I think you may be underestimating how hard this actually is...
How many different power grids feed your company?
Redundant networking is well understood, so is the engineering required to bring in that networking via multiple trenches and facility penetration points so that you are protected against someone cutting a line or opening a trench accidentally with heavy equipment.
This is much harder for electricity.
At some point, probably pretty close to your facility you are going to be drawing from the same grid / same source. Whats the point of multiple trenches for power when you are drawing off the same grid or transformer a short distance away? You still have a potential single point of failure. This is why the facilities invest in local power generation and backup rather than diverse feeds which may be impossible or impractical to obtain.
The only reason I know this is that I've got a colocation cage rented at a Boston area datacenter that through an odd stroke of history actually IS connected simultaneously to 2 different utility power grids. The people who work there love to point out how rare it is to find a building that is attached to completely different upstream power sources.
I'm in the same boat. Had a 15 inch powerbook for years and my first cardbus product was the Verizon EVDO high speed wireless card which I bought about 10 days ago:)
Works great but is it obsolete when I upgrade (which will be soon as I hate having a biz notebook that is out of applecare coverage)?
Google hints of compatability options or adaptor cards but I can't find solid info. Hints, tips and URLS appreciated!
{ excerpting from my own reply made in a different section of this article... }
There are many people posting here who are completely confusing what the word "cluster" means for this particular question.
This article is about APPLICATION CLUSTERING (in this case a very specific relational database) and you are answering the question with information that is generalized to a COMPUTE FARM or a Linux cluster built and optimized for high performance computing.
Broadly speaking the word "cluster" means different things to different people.
There are people who build speedy Linux Clusters where the focus is on High Performance or High Throughput Computing. Then there are people who build ultra-reliable services (High Availibility Clustering).
The difference between "high performance", "high throughput" and "high availibility" clusters can be extreme.
For this particular Slashdot article the only people posting solutions or suggestions should be people who have actually used the RAC product in a production environment. Everyone else just needs to sit back and watch...
There are many people posting here who are completely confusing what the word "cluster" means for this particular question.
This article is about APPLICATION CLUSTERING (in this case a very specific relational database) and you are answering the question with information that is generalized to a COMPUTE FARM or a Linux cluster built and optimized for high performance computing.
The two areas are completly distinct and have different "best practices" when it comes to network topology, configuration and interconnect selection.
Just because GigE works fine as an interconnect on a compute farm does not mean it is going to work fine on a cluster built to satisfy a completly different requirement.
Broadly speaking the word "cluster" means different things to different people.
There are people who build speedy Linux Clusters where the focus is on High Performance or High Throughput Computing. Then there are people who build ultra-reliable services (High Availibility Clustering).
The difference between "high performance", "high throughput" and "high availibility" clusters can be extreme.
For this particular Slashdot article the only people posting solutions or suggestions should be people who have actually used the RAC product in a production environment. Everyone else just needs to sit back and watch...
Re:No one size fits all answer but here is mine :)
on
Linux Clustering Hardware?
·
· Score: 3, Informative
Depends on the specific code to meet your criteria of "twice as fast"...some apps will be more than twice as fast; some will be slightly faster, equal or in some cases slower.
For more general use cases (at least in my field) I can give a qualified answer of "dual Opteron currently represents the best price/performance ratio for small SMP (2-4 CPUs)".
I've also seen cheaper pricing from Sun than what you mentioned. You are right though in that there is a price difference between xeon vs opteron - whenever I consider a more expensive alternative I tend to have fresh app benchmark data handy to back up the justification.
-Chris
Links to additional press coverage
on
Biohackathon
·
· Score: 2
To date we were the Feature National story in Business Day:
LINK
Computer Week online - top story on their home page today:
LINK
sa.internet.com picked up the press release and covered it:
LINK
Some additional media interviews were given today so there will likely be additional coverage. It's nice to see the press get most of the details correct:)
Yeah. We heard about this only after we arrived. We didn't know beforehand.
-Chris Dag
disks not suitable for heavy duty applications
on
Linux On HP Blades
·
· Score: 5, Insightful
The biggest problem I have with these systems (and the ones from RLX) is that they put cheezy laptop hard disks on the blades. The not-so-fast 4300 RPM drives or whatever they are using now are simply not fast enough for I/O intensive tasks.
I'll stick to standard high density rackmounts for my cluster projects that need better local disk IO.
Note the word "concieved" in the first line of the document...there is not much in place yet. The page sounds to me like a breathless plea for lots of venture capital funding.
There are several companies rushing to commercialize this general way of sequencing and haplotyping. The most interesting one I've seen so far is US Genomics.
The problem is still at the stage of "translating our pretty drawings and ideas into working hardware". There are tremendous engineering challenges both with the hardware (detection devices and materials handling) and software (algorithms etc.)
The science is solid and the ideas are clever. Its just going to take a while to build a system that can reliably resolve single molecules as they speed down a tiny channel.
my particular bias is life science it will show with the following URLS: bioperl.org, biojava.org, biopython.org. Find projects like these in whatever dicipline you are interested and you will find lots of science-capable programmers either actively contributing or participating in discussion. Don't spam the mailing lists with job listings of course - be a bit more subtle and find out things like what school programs are producing these folks, what conferences & events they attend etc. etc.
You are correct -- grad school labs are a real world environment research environment. I was wrong to have described industry as being 'real world'. The feeling that I was trying to get across is that I think that I have learned more interesting things faster and with less effort by being involved directly in the industrial/commercial side of things. I have the ability to throw myself directly into interesting problems and areas without having to worry about resource constraints, teaching assistant responsibilities and required-but-not-useful classes etc. etc. This approach worked out nicely for me but may certaintly not be the norm. Heck I still need grad school to pick up those MS level degrees...
I've been working in bioinformatics for 5+ years now and this topic is of extreme interest to me because I went straight from undergrad into a commercial biotech outfit. Generally speaking I'm very happy I did that and I honestly believe that I learned more by working in a real world discovery research environment than I could have in graduate school.
I've often considered going back to grad school but have always come to the conclusion that I am learning more by being in the real world. Getting paid to learn also does not hurt:)
Practically speaking you only need a phD level degree if you want to (a) be in academia or (b) manage your own group or lab someday or (c) apply for grants or teach in a university setting. For everything else the doctoral degree is (I think) pretty much a multi-year waste of time. My eventual goal is to go for the generalist approach -- multiple master's level degrees spanning computer science and molecular biology.
The best thing about industry is that I am judged only on my abilities and accomplishments. They don't give a darn about my academic pedigree and I don't have to deal with status-obsessed ivory tower types who look down on anyone without an advanced degree.
I live and work in the Boston/Cambridge area but if I had to pick any one school for graduate studies it would be UC Berkeley. They have by far the best mix of old school pioneers and young hotshot scientists/faculty. I would love the opportunity to work with people like Steven Brenner and Suzi Lewis. I have a disctinct feeling that over the next 5-10 years some amazing science is going to come out of the UCBerkeley folks.
The reason it is so hard to pick a grad school (or even undergrad for that matter) is that there was a period a few years ago where practically everyone who had done anything with bioinformatics was hired away from the unversities by the big pharma companies. There were whole departments that were essentially gutted by faculty leaving for industry positions. This left a HUGE hole for people looking to learn. This is slowly changing but has not been totally fixed. There are still lots of schools offerning computational biology or bioinformatics programs that only have 1 or 2 'real' faculty capable of mentoring and teaching well. UCBerkeley is (I think) the exception here.
simply put Sun is going to make Grid a core part of their software offerings so that all their own customers can use it pretty easily.
Along with that the sun people are *very* committed to makeing Grid available everywhere. They are committed to supporting as many architectures as the code can be ported to.
I'm not going to name my company because I'm not sure if I'm allowed to yet but we are one of the groups gearing up to offer Grid support for Linux and all other non-Sun hardware architectures.
I really see Grid not competing with code like PBS and Condor. It will probably annoy the Platform Computing people who make LSF because they charge so much $$ per CPU that LSF simply does not make sense on clusters built from commodity hardware.
There is plenty of room for cluster and compute farm companies. Besides technical skills and service which were mentioned in other comments another big differentiator is specific skills or experience in particular domain or vertical market.
As an example, when not working on bioperl:) my day job is with a company that only builds clusters and linux compute farms for the hardcore biotech and pharma crowd. The reason people hire us is that besides the hardware geeks we also have the PhD level computational biologists who understand the algoritihms, software and underlying science. Knowing what your customer is trying to do with the cluster helps greatly in configuring and tuning it:)
Profit margins in hardware, especially in the commodity intel platform are almost non-existant. The way you make money is with the services and software you put on top of the inexpensive hardware.
My last SAN project was 5TB raw disk and the solution that EMC pitched me was close to 1Million dollars more expensive than the competition.
In addition to being cheaper, the alternatives were mostly faster than EMC disk *AND* they played nicely with fiber channel hardware from other vendors (unlike EMC which likes to lock you in to their hardware only).
Not to knock EMC (solid product & killer support) but their obscene (no other way to describe it) pricing only makes their stuff worthwhile in situations where you need a 'black box' solution where some other guy is on the hook for hardware failures. In the life sciences I see EMC disk being used on drug manufacturing process hardware as well as on databases and systems that come under FDA scrutiny (patient outcome and clinical trial data, etc.) Generally people with more money than sense purchase EMC for anything but the most absolute mission critical stuff.
The other thing that annoyed me about EMC was the overly agressive frat-boy style sales force. The internal competition to make sales quotas is killer I've heard.
I ended up going with Brocade fiber channel switches (Silkworm II) and Compaq StorageWorks disks. We needed a SAN that could talk to NT, Linux, Tru-64, Solaris, HP-UX and Irix systems all at once.
I'm not a Compaq cheerleader but I like the StorageWorks line because although they are not always the first to market with the latest buzzword technology when they do come out with a product it is generally really solid and actually reasonably priced. The other cool thing about their new universal drive form factor is that all of their disks are now plug and play from the lowest end proliant server all the way up to their high end systems.
As for RAID levels and such you really need your database architects to tell you what they need. It may end up being a mix of RAID 1+0 and RAID5 for some filesystems and they may ask for solid state disks to store indices and such. Hardware tuning for high-end databases is a whole field in itself and there are lots of people out there who can probably tell you exactly what should be needed.
What you are going to find at the end of this project is that disk capacity is pretty simple and easily handled. The real problem you are going to have is figuring out how you are going to backup 5TB worth of DB data:) Not a trivial task by any means...
The lucent 802.11b cards (sold as 'Orinoco' for home/SOHO users) seem to have the largest amount of Linux suport when it comes to drivers and HOWTO documents. They are not cheap though.
I bought the RG-1000 gateway and an Orinoco silver card a few days ago and after a bit of fiddling had my linux notebook up and running at 11mbs. The windows client software was better at things like measuring signal strength & noise with a shiny little GUI. Range and performance seem to be very good. One problem with the RG gatway product is that it requires a Windoes system to set up the initial configuration.
Another option depending on driver ability would be the 802.11b stuff from Compaq. I noticed that they are selling their gateway software CD for $125 which means you can build your own access point on an existing system for the cost of the software and a PCI wireless card.
I spent almost a year trying to put together a plan for a 7TB storage area network that would be able to play nicely with Alpha/Tru64, HP-UX, Sun, Irix, NT & Linux. I also wanted to hook up that massive fiber channel ATL-P3000 DLT tape library into the SAN fabric as well. The eventual goal was direct backups to tape via the SAN fabric.
We ended up going with five 16 port Brocade Silkworm II switches meshed together. The RAID was Compaq StorageWorks HSZ80 controllers with a mix of their older form factor drives and their new universal drives/chassis.
All and all I've been happy with Storageworks disks and controllers over the years. They are not the cheapest or the first to market with the cutting edge technology but when they do come to market the stuff just plain works.
For people who are considering doing tricky SAN topologies in environments where there are more than one operating system and fiber channel host adaptor type I would strongly recommend at least evaluating the Storageworks product line. They really shine when you have lots of different systems that have to play nicely on the switched fabric.
Building a SAN in a homogenous computing shop is easy. Building one that will interoperate with lots of different systems requires much more work & initial planning.
If you go with NAS instead of SAN I'll also throw my.02 in as recommending NetApp over most other NAS vendors. Be carefull of scaling and backup issues though.
EMC is the choice of dot coms with more money than sense as well as easily frightened IT execs who are from the "you can't get fired for buying IBM" school of thought.
EMC will sell you an incredibly expensive system that is not particularly fast (for what you pay) and does not play well with any other type of storage system (especially in SAN environments)
I can think of very few situations where EMC's advanced software features, bulletproof uptime and excellent service organization make sense and are worth the obscene price premium that EMC charges.
Nine times out of ten, especially when thinking about fibre channel and SAN stuff it makes much more sense to deal with a vendor that makes interoperable hardware based on open systems and standards. You can save tons of money, get better performance and avoid vendor lock-in all at once.
My 7TB+ SAN needed to play niceley with Alpha, Sun, SGI, HP, Linux & Wintel all at once. I ended up going with Brocade FC switches and Compaq Storageworks disks & controllers.
Mind you I like the EMC product line but my personal opinion is that EMC is not worth the price in most cases - especially where the customer has no need or interest in getting the advanced software and hardware features that EMC sells as options.
Rackable (http://www.rackable.com has some really nice high density rackmount systems with interesting cooling and wire management features. If you want you can get the systems without top or bottom covers and they can be installed tray-like into the rack.
They have 1/2 width 1U units that can go dual CPU which means you can get 88 dual-processor systems in a single rack:)
The Olympics people are very, very agressive about preserving their namesake. I am affiliated with a nonprofit that runs an annual athletic event with the word "Olympics" in the event name and the US committee lawyers paid us a visit a few years ago. The only reason we are able to keep using the name is because we have been doing the event for many decades (well before their trademark was established).
That is basically the deal I have with CAIS Internet (www.cais.com). I have the service that they traditionally market to SOHO and small businesses. They call it "HyperLAN" for some dumb reason. I've got about 5 servers now doing misc. things and a small powerPC cluster is in the works.
192k is small bandwith compared to the speed the residential cable modem and DSL users get but I like my quality of service and the pile of static IPs they gave me. It is also cheaper than what I was laying out for 56k DOV ISDN back in the day:)
The freedom to push whatever the heck kind of data/services I want upstream more than makes up for the small pipe and higer cost. The acceptable use policies and terms that get forced on the residential contracts are just too limiting.
Despite some initial problems I would recommend them as a solid ISP to deal with regarding DSL as long as you know enough to 'guide' their first level tech support people.
I'm in the Boston area and CAIS is in the DC area so one downside is that my packets go through NYC and DC before actually getting out on the internet.
Slashdot Mirror
The Register has been on this for a while and although the story is older it is better written and has more interesting details: http://www.channelregister.co.uk/2008/01/16/mysterious_web_infection_continues/
my $.02 of course
I think you may be underestimating how hard this actually is ...
How many different power grids feed your company?
Redundant networking is well understood, so is the engineering required to bring in that networking via multiple trenches and facility penetration points so that you are protected against someone cutting a line or opening a trench accidentally with heavy equipment.
This is much harder for electricity.
At some point, probably pretty close to your facility you are going to be drawing from the same grid / same source. Whats the point of multiple trenches for power when you are drawing off the same grid or transformer a short distance away? You still have a potential single point of failure. This is why the facilities invest in local power generation and backup rather than diverse feeds which may be impossible or impractical to obtain.
The only reason I know this is that I've got a colocation cage rented at a Boston area datacenter that through an odd stroke of history actually IS connected simultaneously to 2 different utility power grids. The people who work there love to point out how rare it is to find a building that is attached to completely different upstream power sources.
Anyway, just my $.02
I'm in the same boat. Had a 15 inch powerbook for years and my first cardbus product was the Verizon EVDO high speed wireless card which I bought about 10 days ago :)
Works great but is it obsolete when I upgrade (which will be soon as I hate having a biz notebook that is out of applecare coverage)?
Google hints of compatability options or adaptor cards but I can't find solid info. Hints, tips and URLS appreciated!
-chris
Not affiliated with them, just like their half-depth 1U servers quite a bit.
Their DC power stuff is quite cool:
http://rackable.com/products/dcpower.htm
-Chris
{ excerpting from my own reply made in a different section of this article ... }
There are many people posting here who are completely confusing what the word "cluster" means for this particular question.
This article is about APPLICATION CLUSTERING (in this case a very specific relational database) and you are answering the question with information that is generalized to a COMPUTE FARM or a Linux cluster built and optimized for high performance computing.
Broadly speaking the word "cluster" means different things to different people.
There are people who build speedy Linux Clusters where the focus is on High Performance or High Throughput Computing. Then there are people who build ultra-reliable services (High Availibility Clustering).
The difference between "high performance", "high throughput" and "high availibility" clusters can be extreme.
For this particular Slashdot article the only people posting solutions or suggestions should be people who have actually used the RAC product in a production environment. Everyone else just needs to sit back and watch...
There are many people posting here who are completely confusing what the word "cluster" means for this particular question.
This article is about APPLICATION CLUSTERING (in this case a very specific relational database) and you are answering the question with information that is generalized to a COMPUTE FARM or a Linux cluster built and optimized for high performance computing.
The two areas are completly distinct and have different "best practices" when it comes to network topology, configuration and interconnect selection.
Just because GigE works fine as an interconnect on a compute farm does not mean it is going to work fine on a cluster built to satisfy a completly different requirement.
Broadly speaking the word "cluster" means different things to different people.
There are people who build speedy Linux Clusters where the focus is on High Performance or High Throughput Computing. Then there are people who build ultra-reliable services (High Availibility Clustering).
The difference between "high performance", "high throughput" and "high availibility" clusters can be extreme.
For this particular Slashdot article the only people posting solutions or suggestions should be people who have actually used the RAC product in a production environment. Everyone else just needs to sit back and watch...
Depends on the specific code to meet your criteria of "twice as fast"...some apps will be more than twice as fast; some will be slightly faster, equal or in some cases slower.
For more general use cases (at least in my field) I can give a qualified answer of "dual Opteron currently represents the best price/performance ratio for small SMP (2-4 CPUs)".
I've also seen cheaper pricing from Sun than what you mentioned. You are right though in that there is a price difference between xeon vs opteron - whenever I consider a more expensive alternative I tend to have fresh app benchmark data handy to back up the justification.
-Chris
To date we were the Feature National story in Business Day: LINK
Computer Week online - top story on their home page today: LINK
sa.internet.com picked up the press release and covered it: LINK
Some additional media interviews were given today so there will likely be additional coverage. It's nice to see the press get most of the details correct :)
-Chris Dag
I'll stick to standard high density rackmounts for my cluster projects that need better local disk IO.
my $.02 of course
There are several companies rushing to commercialize this general way of sequencing and haplotyping. The most interesting one I've seen so far is US Genomics.
The problem is still at the stage of "translating our pretty drawings and ideas into working hardware". There are tremendous engineering challenges both with the hardware (detection devices and materials handling) and software (algorithms etc.)
The science is solid and the ideas are clever. Its just going to take a while to build a system that can reliably resolve single molecules as they speed down a tiny channel.
just my $.02
my particular bias is life science it will show with the following URLS: bioperl.org, biojava.org, biopython.org. Find projects like these in whatever dicipline you are interested and you will find lots of science-capable programmers either actively contributing or participating in discussion. Don't spam the mailing lists with job listings of course - be a bit more subtle and find out things like what school programs are producing these folks, what conferences & events they attend etc. etc.
-c
I've been working in bioinformatics for 5+ years now and this topic is of extreme interest to me because I went straight from undergrad into a commercial biotech outfit. Generally speaking I'm very happy I did that and I honestly believe that I learned more by working in a real world discovery research environment than I could have in graduate school.
I've often considered going back to grad school but have always come to the conclusion that I am learning more by being in the real world. Getting paid to learn also does not hurt :)
Practically speaking you only need a phD level degree if you want to (a) be in academia or (b) manage your own group or lab someday or (c) apply for grants or teach in a university setting. For everything else the doctoral degree is (I think) pretty much a multi-year waste of time. My eventual goal is to go for the generalist approach -- multiple master's level degrees spanning computer science and molecular biology.
The best thing about industry is that I am judged only on my abilities and accomplishments. They don't give a darn about my academic pedigree and I don't have to deal with status-obsessed ivory tower types who look down on anyone without an advanced degree.
I live and work in the Boston/Cambridge area but if I had to pick any one school for graduate studies it would be UC Berkeley. They have by far the best mix of old school pioneers and young hotshot scientists/faculty. I would love the opportunity to work with people like Steven Brenner and Suzi Lewis. I have a disctinct feeling that over the next 5-10 years some amazing science is going to come out of the UCBerkeley folks.
The reason it is so hard to pick a grad school (or even undergrad for that matter) is that there was a period a few years ago where practically everyone who had done anything with bioinformatics was hired away from the unversities by the big pharma companies. There were whole departments that were essentially gutted by faculty leaving for industry positions. This left a HUGE hole for people looking to learn. This is slowly changing but has not been totally fixed. There are still lots of schools offerning computational biology or bioinformatics programs that only have 1 or 2 'real' faculty capable of mentoring and teaching well. UCBerkeley is (I think) the exception here.
Just my .02. Your experience/opinions may vary.
-c
simply put Sun is going to make Grid a core part of their software offerings so that all their own customers can use it pretty easily.
Along with that the sun people are *very* committed to makeing Grid available everywhere. They are committed to supporting as many architectures as the code can be ported to.
I'm not going to name my company because I'm not sure if I'm allowed to yet but we are one of the groups gearing up to offer Grid support for Linux and all other non-Sun hardware architectures.
I really see Grid not competing with code like PBS and Condor. It will probably annoy the Platform Computing people who make LSF because they charge so much $$ per CPU that LSF simply does not make sense on clusters built from commodity hardware.
Just my $.02
-chris
As an example, when not working on bioperl :) my day job is with a company that only builds clusters and linux compute farms for the hardcore biotech and pharma crowd. The reason people hire us is that besides the hardware geeks we also have the PhD level computational biologists who understand the algoritihms, software and underlying science. Knowing what your customer is trying to do with the cluster helps greatly in configuring and tuning it :)
Profit margins in hardware, especially in the commodity intel platform are almost non-existant. The way you make money is with the services and software you put on top of the inexpensive hardware.
just my$.02
-chris
My last SAN project was 5TB raw disk and the solution that EMC pitched me was close to 1Million dollars more expensive than the competition.
In addition to being cheaper, the alternatives were mostly faster than EMC disk *AND* they played nicely with fiber channel hardware from other vendors (unlike EMC which likes to lock you in to their hardware only).
Not to knock EMC (solid product & killer support) but their obscene (no other way to describe it) pricing only makes their stuff worthwhile in situations where you need a 'black box' solution where some other guy is on the hook for hardware failures. In the life sciences I see EMC disk being used on drug manufacturing process hardware as well as on databases and systems that come under FDA scrutiny (patient outcome and clinical trial data, etc.) Generally people with more money than sense purchase EMC for anything but the most absolute mission critical stuff.
The other thing that annoyed me about EMC was the overly agressive frat-boy style sales force. The internal competition to make sales quotas is killer I've heard.
I ended up going with Brocade fiber channel switches (Silkworm II) and Compaq StorageWorks disks. We needed a SAN that could talk to NT, Linux, Tru-64, Solaris, HP-UX and Irix systems all at once.
I'm not a Compaq cheerleader but I like the StorageWorks line because although they are not always the first to market with the latest buzzword technology when they do come out with a product it is generally really solid and actually reasonably priced. The other cool thing about their new universal drive form factor is that all of their disks are now plug and play from the lowest end proliant server all the way up to their high end systems.
As for RAID levels and such you really need your database architects to tell you what they need. It may end up being a mix of RAID 1+0 and RAID5 for some filesystems and they may ask for solid state disks to store indices and such. Hardware tuning for high-end databases is a whole field in itself and there are lots of people out there who can probably tell you exactly what should be needed.
What you are going to find at the end of this project is that disk capacity is pretty simple and easily handled. The real problem you are going to have is figuring out how you are going to backup 5TB worth of DB data :) Not a trivial task by any means...
just my $.02
-chris
I bought the RG-1000 gateway and an Orinoco silver card a few days ago and after a bit of fiddling had my linux notebook up and running at 11mbs. The windows client software was better at things like measuring signal strength & noise with a shiny little GUI. Range and performance seem to be very good. One problem with the RG gatway product is that it requires a Windoes system to set up the initial configuration.
URL would be www.wavelan.com.
Another option depending on driver ability would be the 802.11b stuff from Compaq. I noticed that they are selling their gateway software CD for $125 which means you can build your own access point on an existing system for the cost of the software and a PCI wireless card.
The open letter from Ewan Birney and Sean Eddy that the genomeweb article talks about can be read in full at:b er/001826.html
http://bioperl.org/pipermail/bioperl-l/2000-Decem
If you plan to write to Donald Kennedy check the listserv email thread to read about a correction to his email address.
-chris
I spent almost a year trying to put together a plan for a 7TB storage area network that would be able to play nicely with Alpha/Tru64, HP-UX, Sun, Irix, NT & Linux. I also wanted to hook up that massive fiber channel ATL-P3000 DLT tape library into the SAN fabric as well. The eventual goal was direct backups to tape via the SAN fabric.
.02 in as recommending NetApp over most other NAS vendors. Be carefull of scaling and backup issues though.
We ended up going with five 16 port Brocade Silkworm II switches meshed together. The RAID was Compaq StorageWorks HSZ80 controllers with a mix of their older form factor drives and their new universal drives/chassis.
All and all I've been happy with Storageworks disks and controllers over the years. They are not the cheapest or the first to market with the cutting edge technology but when they do come to market the stuff just plain works.
For people who are considering doing tricky SAN topologies in environments where there are more than one operating system and fiber channel host adaptor type I would strongly recommend at least evaluating the Storageworks product line. They really shine when you have lots of different systems that have to play nicely on the switched fabric.
Building a SAN in a homogenous computing shop is easy. Building one that will interoperate with lots of different systems requires much more work & initial planning.
If you go with NAS instead of SAN I'll also throw my
EMC is the choice of dot coms with more money than sense as well as easily frightened IT execs who are from the "you can't get fired for buying IBM" school of thought.
EMC will sell you an incredibly expensive system that is not particularly fast (for what you pay) and does not play well with any other type of storage system (especially in SAN environments)
I can think of very few situations where EMC's advanced software features, bulletproof uptime and excellent service organization make sense and are worth the obscene price premium that EMC charges.
Nine times out of ten, especially when thinking about fibre channel and SAN stuff it makes much more sense to deal with a vendor that makes interoperable hardware based on open systems and standards. You can save tons of money, get better performance and avoid vendor lock-in all at once.
My 7TB+ SAN needed to play niceley with Alpha, Sun, SGI, HP, Linux & Wintel all at once. I ended up going with Brocade FC switches and Compaq Storageworks disks & controllers.
Mind you I like the EMC product line but my personal opinion is that EMC is not worth the price in most cases - especially where the customer has no need or interest in getting the advanced software and hardware features that EMC sells as options.
Just my $.02.
Rackable (http://www.rackable.com has some really nice high density rackmount systems with interesting cooling and wire management features. If you want you can get the systems without top or bottom covers and they can be installed tray-like into the rack.
They have 1/2 width 1U units that can go dual CPU which means you can get 88 dual-processor systems in a single rack :)
If they post stories too fast then they will cycle down the front page so fast that very few people will get a chance to read/comment/etc.
The Olympics people are very, very agressive about preserving their namesake. I am affiliated with a nonprofit that runs an annual athletic event with the word "Olympics" in the event name and the US committee lawyers paid us a visit a few years ago. The only reason we are able to keep using the name is because we have been doing the event for many decades (well before their trademark was established).
That is basically the deal I have with CAIS Internet (www.cais.com). I have the service that they traditionally market to SOHO and small businesses. They call it "HyperLAN" for some dumb reason. I've got about 5 servers now doing misc. things and a small powerPC cluster is in the works.
:)
192k is small bandwith compared to the speed the residential cable modem and DSL users get but I like my quality of service and the pile of static IPs they gave me. It is also cheaper than what I was laying out for 56k DOV ISDN back in the day
The freedom to push whatever the heck kind of data/services I want upstream more than makes up for the small pipe and higer cost. The acceptable use policies and terms that get forced on the residential contracts are just too limiting.
Despite some initial problems I would recommend them as a solid ISP to deal with regarding DSL as long as you know enough to 'guide' their first level tech support people.
I'm in the Boston area and CAIS is in the DC area so one downside is that my packets go through NYC and DC before actually getting out on the internet.