the hardware) without having to talk to several people inbetween. Basicly, if you can't get ahold of these important people now, as a business prospect, then why expect to get ahold of them at 4am when everything's going wrong? Again, ask questions and be somewhat patient. The big boys often have untrained people answering the phones and Mom & Pop are often busy during key hours of the day. Don't be afraid to send some emails, fax some questions, or leave a voice mail.
Still, would the comparison change drastically when OS X is ready for primetime?
MS recently completed the Mac OS X version of Office (Office v.X) and it should be shipping soon. If that isn't a sign of OS X being ready for primetime, then I don't know what is.
In related news, Apple is gearing up to release Mac OS X 10.1.1, a 0.0.1 point release to address a few minor issues. OS X is looking better all the time.
and MacPerl is available as well
MacPerl is only for "Classic" Mac OS (Mac OS 9.2.1 and earlier). Mac OS X has real Perl.