IBM has recently announced that their original chassis the BladeCenter E will be available until 2015!
This was their original BladeCenter Chassis, it began shipping in 2002. That is 13 years of life. Seems like they got it right the first time while HP, Dell, Sun, etc keep changing and sun-setting their designs. IBM’s designs are compatible with each other. I.E. if you purchased a Blade Server back in 2002 and decided you now want to run it in your newest BladeCenter H….no problem it will fit right in. Likewise the majority of their newer servers will work in the original BladeCenter E that was built in 2002.
The only issue they are running into is the power envelope in this original chassis. They designed it to be energy efficient way back when processors were single cores…. thus now with 4 and 8 core processors coming available they have to upgrade the firmware and possibly the power supplies to run these beefier boxes.
The alternative you get from HP and Dell is to upgrade your entire infrastructure….i.e. their blades and switches do not swap from chassis to chassis! What a waste!!!!
gdrawdy
8 comments
Comments feed for this article
April 12, 2008 at 9:25 pm
Jeff Allen
I know of a little little networking company out west – perhaps you have heard of them – named Cisco? The whole “blade” idea came from the network world where they build a “Chassis” and slide “blades” into it. Because technology evloves so quickly, and it typically changes in both size and performance, Cisco decided long ago to make no chassis backward or forward compatible so that they could focus on building the fastest chassis possible – and not spend time designing in a way to run obsolete blades just so they can say they can do it.
For IBM to think that they got blades perfect the first time and that there will never be a need for another chassis design is not only foolish but also arrogant and idiotic. IBM is the odd man out with their formfactor now that Dell and HP have made 16 blades in 10U the standard. I can’t speak for Dell because they are a copycat of HP, but I can say that HP server customers expected more from HP – more than they could get from their older pClass blades. HP took that challenge and built a far better enclosure than anything available. Within a year, Doug Blalock at IBM all but disappeared from public view – quite a contrast from his famed appearences stating how IBM had “won” the blade market. Apparently their blades are more forward compatible than their seniro VP’s are.
If running the old technology in a brand new chassis is important to IBM, they can have that silly claim to fame – and they can eat crow when the annouce a new chassis (oh yes, it’s coming). According to the latest IDC numbers, customers don’t seem to care what IBM is doing with blades. HP overtook IBM in mid 2006 and has increased their lead every quarter. HP was the #1 blade vendor of 2007 and they are looking like they will repeat in 2008. So again, IBM can have all the “compatability” they want – and I’ll keep buying these “inferior” HP blades
April 12, 2008 at 11:42 pm
gdrawdy
Jeff – to try to validate the 4 mulligans that HP and Dell have done to their install base by using Cisco’s engineering plan is short sited. Cisco’s decision to not put future technology capabilities in their switch chassis does not mean this is how it should always be done. But to that point it is obvious that is what both HP and Dell have done and will obviously continue to do. Don’t fault IBM if they engineered their chassis to go the distance. By your reasoning you could say the IBM BladeCenter E is by far more superior because Intel OEMs it as their Enterprise Blade Server Chassis (http://www.intel.com/design/servers/blades/SBCE/index.htm). You know why they choose it over the half height design that HP and Dell are trying to go with? No single point of failure and heat control. Intel knows it is all about getting the heat out. They should know….their components create most of it.
HP and Dell have abandoned their install base by introducing at least 3 non-compatible chassis’s far to often. If you knew something about the blade server market you would know that moving an older blade to a new chassis is NOT necessarily what a company looks for in forward and backward compatibility.
It’s also the IO infrastructure in those chassis they are looking for. They are tired of shelling out an additional $25 – $40k each time HP and Dell decide its time to have another do over. But since you brought up Cisco….you should know that customers who bought an IBM BladeCenter E in the early 2000s and equipped it with Cisco Ethernet switches (either copper or fiber optic based) can still use those SAME switches in EVERY chassis IBM has ever sold. The same is true of the Fibre SAN switches. Can HP or Dell do this? NOOOO$$$$$$$$$. This is where the value of IBM foresight and engineering comes in and is what I hear time and time again when I speak to those who are moving from HP and Dell to IBM. They are tired of having to buy 2+ new fibre san switches for $8 – $12k each and 2+ new Ethernet switches for $3-$5k. Also they enjoy the fact that they can still get support for and buy newer blades to run in the old chassis in production, a DR site, or in a test\dev situation.
I do agree with you on the “inferior” part of HP blades though. The CIOs and CTOs I meet with are not however as eager to continue buying them as you appear to be. They have tired of the lack of quality support and tired of getting strung out with sunsetted technology. They have indicated it takes very close scrutiny to determine what the HP marketing really means. An example of this is how HP with great fan fare released the WWPN virtualization via Virtual Connect to their product line. With great exuberance many were caught up in it just to find out it only support 4 Systems….and ONLY HP proprietary branded switches. In contract IBMs supported 100 different chassis (YES ONE HUNDRED compared to FOUR) and EVERY switch in the product line…which I believe is from over 6 different networking vendors. (AGAIN THAT IS 1400 IBM Blade Servers with virtualized WWPN compared to 64 HP Blades….thus proving your point on HP being inferior).
That is where I see the major difference….MARKETING. IBM is terrible at marketing…..absolutely terrible. Does this make their products bad? No it makes them a secret and allows the HP and Dell hype machine to run rough shod over people who don’t investigate what their marketing words really really mean.
I am amazed over and over again when I speak to people such as yourself who tout the banner of greatness for either HP or Dell but can’t step back and look independently on the fact without having their vision blurred by marketing hype. Here is another example of marketing hype….HOT SWAP drives. HP and Dell both throw this up time and time again to insinuate their blades are far superior. BUT if you truly look at their blades you will see that…by putting these drives in front of the CPUS, and the CPUs in front of the memory chips, they are FRYING their memory chips. Thus by marketing HS drives they kill their memory which performs poorly and must be swapped out. Guess what….you aint hot swapping memory in a blade. Why not just san boot your blades, which is what a majority of my C level exec’s want anyway. They want to virtualize the blade so that if any component fails they can continue to run….and they can do that across much more blades and chassis configurations than either HP or Dell. Does IBM do everything right…NO….but they have made a commitment and have left far fewer people hanging on to unusable technology at least in the Blade market.
I have on my desk a Network World (April 7 2008). The back cover is a Dell Blade System ad blasting HP’s Blade System. The claim is “19% LESS POWER CONSUMPTION”. So while you say they are a copycat apparently they are now going after their mentor. Of course Dell doesn’t point out that their Blade System support much less I/O or processing capability. But this shows how HP is getting bit by someone doing the same thing to them that they have been doing to IBM for years and years.
Another issue I have with both the HP and Dell chassis is the CIOs and CTOs I work with demand no single point of failure. Neither of these two vendors provide redundant paths from the blade through the midplane, to the rear of the chassis. IBM does. I have schematics and pictures that show it. Keep watching though because just as you say IBM is coming out with a new chassis….I suspect that HP and Dell will also. They have had 3 chassis over the past 6 years or about 1 every 2 years. Their current chassis is a bit over 1 year old so it is about time for them to call their 4th mulligan and try again…..there by stranding once again their install base….and forcing many admins such as yourself to once again have to learn a new piece of hardware….just because they didn’t get it right once again.
April 13, 2008 at 12:06 am
gdrawdy
I just have to add a bit more:
- IBM blades have Solid state disk drives – they, along with Apple have solved the memory problem in SSDs. HP – NOPE.
- Externally connected SAS – HP – NOPE.
- 2 current shipping chassis that support 10gb – HP – NOPE.
- Layer 2-7 ethernet – HP – NOPE.
- Managed Infiniband – HP – NOPE.
- Choice of 4 processors Intel, AMD, Power, Cell BE – HP – NOPE
More if you want….I would love to help you see the light….
April 13, 2008 at 10:54 pm
Jeff Allen
OK – thanks for the spanking, but I’m still apparently not seeing the light. But while I’m bumping around here in the dark, I have to correct some errors you have made:
1) Virtual Connect support 100 enclosures and has for a number of months since HP released Enterprise Manager. I believe that makes HP’s solution 1600 vs IBM at 1400.
2) “Redundancy” means I can do without that part and still function AND that I can replace that part without affecting the entire solution. I cannot replace IBM’s midplane without taking down the enclosure – same goes for HP. I believe I have the same schematics you reference and it shows a SINGLE PCB in IBM’s design. Since you brought up IBM’s midplane though, let’s talk about that for a minute. IBM chose an “active midplane design” which means they use eeproms and other active components on their midplane. Every eeprom is a single point of failure and IBM has over 50 on their midplane. HP chose a passive midplane – so did Cisco, Brocade, Nortel, Juniper, 3COM, Extreme, etc. All of these vendors chose a passive design and brag about doing so because they know how rock solid it is(oh, and none of them have “redundant connectors” because they know they don’t need it). I challenge you to find another vendor that even mentions the use of an active midplane and uses the word “redundancy” in the same paragraph. And don’t look to IBM’s storage division, they already have a redbook “Designing and Optimizing an IBM Storage Area Network” in which they quote “A Passive backplane Leads to very high level of reliability of the unit as a whole”.
3) IBM had to move to SSD because they can’t physically fir 2 drives in the system and not compromise other options. HP solved all that and has never made me choose between options and hard drives – and yes, they are hot plug (like every other server in the data center). But you didn’t mention that IBM actually does have hot plug drives for their blades – and what a great solution it is – I can get 7 whole blades in the chassis when I want hot plug drives.
4) HP does have 2 enclosures that support 10GB – both cClass enclosures support 10GB.
5) Your argument for processor types is invalid. HP supports every processor in blades that they support in their other servers just like IBM does. HP is obviously not going to support IBM proprietary CPUs.
6) IBM’s OFM solution where they use Cisco and Nortel (and Brocade for FC) switches is complete crap. Here’s why. With HP and VC, I can have 1 single server act as a cold-standby for whatever number of blades I feel comfortable with. All of those blades can be on any number of vlans. If a blade fails, VC not only moves the MAC and WWPN (a technology that HP invented), it also moves the vlans and fabrics that the failed blade was on to the spare blade. OFM cannot do this – it means nothing to me to have the same mac and WWPN if I am on the wrong vlans. With OFM, I have to have a spare blade for every vlan that my blades are on – that isn’t practical or affordable.
7) Anything above Layer 3 inside the enclosure is a complete waste of that resource since I can’t benefit from it elsewhere. A waste of money and HP got that right by not providing it.
April 14, 2008 at 7:56 pm
gdrawdy
Wow. Seems like you either work for HP or you have seriously had a Kool-Aid IV administered. And all of this started because I pointed out that HP and Dell continue to strand their customer base by taking mulligans on their chassis design. Thank goodness the chip manufactures don’t do the same thing.
I must have missed the VCE coming around a year late but obviously not a $ short. IBM’s OFM is 1499 per chassis compared to $9k per chassis for the HP solution. Who could afford 100 chassis of that? 100 chassis of the HP solution = $900k. WOW! Nice move HP. I’m sure a lot of big datacenter are jumping all over that. This is the HP game plan I run into time and time again. Show a low price and then bait and switch the CIO at the closing table….”Oh I’m sorry sir…you meant you wanted a optical drive with that???? Errr…that’s extra sir”.
The financial institutions I work with DEMAND the layer 3-7 INSIDE the BladeCenter. They love it for “Bank in a chassis” type solutions. I’m sure that soon HP will have a layer 3-7 solution and will then begin touting that everyone needs it. This was what Dell did last year without a Blade Solution. They told all their customers blades were dead….now that they have a lower-end blade system….look out…now they say everyone must have blades.
I’m sure in the bowels of HP they are working on figuring out how to spin a layer3-7 offering, a real KVM over IP solution, etc. When IBM releases their deep packet inspection blade later this year I’m sure the HP marketing machine will kick in….as surely it will when they release their I Series (AS400) blade just a bit later.
As far as redundancy goes I am glade you came up with your own definition of what Redundancy means….as surly this will make the HP customers feel better at buying an inferior product. Here are few definitions I found that are a bit more accurate than yours:
• Wiki: is the duplication of critical components of a system with the intention of increasing reliability of the system, usually in the case of a backup or fail-safe
• Dictionary.com: Duplication or repetition of elements in electronic equipment to provide alternative functional channels in case of failure
I am sure you claim redundancy isn’t redundancy unless you can replace the failed component without affecting the entire solution because this is one of HP’s failures. Airplanes have redundant systems….but they sure don’t change them out in flight….but thank God they decided they would go with redundancy anyway….no matter how you or possibly HP define it. Maybe AA should have redundant wiring harnesses….it would have kept me from being laid over in San Juan for so long
.
The combination of IBM’s OFM (the feature you descried as crap) and the active passive mid-plane (you are critical of) will enable IBM’s customers to move their servers during the day to another chassis during the work day…after a normal shutdown if needed….where as any failure in the HP mid-plane…BOOM…up to 16 servers hard down. Explain now how that is a better design on HP’s part? You may have sold your customers on that but none of mine will buy it. But if you do sell it…be sure to sell at least 16 spare blades in case they have a mid-plan failure….because they are going down hard!
Those chips on the mid-plane buy the way provide rich features that IBM’s customers love. Maybe you should pass your designs over to HP design to get them to put a few on the HP mid-plane. I’ve been working with these boxes since inception and by the way I have never had one go down because of a chip failure. I do hear that HP has between 2% and 4% fail because of failed mid-planes. Thats a lot for a simple circut board with imbedded cables. That is way too much for my customers. Of course some of these are probably because admins push in the blades sideways or put in the power supplies upside down or any number of other ways that HP didn’t fisher price their BS. I can hear it now….you telling a CIO who wants 5 9s of uptime….yes sir that’s right….HP only has 2%-4% failure rate on the mid-plane. Presentation OVER!
One point that you make on the internal drives I do think is a debatable. But it is HPish to claim that the reason IBM went to SSD was because they couldn’t fit additional drives. IBM has many blades that hold two SAS drives. When they first released the SSD I was not a fan of it because of its size. Now that it is in the 30g range I am much more of a fan especially now that I have run performance testing on it and discussed its MTBF(8-15 years) as compared to that of spinning disks. Also while I am bumping around in the dark as you say….look at the BCS that holds 12 internal disks (the shorty doesn’t have that) that can be shared between the blades and look at the BCS and BCH that has the ability to connect drive cages directly to the chassis. Choice is what I see here….if a customer wants a san give em a san….If they cant afford a san…give em sas.
As far as the 10g solution…..I got that off of the HP web site for the c3000 where it states “the c3000 Enclosure can support today’s Gigabit networks with an easy migration to tomorrows 10 Gigabit solutions”. I take your word that tomorrow is here at HP.
IBM has 5 chassis, 7 blades, and 5 different fabrics….more blades on the way!. HP can’t touch this in terms of choice. Seems they are more of a Ford type company…you can have your Model T painted any color as long as its black….remember that? Obviously their marketing group has done a good job at selling the c class but the customers I speak to are ticked. Why are they ticked? It appears that HP firmly believes that blade technology is progressing so rapidly that blade customers do not want to be tied to old technology. It seems that HP considers the old blade chassis and blades to be expendable. It seems that their goal is to upgrade each customer with an all-new blade server every five years or so. This view differs sharply from IBM’s strong investment protection plan. So how does HP handle IBM’s investment protection??? They have folks like you scour the blogs and try to blur the truth and in some cases it has worked. Customers want their servers to last more than 5 years…and they want to be able to buy a chassis now that will run 6 and 8 core processors or what ever the latest technology is that will cause HP to take their next mulligan. They want to move these systems into a test/dev/dr/cob environment and continue to use the investment they made in the chassis and infrastructure components. This is not possible with HP or Dell.
If you are looking to teach me something….teach me about why HP decided to design their blades so their memory chips run at McDonald’s grease temperatures….seems like those hot swap drives you tout are also good for creating an air dam to keep cool air out of the blade.
April 14, 2009 at 11:08 am
Ben
WTG IBM, for maximizing your customer’s investment and allowing them to reuse machines that would other wise be EOL.
October 7, 2009 at 10:29 am
Romulo Cholewa
Hi there,
I currently work for an IBM partner and I would like to confirm this info for a project we are working on. I searched the Internet for any other sources but couldn´t find one. Where did you see the IBM announcement stating that it extended the original BC mfg to 2015 ? Thanks in advance.
October 7, 2009 at 11:29 am
gdrawdy
If you currently work for an IBM BP check the road map for the BCE. i.e. if you purchase one you now or even near the end of 2012 will be able to have default support on it until 2015. Even in 2012 a 5 year upgrade on support will extend your service until 2017.