|
||||||||
|
|
![]() |
|
|
Reader Feedback and Insights The feedback on my idea of putting an iSeries blade server inside the eServer BladeCenter ["It's Time for an iSeries BladeCenter"] has been completely positive. Here's a sample of what some readers had to say. --Timothy Prickett Morgan I read your article and was left wondering about the unwritten performance issues that a BladeCenter may have issue with. I ask this because I haven't followed the BladeCenter story that much, as we are primarily an iSeries shop, and as you so clearly point out, BladeCenter isn't there for us. Being a customer who just did a massive server consolidation, I was wondering about the performance of a BladeCenter with I/O. We have found that we have gained significant unplanned performance from VLAN and our logical partitioning strategy. The HSL link between iSeries boxes also offers impressive benefits. Can you help me understand these aspects of a BladeCenter? --Nigel Timothy Prickett Morgan responds: Within a single two-processor iSeries blade, all the same VLAN capabilities across partitions apply. You could, in theory, put eight OS/400 partitions or 20 Linux partitions on a single blade server using the current hypervisor technology IBM has for the iSeries with OS/400 V5R2. The memory bandwidth on the iSeries blade server won't be anywhere near what you can get out of a 32-way Model 890, but it should still be pretty significant. As for talking out to SAN disk arrays, HSL is no faster than 2 Gbps Fibre Channel, as far as I know, so if IBM made the iSeries blades link to storage with these Fibre Channel or modified HSL links, there is no reason to lose that performance, either. And if IBM really wanted to be clever, it could use an InfiniBand switched fabric to connect the blades to each other, as well as to the switches and peripherals that feed those servers. I/O doesn't have to be a problem. We're a small AS/400 shop running V5R2 on a two-way model 720, which we purchased new several years ago. At V5R2, we're at the end of the line as far as software upgrades are concerned because our hardware (disks primarily) won't be supported beyond our present software level. Our tentative upgrade plans are to lease an iSeries Model 810 sometime in 2005. The 810 would also include Integrated xSeries Servers, which would replace our small Windows 2000 server farm. Rather than upgrading to a Model 810, having the option to put an OS/400 powered iSeries BladeCenter with Integrated xSeries Servers in our server rack is my idea of a perfect upgrade path. It could replace all of our servers with the latest equipment and software. Running OS/400 retains the stability and database functionality we need and are used to. Disk storage mapped over the whole affair gets us a simpler and much faster backup scenario (LTO-2 would be my choice) and provides RAID 5 for all servers. Our Microsoft network retains its profile. And, if IBM would just do it, the company could price it very competitively to sell a ton of them, configured just like (or similar to) this! Again, thanks for your persistence on this issue! --John This is a great idea, and I agree with your assessment 100 percent. I'm surprised IBM didn't think of this when comparing architectures. It appears that with a single-level store and separate I/O processors, the iSeries would actually be more appropriate than the big iron. --Dave Yes, the OS/400 iSeries blade server concept is a great idea whose time has come; however, IBM has bigger issues to deal with at the moment. I see the biggest problem right now is the pricing structure for iSeries shops that partition large P40 servers into a multitude of smaller partitions and have to pay high dollars for the software due to the P40 group tiered-pricing structure. Once that issue is resolved, the blade server concept can fly; otherwise, how will software be priced under the current tiered pricing model? But then again, is IBM in the business to sell hardware, software, services, or what? My company couldn't bring over an application from a European subsidiary and run it on its own partition here in the States, because our partitioned server is a P40 tiered server. We can run it on a Model 810 much cheaper, from a software licensing perspective, than on a partitioned Model 840 12-way server. Hence we have to pay annual maintenance on a server that we really don't need, but it's cost justified due to the software cost. --Greg I, too, would like to see an OS/400 blade server, but I think that's in direct opposition to IBM's new concept of on-demand processing. Think about it. They are giving away hardware with the off chance that someone will "turn it on" and then pay for it. That tells you the hardware doesn't really cost very much. The iSeries is way over-priced, from a pure hardware perspective. But we don't buy them because of the hardware; we buy them because of the software and unparalleled reliability. "Downgrading" an iSeries to a blade in a BladeCenter would increase the complexity of the hardware and probably decrease the reliability. The big selling point for the iSeries has always been to consolidate your server farms into one box. Having homogeneous hardware with other platforms would eliminate one of their selling points. The whole enterprise pricing scheme and exponential price increases for bigger iSeries models is nothing more than a way to keep IBM's revenue up to pre-competition prices. Before the PC and the client/server concept, IBM could charge what ever it wanted, because it really didn't have any competition in the multi-user computing environment. Now that PCs and Unix servers have the same or better horsepower, at much lower prices, they are scrambling to justify their ridiculous prices. Until IBM accepts the fact that big iron prices are gone forever, they will continue to lose market share. What difference should it make to IBM if someone uses their iSeries for 5250 or client/server? They have abandoned the workstation as a revenue stream. IBM could charge a client license for 5250, just like Microsoft does for access to Windows servers. IBM already charges for iSeries Access; shouldn't that be enough? --Anthony I agree with your assessment. I personally am at a crossroads in my career as a developer. I have missed quite a few opportunities to "convert" small businesses in my area over to the iSeries due to size or start-up costs. Something like this a few years ago would have changed a lot of outcomes for me. If IBM doesn't make some serious changes, my career in iSeries development will end and I will have to become a .NET developer. I have already made the skillset shift but would prefer to develop on the iSeries the rest of my days. Unfortunately, I have to go where the jobs are. --Tim We value your feedback and insights. Feel free to send a letter to the editor. Letters may be printed, unless otherwise specified, and edited for clarity or length.
|
Editor
Contact the Editors |
| Copyright © 1996-2008 Guild Companies, Inc. All Rights Reserved. |