System Admins. R.I.P.?

January 20 2011


The advent of the cloud has brought a wave of changes in IT from the way we provision servers to the applications we use. At MuleSoft Google Apps provides email and office productivity services, Salesforce provide us with sales, support and product tracking, Marketo provides our marketing automation services and Atlassian gives us issue tracking and development productivity tools. All of these applications are hosted by our services providers, so what does our trusty sys admin do? Well, we will always need application infrastructure right…

A couple of years ago building an application meant building the infrastructure to support the application. Typically this entailed installing a database, an app server and often a messaging server and email server. All tedious preamble for actually building an application. Now you have free access to these infrastructure pieces on the web – yay cloud!


As NoSQL providers gain widespread adoption some are now offering hosted versions making it super easy to add storage capabilities to your weekend application. Typically NoSQL datastores are easy to set up, but now there are a few free hosted solutions, making it even easier to get started.

MongoHQ – fully featured document- based storage.  no fuss.

– Hosted CouchDB.  Couch is pretty interesting as it has a mobile storage offering too.


The daddy of cloud infrastructure services, AWS have offered S3 for a few years, complete with zoning for more efficient data access, but there are others gaining popularity.

Cloud Files – from Rackspace, this service seems to be a very similar offering to Amazon S3.

Cloud Storage – from GoGrid offers mountable storage with a REST API.

Cloud Layer Storage – from SoftLayer offers simple storage for a variety of of clients and CDN support.

Also, there are options like DropBox and which are more consumer focused but have REST/SOAP APIs also.


This category is set to get interesting this year. Amazon led the charge here with SQS (Simple Queue Service) . While falls short for anyone coming with enterprise messaging expectations it is ok for casual messaging. Now there are more interesting options.

PubNub – super simple and super fast, is becoming the messaging service of choice for gaming, thanks to it’s impressive high throughput and low latency. It has huge potential outside of gaming too.

Amazon SNS – Amazon Simple Notification Service is a pub sub messaging system that compliments the AWS SQS service. Amazon SNS seems aimed at casual notifications that can be published over different channels such as email or HTTP.


Postmark – is a really easy way to create virtual SMTP servers that have a simple REST API. The advantages of this may not seem obvious since you can set up a Gmail account and send mail till your hearts content. But postmark offers a REST API (anyone that has used JavaMail will appreciate this) as well as an SMTP interface and the service provides delivery monitoring (good for transactional processing) reverse DNS, whitelisting and more.

With all this infrastructure available to developers with a few clicks, I wonder what our system administrators will spend their time on?

*I have definitely missed some providers here.  Please add anything you think I should have covered in the comments.

Follow: @rossmason@mulejockey

We'd love to hear your opinion on this post

33 Responses to “System Admins. R.I.P.?”

  1. I’d like to add:

    – Xeround, the up and coming MySQL in the clouds:
    – Hoptoad, awesome error management console:
    – Loggly, logging as a service:
    – Scout, monitoring in the clouds:
    – Twilio, awesome Saas for communicative applications:

  2. The major ones to add would be services like Heroku, NodeJitsu, and AppHarbour.

  3. I wonder… who manages the Cloud? SYS ADMINS.

  4. I have to agree with anon. In certain situations you are correct, it only takes a few minutes to setup a fully functioning business. In some cases (with rightscale or similar) everything may be automated. I think most companies right now can’t use document databases (nosql), so you still have database maintenance t deal with. Also backup is much more then having yesterday’s server image.

    Then there is the dns, the server monitoring, the daemon monitoring, I happen to have a VPN nightmare to deal with.

    Our company currently doesn’t have a sys admin because we are on the cloud and have no servers. So what has ended up happening is I have to deal with all of the cloud system administration headaches.

  5. sysadmins run all those services you are talking about… and the margins are huge

  6. Exactly what anon said.

  7. lol, the sys admins will be waiting for all the cloud computing to migrate back in house (at great cost to the businesses) once it is realized how poor the current cloud technologies are.

    just wait until amazon ec2 ip blocks are blackholed by all major isps due to the rampant abuse from their nets.

    enjoy this bubble while it lasts.

  8. this article has no grounds and makes no sense. you are a moron. first commenter is right. Who will manage your “cloud”? I sure hope you don’t manage anything. asshat.

  9. I have to laugh, to think that most companies of any noticible size could get away with the list of things you have provided here is laughable, there is almost always glue script to hook between most infrastructure components.
    As an example, what manages accounts between
    Google Apps, Marketo, Sales Force and Atlassian….oh right manually…thats awesome for about 5 people.

    I could go on and on about other aspects of how flawed this logic is but i wont.

    In short cloud services have their place but its definitely not a replacement in most situations.

  10. We’re deploying a new outsourced infrastructure and finding that the system and network admin skill set is wanting. Every network related problem is being “bumped up to Cisco” and everything hardware related either takes our service vendor days/weeks/months to resolve, often via “change all the tires and see if the engine will start” methods, or we end up having to resolve it ourselves. Tasks that we’ve taken for granted before (firewalls between tiers, VPN setup, cluster interconnects, SAN configuration) seem to be extremely fault-prone and often result in outages elsewhere due to ham-handed change control and lack of professional responsibility.

    I think the sysadmins _have_ died and moved on and we’re left, for the most part, with inadequate substitutes. I don’t know if it’s going to get worse or better. Someone has to do the work, as the abstractions are still very brittle. Unfortunately, while paying to make it SEP (Someone Else’s Problem), it bounces back to our application development and release engineering teams.

  11. is a great filesystem provider that ticks a lot of boxes: rsync-over-ssh “API”, geo-redundant, developer-friendly, clueful.

  12. As “anon” said: the sysadmin will manage the cloud. That’s my job right now 🙂

    cheers! great post 🙂

  13. You *could* send emails through Google Apps, but they have a rate limit, and it’s in the low hundreds per hour, so PostMark is a much better solution. And PostMark gives you feedback on bounces, etc.

  14. MongoDB requires a priest.

  15. Any company that uses the cloud, to me, is stating they don’t care at all for the privacy and security of their customers information.

  16. The definition of what sysadmins are doing is changing, but they won’t go away. You find more and more “devops” type roles.

    They often become important when you reach the threshold where the convenient hosted services are no longer scalable in terms of cost or the lack of visibility. They are leaned upon for the knowledge of how what the app does impacts the system and the limits it can run up against.

    One problem with some of these hosted services is the lockin you get from the convenience. I’d had to migrate out of several of these types of services and it is frustrating to undo the lockin they produce.

  17. Frank, I was saving the app platform for another post. But yes agree that less and less infrastructures needs to be managed in house

  18. I think you have confused R.I.P. with outsourced.

  19. Kris, I assume you are against SalesForce then?

  20. Art and Ryan, good points. I can see one problem with cloud services is that you delay the need to get in house help for everything else and operate by shoving problems around without ownership

  21. Anonymous, erm, SSO?

  22. I completely agree here Kris, most cloud type places pretty much have a ‘we take no responsibility’ clause in their terms and conditions.

  23. I think the real thing working against the cloud is the stability of your connection to the outside world. If you’re in midtown Manhattan, you’d probably be fine. But where I’ve worked as a system admin, in suburban areas where the ISPs available are residential providers first-and-foremost and business customers are secondary, I don’t think our IT office nor the president of the company could trust the cloud at this point.

    Could a company go all cloud at some point? Of course. But I’m not sure that infrastructure (and as others have said, integration) are fully there yet.

  24. There’s always going to be asshole who puts up roadblocks and eats away at everyone’s productivity. Just look at what Amazon demonstrated with wikileaks.

    The BOFH is far from dead.

  25. Typo: compliments should be complements.

  26. Cloud has helped sysadmins.. Samething can be said for software like Mule. With many frameworks integrating REST and other middleware components, why does one need software like Mule?

  27. System Admins. R.I.P.?
    Of course not, sysadmins are only evolving into something new and better.. cyberbofhs ! 😉
    They will be more cyberspace oriented, but still sysadmins are needed, because who else will make it work?

    Also, a lot of people forget that “the cloud” is _not_ holy grail of all IT solutions!

  28. I’m a sysadmin using EC2 to deploy and support SaaS and PaaS solutions. The ‘cloud’ doesn’t eradicate the sysadmin- it is only evolving the set of tasks a sysadmin does.

  29. Good luck letting your users solve latency/performance problems between all apps. I guess they’ll get used to “damn! the internet is slow today isn’t it? let’s have coffee and wait since we don’t have a sysadmin and the service provider guys are as clueless as us” kind of things.

  30. You know what’s worse? The same clueness when it comes to complexity that this article expects users/companies to have… is probably being used withint by SaaS providers to choose other providers that they will outsource some pieces too. So if the service you hired from a given provider is slow today he’ll blame on another provider and show you how the contract provides no guarantees. It’s the same mentality over and over.

  31. a) what everybody else said–I’m a devops sysadmin working for a big web company, and it’s painfully hard to hire, so I doubt the jobs are going anywhere.

    b) Mongo: as a part-time mysql priest, I can tell you that sql and nosql both require sacrifices way beyond schema.

  32. Also,

    There’s no cloud access to a laptop sitting in an airplane. Though granted it’s getting better, and there are ways of getting stuff offline. It still depends entirely on being connected 24/7.

  33. One more comment, perhaps the sysadmins aren’t managing “systems” anymore, they are managing integration, but then a system can be one machine, multiple machines, or an abstraction that means a bunch of moving parts. Administrating that system will never be dead.