A 'Hot' New Option for Dealing with SSDs

For a while now customers have been asking me how to identify the filesystems and physical and logical volumes that are good candidates to move to solid state drives (SSDs). Given these drives' I/O characteristics and the higher prices they currently command, you should be selective about the data you move from traditional drives to SSDs.

Obviously you don't want to put seldom-used data on these pricey new drives--just leave that on your older, slower disks. Your SSDs should house critical, heavily used data. Fortunately, figuring out how to get the most from your SSDs just got easier.

The latest AIX releases--5.3 TL11 and 6.1 TL4--include new options that can help you determine the data that's best suited for SSDs. I updated some of my machines so I could test out these new options, and they seem to work as advertised.
 
The key new option is found with the filemon tool and its new flag, -O hot, which creates a report that shows you your most frequently accessed data.

To receive my filemon output, I ran:

filemon -O hot -A -x "sleep 20" -r fmon -o fmon.out

Incidentally, -O hot must run in conjunction with the -A flag to enable automated offline mode. It cannot run in realtime mode. I know this because I tried and received this error:

filemon -O hot  -o fmon.out
hot option not supported in realtime mode.

Once I had my output file, I ran trcrpt:

trcrpt -r fmon.out > fmon.rpt

Then I viewed a report showing me hot files, hot logical volumes, and hot physical volumes by running: more fmon.rpt

I intentionally ran the trace ran at a time where my disks were heavily utilized. Actually, it may make sense for you to run these traces several times, especially during peak workloads. This will give you a good idea about which files and physical and logical volumes may be good candidates for moving to SSDs.

This data will greatly help you make informed decisions about moving filesystems and physical and logical volumes to SSD.

While I do regularly work with and speak to customers, SSD is fairly new technology, so there's plenty more to be learned about how people are using these new drives. So please enlighten me and your fellow readers by sharing your experiences with SSDs in Comments.

On an unrelated note, I saw these interesting articles about tech support for the International Space Station that I thought I'd pass on.

Tools Gold

AIX pros have been using VNC and Screen for a long time. Years ago I wrote about these handy tools.

For a more recent and more in-depth look at Screen's capabilities, check out this article from IBM developerWorks.

I literally use Screen every day, yet I'm amazed whenever I find people who still don't know about it. I guess the methods and tools that I take for granted are always new to someone.

Another tool that I use daily is called Synergy.

I can remember showing a then co-worker how Synergy allowed me to have my Windows-connected mouse and keyboard also control a Linux machine via the network. (And, yes, the keyboard was a Model M. Of course it was.)

I had two systems, each with a monitor attached. Using Synergy, I can move my mouse seamlessly between the desktops, and even cut and paste between systems. This can work with multiple Windows machines, Linux machines or combinations of each.

I had a chuckle a few years later I visited that co-worker; he was still running Synergy and had been spreading the word. Now others in his office were using it, too.

I always like hearing about how people use their systems. How are they being backed them up? How many monitors are there, and are they extended desktops or different machines, with each running synergy? Are vnc and screen being used? What different things are being done on the command line?

For chatting, I like pidgin.

From Wikipedia:

"Pidgin (formerly named Gaim) is a multi-platform instant messaging client, based on a library named libpurple. Libpurple has support for many commonly used instant messaging protocols, allowing the user to log into various different services from one application."

For a good ssh/scp/sftp client on Windows, there's PuTTY, which, as explained here, is "a free implementation of Telnet and SSH for Win32 and UNIX platforms, along with an xterm terminal emulator. It is written and maintained primarily by Simon Tatham."

Whenever I'm forced to use a Windows workstation, I rely on several options that will give me a UNIX-like experience, including cygwin, VMware (also here) and Damn Small Linux (DSL).

As explained here, DSL is "a very versatile 50MB mini desktop-oriented Linux distribution. Damn Small is small enough and smart enough to... boot from a business card CD as a live Linux distribution (LiveCD), boot from a USB pen drive, boot from within a host operating system (that's right, it can run *inside* Windows)."

Which tools make your life easier? Please share your experiences in Comments.

More From the Grab Bag

Every now and again I like to present a grab bag of links that I find to be interesting, even if they don't warrant their own post.

For instance, some weeks back I wrote that more people are talking about AIX on Power and its industry-best downtime numbers.

Here's more on that topic.

"Among the customers surveyed by ITIC, IBM's Power Systems running AIX experienced (this includes older System p and pSeries iron) the least amount of downtime per year, when averaged across all customers using these platforms. AIX shops reported an average of 0.42 Tier 1 incidents per year and 0.34 Tier 2 incidents, and not one customer reported a Tier 3 outage on their AIX boxes. The Power Systems machines (and this includes older System i and iSeries iron) had an average of 0.56 Tier 1 outages per year, 0.44 Tier 2 outages per year, and 0.12 Tier 3 outages. So in 2009 at least, the i platform fared a little worse than the AIX platform running on Power iron."

I also liked these recent articles:

* From ComputerWorld:

"The National Oceanic and Atmospheric Administration (NOAA) system, which went online in August, is compromised of 'Stratus' and its backup, 'Cirrus,' two separate supercomputers with about 5,000 IBM POWER6 chips running AIX operating systems."

* IBM developerWorks has this piece on a script that can gather information about your systems. Then you run the diff command to find the differences between them.


* Here, the reuse of SAN disks is examined:

"The support of these scenarios in which remapping, copying, and reuse of SAN disks is allowed and supported has never been officially documented. There have been some documents and IBM Redbooks that have claimed support for specific scenarios, but they do not list the specific steps or restrictions. The scenarios detailed here guide systems administrators through the steps taken to achieve the specific environment desired. They also attempt to explain why the setup must be followed to achieve the desired results. If the steps are not followed, in some cases the system may not boot."

* Here's an IBM Systems Magazine Web Exclusive on using the mkramdsk command:

"One of the most common concerns for system administrators is maximizing disk performance. The AIX command mkramdisk is ideal for producing very high speed I/O by letting the memory do all the work. Database administrators are well aware of the benefits of keeping frequently accessed data in memory in order to reduce the need to retrieve that data from disk. The AIX mkramdisk command allows system administrators to create memory-resident file systems. The performance benefits of using RAM disk can be astonishing. The unload of a large TSM database was reduced from 40 hours on SAN disks down to 10 minutes using RAM disk."

* Finally, here's something to look forward to with the POWER7 processors:

"IBM's Hot Chips presentation on its forthcoming 45nm POWER7 server processor had a wealth of information on the chip… POWER7 will come in 4-, 6-, and 8-core varieties, with the default presumably being the 8-core and the lower-core variants being offered to improve yields."

If you've found interesting articles online, please share your links in Comments.

Migrate When the Time is Right

I know that many of you have been upgrading from older technology to POWER6 servers. As one customer recently told me, "The machine is working great. Performance is better than I expected." That seems to be a constant theme: people are happy with the newer hardware when they get it deployed.

Still, I'm sure that many others have been delaying their upgrades. They figure that POWER7 will be here soon enough, so why not just put that upgrade on hold for a while?

As I noted, IBM recently issued a statement of direction that I consider to be good news for these customers. If you've been debating whether to order a POWER6 server now or wait until POWER7 comes out, you can have the best of both: POWER6 performance now, and POWER7 when it ships.

From IBM:

"IBM plans to provide an upgrade path from the current IBM Power 595 server with 12X I/O to IBM's next-generation POWER7 processor-based high-end server. The upgrade is planned as a simple replacement of the processor books and two system controllers with new POWER7 components, within the existing system frame. IBM also plans to provide an upgrade path from the current IBM Power 570 server with 12X I/O to IBM's next generation POWER7 processor-based modular enterprise server."

More details are emerging about POWER7 processors.

"'POWER7 is an 8-core, high performance server chip. A solid chip is a good start. But to win the race, you need a balanced system. POWER7 enables that balance...' Starke noted that the POWER7 offered 'multiple optimization points,' such as improved energy efficiency, upgraded thread performance, dynamic allocation of resources and an 'extreme' increase in socket throughput. In addition, the POWER7 provides scalability up to 32 sockets, 32MB on chip eDRAM shared L3, dual DDR3 memory controllers, 100GB/s memory bandwidth per chip (sustained), 360GB/s SMP bandwidth/chip and 256KB L2 per core.

Also:

"Basically, the POWER7 is an 8-, 6, and 4-core chip with 1.2 billion transistors, running at an undisclosed clock speed. A shared L3 cache of up to 32 Mbytes in size will use eDRAM. The POWER7 will scale up to 32 sockets and 1,024 threads. Not surprisingly, it will be backward-compatible with the POWER6.

"Not surprisingly, the performance of the POWER7 exceeds the POWER6 by a significant amount, although IBM has left off actual numerical comparisons. Application comparisons such as integer workloads seem to indicate an improvement of about 20 percent across the board on a per-core basis, and a 4X to 5X performance when compared chip to chip."

Hardware constantly evolves, and every organization needs to evaluate the tradeoffs when comparing performance to the costs of acquisition and running new machines. Keeping up on all of the benefits of POWER6 and POWER7 compared to older technology that you might be running today can be a challenge. However, by creating an upgrade path that allows your organization to migrate when the time is right, IBM has made it easier to protect your investment.

Changing the padmin .profile

Recently I worked with a customer who was frustrated with the VIO server. When he logged in as padmin, he was faced with a $ prompt. When he ran oem_setup_env and became root, he had a # prompt. Whenever he switched users, he had to remember to run set –o vi to get access to his command history. He would log into two VIO servers and couldn't easily keep track of the one he was working in. For all of the other machines in his environment, he had customized his .profile to his liking. He customized his prompt and the name on his title bar, and would automatically run set –o vi, etc. However, those oddball VIO servers weren't playing nice.

I know that we're supposed to treat the VIO server like an appliance. We're supposed to set it up and let it run. But in this instance, we were still in the middle of setup. We were mapping LUNs to different vhost adapters for different clients, and we were going into oem_setup_env to install additional drivers.

Often I see customers set up their .profile with things like set –o vi, or customize their prompts with things like the user ID they're logged in with, their current working directory, their hostname, etc.

While you can certainly make changes to the /home/padmin/.profile file, they will usually go away whenever you upgrade your VIO server. That's fine; since we're supposed to treat it like an appliance, I understand that we shouldn't change things. In reality though, I continually find customers who alter the padmin .profile.

Even when customers change the .profile, those changes don't carry over when running the oem_setup_env command. For the sake of usability, something had to give in this case. Whenever we ran the oem_setup_env command, we had to run manually run /.profile. This customer was getting fed up.

After asking around for a more elegant solution, someone reminded me to just use a .kshrc file.

I was told to create a /home/padmin/.kshrc file. I put set –o vi and my other desired prompt settings into that file. Then at the end of the /home/padmin/.profile, I added:

export ENV=/home/padmin/.kshrc

This allowed the customer's desired environment to be set up automatically when he logged in as padmin. Then when he switched over to root by running the oem_setup_env command, that was set up as he wanted it.

Yes, I realize that .profile and .kshrc will have to be revisited when this customer upgrades his VIO server. But in the meantime, this technique allowed us to customize the environment to his liking. It helped me keep a customer happy. Maybe it will make you happy as well.