Quantcast



Docs


Blogs


Forums


Samples


Media


Labs


Resources

 




DevCentral > Weblogs > Persistently Different - Not right, just different.
 Storage? We bought Storage?
posted on Monday, August 13, 2007 1:09 PM

Okay, having been the Storage and Servers Editor for the (now defunct) pre-eminent networking geek magazine, I've been feeling like I should say something about our purchase of Acopia Networks. The reason I haven't yet is twofold -

1. I found out we had made an offer when you did (or could have), and by the same manner you did - Mr. McAdams' press release.

2. I wanted to understand how Acopia "fit" with us, and how that effected DevCentral.

I can (and have!) blog for days straight about storage, its market, and the special issues of that market. Just my belief that smart enterprises don't want or need multi-tiered storage in the age of RAID and replication, and my further belief that the storage industry fails when they don't understand that fact could take up pages and spawn some interesting conversations, but I'll keep this on-point for the most part.

In the long term, we at F5 are working on building even higher reliability. Lori was right in her blog post in that sense, but I'm going to take that and run with it from a DevCentral perspective.

Most people view the importance of iRules and iControl to be in allowing customers to do things that networking vendors can't foresee, things that are focal to your architecture and your organization. That's true, but I like to add the tiny little caveat on there that reducing your workload is at least as important. IT staffs are small and networks seem ever bigger. Anything to avoid one more call or cut one hour out of the week for you all is a worthwhile endeavor, in my opinion. Couched in terms of "one less body" or "increased productivity", it's worthwhile to management also. And that's a balance we have to walk, right? The job has to get done, but you can't get burned out doing it.

So my focus is on the future. I can't talk to the folks over at Acopia until the deal closes, so this is just me, knowing both spaces, offering you possibilities... Things we could be doing in a relatively short time.

It's two AM and a server goes down with errors on its NAS. BIG-IP takes it out of the pool, and you get a call, right? But what if iControl could get notified instead, and an iControl application automatically points the server at the replica that Acopia has been producing. Or better yet, an iControl App tells Acopia to move the mountpoint to the replica so that nothing changes on the server. You stay in bed, the system stays at peak efficiency, on Monday morning you go figure out what happened to the NAS. Everyone's happy.

It's your day off and there's a rush on your servers. Seeing too much traffic, they start to struggle to keep up. You get a call. Or an iRule notices the problem and tells iControl that it needs another server. iControl pulls a machine that is online but not associated with any of your web applications up, points it at the NAS storage that this web application uses, then adds it to the pool for the application in question. Load reduces. Blade servers do this sort of thing within their frame - Egenera has had this capability within their BladeFrame for quite a while now... But only within the frame. BIG-IP could do it for your data center.

Throughput on a given pool has dropped so that servers are seeing a very low utilization for that pool, but another web application - served by a different pool - is struggling to keep up. An iControl application takes one member out of the pool that has low utilization, points its NAS connection at a different Acopia mount point, and adds the machine into the overburdened pool. Self adjusting resources, brought to you by F5 with Acopia.

These are really just three examples of a self-aware, self-healing network that Acopia helps us enable. The next while on the DevCentral team should be fun as we figure out even more ways to make these two technologies work together to add value for you.

I enjoy storage nearly as much as I enjoy F5, so for me this is a great challenge, and I'm looking forward to figuring out ways to make your life easier. Frankly, my time in IT and IT Management left me with an "if it can be automated reliably, automate it" attitude, and that's where I see Acopia helping out a lot. We can do a lot of things, but like Lori tried to say, your webapps aren't any more reliable than the disk they run on. Unless "disks they run on" can be expanded to include several NAS boxes, some RAID and some Replication... RAID is handled by the NAS vendors and smart folks like you, I guess we'll get the rest ;-).

 

That's it for now, have some other items to look in on.

Don.

Reading: Re-reading Panzer Aces II, Franz Kurowski

Imbibing: Mountain Dew



Email This
  del.icio.us
      

Feedback


8/18/2007 4:01 AM
Gravatar You are dreaming! If only it could be true.
Storage Guy

8/18/2007 11:17 AM
Gravatar So as a consumer of such storage let me explore this just a bit further, and equate this back to business dollars, after all that is why all this get purchased in the first place. Yes, as an engineer, the thoughts of a peaceful and restful night are truly utopia, but even better not having to stand on the carpet because a system you engineered failed.

Acopia products are so compelling to the NAS world, this is a powerful and open standards based product. What make this so important is the NAS, or file level storage, is able to keep up with it's more expensive big brother, Fibre Channel. Virtualized Fibre Channel is a common place now with offering from the infrastructure vendors like Brocade, and Cisco, software solutions by Symantec/Veritas, and array based solutions from Hitachi and EMC, just to throw some big names out there. The power of virtualizing thousands of LUNS under one single pane of glass is a storage architects dream. After all it not the failures that end up killing us, it's the management. Back in the good ole' days I remember managing 15 different arrays independently, but spreading a single host across all 15. Spreadsheet, home LAMP based management, and a huge amount of "oops I forgot about that one was going on"

Now NAS comes to age with File based virtualization. For those of us that have managed file servers, shares, users directories, NAS heads, and the like it was a full time job. In the tune of 500-1000K users per head count. As a technology decision maker, that keep the lights on cost was painful. And you saw it creeping in year after year after year. I have colleuges in other companies that are seeing 75-100% growth YOY on all storage. HUGE EXPENSE.

Enter Acopia, yes we still have to buy it, but the access point becomes a single pane of glass. Need access the finance share, just \\servername\finance and TADA you files are there. It does matter that IT has been scrambling to move data around to different costed tiers because Alice in marketing just put up 100GB in HiRes Tiff's for the June advertising glossy. It doesn't matter that you data just go moved from NYC to St. Louis because that is where the most space was. To push Dan's theory just a bit further, WANJET kills it here. With the power that we have with CIFS proxy, it really doesn't matter now, where the backend storage is. To the end user they might see another 3-4 seconds, but that is close enough to LAN times that it will be more of a wonder than call the support desk.

I have been preaching around the Country and the Southeast for awhile now, and the biggest thing that I continue to say, is for True Application Resiliency you have to have a view from the click to the disk, period. This will give us that view. And the ability to modify it as needed depending on the needs of the consumer of the application.

It's almost like the final piece of the this big puzzle has been put into place. We know have all the major pieces of the ADN column in our portfolio; Storage, Network, Server, Acceleration, and hooks into the very applications that we manage.




Brad Wood

8/20/2007 11:58 AM
Gravatar Wow, self responding readers! :-)

Actually, there is no place to start a thread for this on a forum, I should find one.

It can be true, Storage Guy. One of my big complaints when Editing for NWC was the failure of the storage market to account for the network they impact so heavily... This is a chance to rectify some of that.

But what can be done, and how much will depend upon how the two technologies mesh, which we won't know really well until after the deal closes in September. We know they compliment and can work together - they're in use together - but the advanced stuff I'm talking about will take some looking into.

Don.
Don MacVittie

8/23/2007 8:14 AM
Gravatar I would think so, we don't have a magic converter box to make all of this work. What are you guys doing in seattle.... :D
Brad Wood

8/23/2007 10:43 AM
Gravatar No converter box? Darn, that means we'll need a budget... ;-)

Ah ha! I'm not in Seattle - Green Bay, Wisconsin is home to me. Conveniently located near neither coast.

Don.
Don MacVittie
 Leave Feedback
Title  
Name  
Email
Url
Comments   
Please add 7 and 7 and type the answer here: