Understanding The SolarWinds Value Proposition #NFD6

1,229 Words. Plan about 5 minute(s) to read this.

SolarWinds, for those that don’t know, is a company selling a network management system (NMS). NMS platforms are the bane of the networking industry, ranging in capability from single-purpose applications designed to manage one specific piece of gear to full-fledged framework applications that purport to manage everything you can imagine…and probably more if you write a big enough check.

I both love and hate NMS’s. I love the idea of taking data supplied by a network device and presenting it in some useful way to network operators, IT managers, and executives. In principle, it sounds so right. In execution, most NMS’s are fairly weak presenters of information, rendering the same graphs of bandwidth utilization and error counts that every other NMS seems to render, but never really making the information useful. There’s never any context.

Certainly, there are some standouts in the crowd, and yes, it’s possible to customize an NMS to your liking in many cases. And if you really want to go custom, the open source community has a bevy of packages and libraries so that you can roll your own NMS nirvana. $Deity help you sort out the MIBs and OIDs, but just sayin’ – you can do it if you like.

But let’s face it – most of us don’t have that kind of time. We need an NMS that out of the box just works. We want a network management tool that gives us useful information about our network in a hurry. So let’s switch our focus back to SolarWinds, where they’ve got a really strong “it just works” solution.

SolarWinds as an NMS has, in my opinion, the following high-level features to commend it:

  • A straightforward web interface presented in a columnar format populated by web parts. Any page is customizable by changing the number and width of columns and the web parts that are present in each column.
  • A modular framework that allows companies to augment the system’s capabilities by purchasing and installing new SolarWinds modules. For instance, the core “Network Performance Manager” product can have the “Network Configuration Manager” module bolted on to add automatic backup of network devices to the system. When installed, additional modules have the same look and feel as all the other modules, so the system feels unified. This is no small accomplishment considering that SolarWinds has grown largely by acquisition.
  • A robust set of SNMP MIB knowledge. SolarWinds works hard to integrate the MIBs of devices requested by their customers. For example, I have had good success monitoring Cisco Catalysts, Cisco Nexus gear, Cisco wireless gear, F5 LTMs, Riverbed Steelheads, and Check Point firewalls. Of course, it can also monitor a variety of operating systems such as Windows, and can hook into VMware hosts as well.
  • If you want to monitor custom OIDs, there is a “Universal Device Poller” tool that allows you to build your own object to poll.
  • The “Network Atlas” tool allows you to build a live network map, including live links where you can even embed the link utilization right into the map. The resulting maps aren’t a pretty as you might generate with Visio, but…live maps! That’s a great feature.

On the downside, SolarWinds as an NMS has a few rough spots.

  • The platform only runs on Windows. SolarWinds is not an appliance, nor does it ship as a canned virtual machine. Yes, you can certainly run it virtually (and SolarWinds has a sizing guide to help you do just that), but no matter what, you need to have Windows underneath. If I need to explain why Windows as a server platform in this context stinks, then this point isn’t a worry for you.
  • Performance is SolarWinds’ weak link. And by “performance,” I mean that the interface is slow to use – frightfully slow at times. Much of this is tied to SQL, where SolarWinds keeps almost all of its data. It’s easy to load up a SolarWinds system with lots and lots of database transactions when polling many devices & interfaces, especially if you add the netflow collector module.
  • Unrelated to the NMS specifically but worth mentioning is that the SolarWinds sales process is a wretched evil, easily one of the worst I have ever dealt with in IT. There is no assigned rep that knows your account. If you have more that one group in your company doing business with SolarWinds, SolarWinds will almost assuredly get the separate accounts confused, fix them for you, then confuse them again at the next service renewal. Account reps are relentless at renewal time, blasting every company contact with repeated renewal requests and phone calls, even after a PO has been sent over, and even if you’re the wrong person to be contacted about the purchase and even if you explain this to them, repeatedly. I got smart after a while and just stopped picking up inbound phone calls from Austin. Yes, that was a whiny aside, but I bring it up because several other people I’ve run into have had this same sigh-inducing treatment. Believe me, it’s a frustrating enough process to make you rethink your career choice in networking sometimes.

At Networking Field Day 6, SolarWinds had a conversation with the delegates regarding network virtualization, along with a mention of the performance issues. I think I can sum up SolarWinds’ position on network virtualization this way: they’re aware of it, but aren’t in a rush to implement fancy new monitoring to handle it. Why? Because of SolarWinds target market, which is the small & medium size business. SMBs aren’t rushing into network virtualization. Heck, most SMBs don’t even know what it is or what they’d do with it if they did. That being the state of SolarWinds target market, they can choose to put their energies in other directions.

The direction I was pleased to hear about was the “performance improvement” direction. Specifically, SolarWinds is rewriting the database backend for the Network Traffic Analyzer (NTA) component so that it no longer relies on SQL. NTA is the netflow collector I referred to earlier, and it definitely generates a ton of SQL traffic that burdens the overall system performance. The new NTA database will (a) not be SQL and (b) be optimized to handle netflow data structures and the sorts of queries that are made against those tables. This is no small feat, as the finished product must not only work as advertised, but also must successfully and consistently import existing customer data.

Whatever criticisms I might have expressed about SolarWinds, I’ve used the product a lot over the last 3 years. I know it reasonably well. And I can say that it is a useful NMS for smaller sites. Could it be better? Sure. Would I buy it again? I’d certainly consider it, despite having no love their sales process. Will the performance improvements to NTA fix everything once SolarWinds gets them done? I doubt it will fix “everything.” But it will be a good step in the right direction.

Disclaimer: I attended Networking Field Day 6 as a delegate, but was not compensated to attend. I am free to write (or not write) about any of the presentations as I see fit without fear of censorship by the vendor or Tech Field Day. My approach is that if I was inspired by a vendor presentation enough to write about it, I will.

Ethan Banks writes & podcasts about IT, new media, and personal tech.
about | subscribe | @ecbanks

2 thoughts on “Understanding The SolarWinds Value Proposition #NFD6

  1. Ethan,

    Regarding SolarWinds, what do you mean “useful for smaller sites”? What size of sites?


    1. Businesses that would fall into the “small & medium” category would be fine with SolarWinds, and that’s their target market. If you are large enterprise, the SolarWinds platform gets difficult to scale, although it’s possible. SolarWinds has the ability to separate polling tasks into multiple engines, offload SQL to a separate box, and I think offload the IIS engine to a separate box.

      But realistically, if you end up polling many thousands of elements (i.e. an interface is an element), the SolarWinds engine struggles to get information out of the database and build a web page. It’s just not a fast performer under a heavy load.

      In addition, the SolarWinds platform can handle many events and react to them, so that you can have it do things like parse syslog, manage SNMP traps, take in netflow data (if you bought the NTA module), and much more. The more you layer onto the system, the more you’re asking of it, and the slower is seems to get if you have a lot of devices that are a part of the system.

      For a more concrete example, I ran a SolarWinds platform for about 3 years with the core NPM module, and added the NTA (netflow) and NCM (config management) modules. I managed roughly 120 devices with it, ranging from simple Cisco routers with 2 or 3 interfaces up to stacks of 3750X switches with 400+ interfaces. The SQL backend was its own box, and the main SolarWinds box handled polling, web services, syslog, SNMP traps, alerting, etc. It was…okay. Slow for sure, but not unusable. It was right on the edge of frustrating to use. It was enough of a headache that we were looking at building out a second polling engine and wondering if it was possible to move NTA to a different box, but still have the integration. We were also looking at Cisco Prime as an alternative, since we had a few Prime pieces in house already.

Comments are closed.