-1

Thanks to this article (in Norwegian), I started researching a bit about the new Intel Solid-State Drive 710 Series, which is a MLC with about 30 times improved endurance. It seems like they are really making advances. My question is, therefore, whether I can expect significant improvements in read/write performance of file servers in large companies any time soon? In other words, can this technology really convince the administrators of these servers that it beats high-end hard-disks on all levels? If not, which is the critical spots that the SSDs still need to crack to start taking off in this context?

The reason why I ask is because this could significantly change the way file servers are used.

6
  • the answer is "it depends" is content being served by the file cache or not and if not why not? The basic cifs protocol will probably be the limiting factor in a true file server situation. Now if this was an sql server etc then things may change
    – tony roth
    Sep 18, 2011 at 20:07
  • @tony roth Maybe I am using the wrong word, when saying "fileserver". What I mean is the server in a traditional large company, which contains all the .doc, .xls and .ppt, files which its employees share with eachother. Typically this server is named x: or similar. Is there a more specific word for this than fileserver?
    – David
    Sep 18, 2011 at 20:10
  • 1
    It still doesn't win on one of the most impotant factors. cost/MB. For many applications size is far mor important then speed.
    – Zoredache
    Sep 18, 2011 at 20:15
  • @Zoredache You are completely right. The article, though, states that the cost of the 300 GB version is only 1929 dollars, which does not seem frightening. That means that it is quite cheap per .xls file.
    – David
    Sep 18, 2011 at 20:18
  • you'll be limited by other factors, on a pure fileserver the disks are rarely the limiting factor. And you are using the word correctly!
    – tony roth
    Sep 18, 2011 at 20:29

1 Answer 1

1

Disk throughput and latency are not often an issue on file servers serving up mapped-drive data (documents & images). It's fairly easy to scale a file server's disks up to accommodate more concurrent users (by adding disks to an array, using junction points to add additional arrays, splitting shares), and scale them out across multiple file servers (with DFS, junctions and a few other tricks). Usually you hit a limit in network congestion or latency before you hit a serious problem with disk throughput and latency.

However, there are much faster use-cases for enterprise storage which are already around and have been in use for many years now. These are iSCSI and Fiber-attached Storage Arrays, and are capable of serving data for much faster usage than is needed for simple file servers. Typically, they are used to concurrently serve data on a dedicated storage network to multiple servers concurrently. Those servers could be fulfilling nearly any function, such as email, file & print, web, database or application hosting.

The impact that SSDs are having on these systems is increased availability of automated Tiered Storage capabilities, even in relatively low-end storage appliances. Administrators will usually purchase a few SSDs and add them as the highest storage tier, effectively utilizing them as a large cache.

1
  • I do not see how my question could deserve to get closed, when I got such a good answer! Thanks!
    – David
    Sep 18, 2011 at 21:07

Not the answer you're looking for? Browse other questions tagged .