Tuesday 30 April 2013

Intel Announces the End-Of-Life for the Intel Modular Server Platform

Bummer ...

Wow, we had a good run with the IMS platform.

It was the platform we build our first Hyper-V failover cluster on.

It was the platform where we learned what VLANs do, Trunking means, and how to configure centralized storage for shared access.

image

We have done _a lot_ of testing on this platform, Hyper-V cluster deploys, and so much more. We have had the fortune to have been in touch with folks from all around the world that reached out to us for assistance with their IMS questions.

We do see some neat changes coming down the pipe including a new Intel Storage System JBOD2224S2DP platform that is perfect for Windows Server 2012 Storage Spaces and VHDX delivered via SMBv3 to Hyper-V Failover Cluster setups. The new Intel Server System R1208JP4OC 1U single socket unit looks to be a perfect fit for entry level nodes with the ability to tie in two SAS HBAs for redundancy.

So, time to move on with our individual node configurations utilizing Promise, IBM, and now we are even looking at Dell MD3220 smart storage as an option.

Philip Elder
MPECS Inc.
Microsoft Small Business Specialists
Co-Author: SBS 2008 Blueprint Book

Chef de partie in the SMBKitchen
Find out more at
www.thirdtier.net/enterprise-solutions-for-small-business/

Windows Live Writer

11 comments:

Joshua Liberman said...

We too cut our HyperV clustering teeth on the IMS platform. I cannot imagine a better platform more "born for" HyperV clustering. I believe it is a HUGE mistake to kill this platform and cannot help but wonder why. They say it is all about the money, but the reality is that in well over five years on the market, it was never promoted in any meaningful way. I have to wonder if a product this far out ahead of the OEM competition wasn't just killed for that reason, and nothing more.

Philip Elder Cluster MVP said...
This comment has been removed by the author.
Philip Elder Cluster MVP said...

Joshua,

Intel's big push to build an in-house virtualization stack with failover capabilities was a real surprise to us and no surprise when that project was killed.

EPSD has been really good lately with producing very competitive systems. They have always seemingly struggled with vision for those products.

We were very active with the Intel IMS team. We worked with the engineers in storage and networking when we were looking to create all kinds of elaborate setups for tenancy setups.

The biggest drawbacks for the platform at this point:
+ 3Gbit SAS
+ 10Gbit Ethernet limits
+ 3Gbit SAS to Compute (lane pair)

Probably the biggest two catches for us were the sinle SAS lane per storage controller on each compute and the 3Gbit limit.

Philip

Anonymous said...

I believe you're one of the lucky one without any issue with IMS. Most of our SEA channel partners experienced dreadful issues and failures a lot. From over 90% units we sold dated since 1st generation facing so much problem that one of our customer sued us. It's not a big surprise Intel killed it since Intel they themselves having issue to solve majority of the issue.

Philip Elder Cluster MVP said...

A.

We've burned a _LOT_ of Hyper-V cluster hours into the IMS platform over these last four or five years.

We are talking about clusters being up for four years straight here!

All on the Intel Modular Server platform.

I'm sorry, but I view that claim with deep skepticism.

One of our client IMS units had two SD card failures on it. Since we were a part of the IMS Champions Program we had front-line access to the team.

The request was for us to send the second midplane in to Intel's labs directly.

It turned out that the hole the client poked in the double brick (old insurance company vault) wall released a bunch of asbestos fibres. Guess what? They are conductive.

The platform has been rock solid for us.

However, it is _not_ easy to set up and get functioning with stability and performance in mind.

I'm sorry to hear about the problems experienced, but after a very long, and successful, run the IMS platform will be sorely missed by us and others.

Philip

Anonymous said...

Any thoughts on the Dell VRTX? I find the coincidence of the introduction of VRTX and the killing of IMS suspicious.

We had championed the IMS for 3+ years, and sold many (dozens) to our customers in the North East. It needed a technology refresh, but it was a "cloud in 6U" way ahead of its time.

Philip Elder Cluster MVP said...

Not suspicious at all.

The Dell team is having struggles virtualizin the two PERC controllers.

So, to date the VRTX only sells with _one_ RAID controller.

Defeats the whole purpose IMNSHO.

Philip

Anonymous said...

I noticed the fact that they only have the one PERC as well. I figure they will resolve that in time. Perhaps I have been doing this too long...with my conspiracy theories and stuff.

Great blog. I enjoy reading it and learning from it.

Philip Elder Cluster MVP said...

I find it interesting that Intel could do the dual SAS controllers, and the manufacturer of the PERC (LSI) has just released a virtualized RAID setup in the Syncro.

Syncro allows one to attach two DAS JBODs and set up redundancy across the two.

I do hope that Dell gets it figured out since we have a number of cluster deploys in the upcoming months that the VRTX would fit in with perfectly.

Thanks! :)

Philip

Breck Jeff said...

I'm currently trying to figure out how to replace the Intel Modular Server. I see the Dell VRTX, but am looking for something maybe a little bigger and expandable. I've looked into IBM's Flex and NEXT systems and they are a bit pricey. Maybe Intel's 1U or 2U server systems with attached storage is the way to go--SAN or DAS? And then there are some software solutions on the horizon that tout software defined storage that would eliminate the need for a SAN (Maxta, Starwind). I'm curious what others who are trying to provide enterprise-level services to small businesses are currently using or looking into.

Breck Jeff said...

I'm also figuring out what's next after the Intel Modular Server. We're running Hyper-V with clustering. What would replace it? What are others using? I've looked into IBM's Flex & Nextscale systems, but they're a bit pricey. If I were to stick with Intel, what's the best products--1U or 2U with SAN or DAS? Or is software defined storage (from the likes of Maxta or Starwind) the answer?