Which SSD is best for backtesting
- Thread starter ideabox
- Start date
The only SSD I've ever owned is this:
http://www.newegg.com/Product/Product.aspx?Item=N82E16820167040
I run it in my netbook in three equal boot partitions (W7x86, Ubuntu LTS (Currently in 12.04 LST BETA) and Windows 8 DP or other trial OS.
EDIT: that's not entirely true, I've bought & installed a bunch but that's the only SSD I've ever had in any of my own personal machines.
I run the SSD in the netbook because it extends battery life. For $200 I didn't care about anything other than energy consumption - even if it was slower than my spinny HDD I would have installed it.
ideabox - you have to answer your own question and think about it. If you are storing historical data on the SSD then your back test engine will be reading from it right? If this is the case reads will be much more important than writes.
How will your database be structured? Sequentially tick by tick or by header or ticker? If you store each message sequentially only by timestamp vs. by ticker then timestamp then sequential reads would be more important than random.
I really can't answer these questions for you - you need to tell us how the database will be setup and how the backtest engine will access the data. If you can explain that to us then I'll bet you can answer your own questions.
My gut tells me that you don't need a SSD at all - perhaps a small one for a boot/OS partition and then a decent size spinny HDD for your data.
The reason why you didn't get an answer is because only you can answer these things. SSD speed isn't going to help you out at all when running ninja-trader. The average seek time on a large platter 7200rpm HDD will be a fraction of the time it takes you to send an order or connect to your broker. It's so trivial it isn't worth thinking about.
This is my understanding, as well. The first data access will be against the drive, so you will get some performance improvement using the SSD for the first pass against the data. For subsequent access, the data will be resident in memory and the drive is not a limiting factor. So the more important design factor for backtesting is to use a 64-bit OS with a lot of physical memory.
My personal view is that the two greatest benefits of SSD's are limited to cases where 1) you are constantly turning off and on your computer (for some reason) and the boot time is quicker when your OS is on the SSD, and 2) your laptop is subjected to a lot of physical movement because you are mobile (not just sitting at a desk) since the SSD drive is less vulnerable to physical shocks. There are other benefits of SSD's, but they are largely offset by the drawbacks (e.g., more expensive, smaller capacity, requires periodic flash updates to controller, no TRIM support in a RAID array).
Actually SSD have a ton more advantages bu you need to have a scenario in which they work.
* Databases are great on SSD. An SSD these days is 100 to 150 times faster than a nomral disc and database systems are regularly mostly limited by IO per second (IOPS) budgets.
* Virtualization
I run a smaller virtualization server using 8 Velociraptor in a RAID 10 as discs for the virtual machines and at tiems they are OVERLOADED. Like really badly overloaded. Adding a pair of SSD as second level cache (for now: keeing a common base image, in 2 months: serving as RAID card integrated read and write cache) really takes a ton of load from the system. Goes down to IOPS.
* Same for development. Compilers / Visual Studio are often very imtied by IOPS - accessing tons of files, generating tons of files. C++ is particularly bad here with all the header files and the linking in a second step
I think by now I have 7 or 8 120gb OCZ Vertex 3 in use and I love them.
* Databases are great on SSD. An SSD these days is 100 to 150 times faster than a nomral disc and database systems are regularly mostly limited by IO per second (IOPS) budgets.
* Virtualization
I run a smaller virtualization server using 8 Velociraptor in a RAID 10 as discs for the virtual machines and at tiems they are OVERLOADED. Like really badly overloaded. Adding a pair of SSD as second level cache (for now: keeing a common base image, in 2 months: serving as RAID card integrated read and write cache) really takes a ton of load from the system. Goes down to IOPS.* Same for development. Compilers / Visual Studio are often very imtied by IOPS - accessing tons of files, generating tons of files. C++ is particularly bad here with all the header files and the linking in a second step
I think by now I have 7 or 8 120gb OCZ Vertex 3 in use and I love them.
Price / Benefit analysis and jsut the fact that the are new
Sorry - at this point it makes little difference what you choose, SSD are so far ahead of the pack That said - I paid a heavy price. They were unstable like hell, dropping out of RAIDs regularly, taking down hole arrays (because it not answering also blocked the SAS bus that connected the discs). Took some firmware updates until they go where they are now, which is flying away and being stable for months. And quite fast
Quote from NetTecture:
Actually SSD have a ton more advantages bu you need to have a scenario in which they work.
* Databases are great on SSD. An SSD these days is 100 to 150 times faster than a nomral disc and database systems are regularly mostly limited by IO per second (IOPS) budgets.
* Virtualization
* Same for development. Compilers / Visual Studio are often very imtied by IOPS - accessing tons of files, generating tons of files. C++ is particularly bad here with all the header files and the linking in a second step
I think by now I have 7 or 8 120gb OCZ Vertex 3 in use and I love them.
6 days ago I just bought and implemented my first set of SSDs in all of my VM hosts. I can't even tell you how awesome it is.
I bought CHEAP Patriot 32gb and 60gb SSDs - that I hope to burn out before the warranty expires
I'm a huge fan for VM/Datacenter environments. To the point that I regret not getting them sooner.