Slices Of i For The Little Guys
September 16, 2013 Dan Burger
Replacing chaos with calm while solving the mysteries of computer voodoo could earn you a pretty good reputation in the which-planet-am-I-on world of IBM midrange computing. It’s not that you’re more likely to get zapped here in the i zone than in any other IT fun house, but coming to grips with things like managed services and cloud computing have their own i-flavored outcomes here. Even for the little guy who is wondering how he fits into the MSP and cloud action.
Larry Bolhuis has an idea or two about the MSP and cloud computing topics and a few other related topics such as disaster recovery, high availability, independent ASPs, and PureFlex systems as well.
Some of you know Bolhuis. He’s a low-key, but high-profile guy in the IBM i community. His specialty is systems administration and he’s a well-traveled subject matter expert who is familiar to those who attend COMMON conferences and participate in local user group meetings where he shares tips, techniques, and stories of successes, near successes, and you can’t even see success from where you’re standing. His rent-a-slice-of-the-IBM-i business, called iDevCloud, is popular with colleges, local user groups, and individual developers who otherwise are cut off from IBM i connectivity. For the past year, iDevCloud has been part of a new business venture called i In the Cloud . And Bolhuis’ partner is Pete Massiello, owner of iTech Solutions Group . Massiello is a former president of the COMMON User Group and a current member of the board of directors. He is also a well-known speaker at tech conferences and local user groups as well as a subject matter expert in system administration. Bolhuis and Massiello launched i In the Cloud in August 2012. Their combination of skills and existing customers put i In the Cloud in a good position at launch time. And their reputations and customer referrals should serve them well into the future.
Small Fries’ Dilemma
The strategy going forward is to cater to small IBM i shops looking for the IT equivalent of a nicotine patch. Many would like to drop IT like a nasty habit if they could. Buying hardware and software, and hiring a staff to manage systems, is sapping their strength. The idea of renting the equipment and the skills it takes to run it–rather than owning–is sounding better with each passing year.
Processing power, even on the smallest of Power Systems boxes, has gone through the roof, leaving many small IBM i shops wondering why not give the care and feeding of the 800-pound gorilla to someone else, preferably someone who will selling exactly the processing power they need and when they want it.
There’s enough of this thinking going on to encourage Bolhuis and Massiello.
The smallest slice of the i In The Cloud is a tenth of a processor, 2 GB of memory, and 70 GB of disk. That’s going to work for a lot of small companies. Additional upgrades can be purchased in tenths of a processor, 1 GB nuggets of memory, and 35 GB chunks of disk.
The slices are shaved from an eight-core Power 720 machine with 5 TB of disk. Currently more than 40 partitions have been carved off for customers. (The technical limit for partitions on a box like this one is 80.) There are multiple Internet feeds to the co-lo site and customers with production environments there are on a virtual private network (VPN).
That server is located in a commercial-grade data center located in Michigan. It has all the expected redundancies and security built in to the infrastructure. You’ll find uninterruptible power supplies, generators for emergency power, multiple cooling systems, and logged, keycard-only access with always on duty security cameras.
There’s a lot more and a lot better equipment here than small businesses could hope to buy on their own. So who pays for all that stuff?
“We know that the little shops can’t afford more than $1,000 per month for the hardware, operating system, hardware and software maintenance, bandwidth, and backup media,” Bolhuis says. That price does not include any administrative work and the shops still have whatever software costs they have to deal with.
This pricing is in line with what I’ve found other MSPs charging, but there seems to be some flexibility in this to the degree of maybe a couple of hundred dollars, along with some negotiating on what services will ultimately be provided. Competition will keep the lid on this to a large degree. And one factor to take into account is the level of service and degree of expertise the buyer is receiving.
Once the monthly payment becomes $1,500, a company starts comparing the cost of buying or leasing a Power 720. Of course other factors also come into play such as the associated costs of power consumption, cooling, UPS, and maybe a generator for backup power are a few considerations. On the other side of the equation, the MSP isn’t likely to survive if too many of its customers are swarming close to the $1,000 per month figure. They will need a decent percentage of customers paying $1,500 and up to cover costs and make this worth their while.
The first step for many small IBM i shops is to come to the cloud for the system redundancy. It’s a system backup and disaster recovery plan and it’s the foundation for pretty much every cloud computing start-up business I’ve talked with. The majority of the replication work is being done on IBM i hardware, but you will also find smaller Windows and Linux servers in the mix–systems that are dependent on data stored in the DB2 for i database.
Soft or Hard HA
Thanks to a maintenance company that managed to crash a company’s system twice while attempting to deploy a high availability system, Bolhuis was able to bring that company to i In the Cloud.
Bolhuis has done the homework and become familiar with PowerHA and the independent Associated Storage Pools (iASPs) and storage area networks (SANs) that are integral to hardware-based high availability. He is working with several other companies that are in the process of moving to PowerHA.
Delivering HA at a reasonable price point could be a business magnet for i In the Cloud.
For companies that have not already invested in software-based HA, PowerHA is a good option, according to Bolhuis. Compared to software-based HA that uses journaling and is sometimes stymied by objects that won’t be journaled, PowerHA doesn’t take into account what kind of object is being replicated. Uncooperative IFS or database objects, for example, can be put into an iASP and mirrored. Developers who do not need to understand journals and journaling are generally less grumpy. (At least that’s what I’ve heard.) And although application changes will be required and iASPs need to be set up for PowerHA, he says neither task is difficult. He describes the iASP “just another set of disks” with one storage pool set up to contain the IBM i operating system, the spool files and subsystems, plus DB2 and the IFS.
PowerHA, however, doesn’t fit in a company that has hundreds of terabytes, he says. That’s partly because the initial save-to-tape process takes forever to back up. Software-based HA can synchronize much quicker than hardware-based HA because it recognizes the save markers (the restore markers follow the save markers) in the journal. It writes just the data designated by the save markers. PowerHA writes all the data every time it’s called on to mirror.
“To me, PowerHA is becoming very mature and very useful to the small shops,” Bolhuis says. “Companies can come into the co-lo and get small virtual disk, for instance, a total of 20 GB of disk for the OS and the data. That’s not a lot.”
“If you were buying a Power 720, the best you could hope for is to buy two 140 GB drives and mirror them. That would be weak in terms of performance, so you would buy four 140 GB drives to get decent performance. And with that the shop would have 400 GB it doesn’t need. Having enough disks to make an iASP for that guy is not an issue. PowerHA is made to work with storage area networks, which are designed to be multi-platform external storage systems. It doesn’t make a lot of sense to build i-exclusive SANs. The best reason I can give is because internal storage will provide much better performance, unless you spend a lot more money building the SAN.”
Another consideration to factor into PowerHA deployments is that companies using the Integrated File System (IFS) for Web serving run into more work because moving a Web server into an iASP requires changing the links in all the applications.
Flex Not Built For SMB
Several managed service providers I’ve talked with have brought up the potential for using the IBM Flex System to serve their customers. No one is using Flex for that yet, as far as I know.
Bolhuis says he has no customers that are a good fit for PureFlex. One reason is because PureFlex is too large for customers that will likely never need more than three processors for IBM i.
We’re still hoping for a smaller, less expensive PureFlex system to be released.