I have a relatively big file - around 10GB to process. I suspect it won't fit into my laptop's RAM, if MRJob decides to sort it in RAM or something similar.
At the same time, I don't want to setup hadoop or EMR - the job is not urgent and I can simple start worker before going to sleep and get the results the next morning. In other words, I'm quite happy with local mode. I know, the performance won't be perfect but it's ok for now.
So can it process such 'big' files at a single weak machine? If yes - what would you recommend to do (besides setting a custom tmp dir to point to the filesystem, not to the ramdisk which will be exhausted quickly). Let's assume we use version 0.4.1.
I think the RAM size won't be an issue with the python runner of mrjob. The output of each step should be written out to temporary file on disk, so it should not fill up the RAM I believe. Dumping output to disk is the way it should be with Hadoop (and the reason why it is slow due to IO). So I would just run the job and see how it goes.
If the RAM size is an issue, you can create enough swap space on your laptop to make it at least run, thought it will be slow if the partition isn't on SSD.
Related
I am using Pentaho for reading a very large file. 11GB.
The process is sometime crashing with out of memory exception, and sometimes it will just say process killed.
I am running the job on a machine with 12GB, and giving the process 8 GB.
Is there a way to run the Text File Input step with some configuration to use less memory? maybe use the disk more?
Thanks!
Open up spoon.sh/bat or pan/kettle .sh or .bat and change the -Xmx figure. Search for JAVAMAXMEM Even though you have spare memory unless java is allowed to use it it wont work. although to be fair in your example above i can't really see why/how it would be consuming much memory anyway!
I realize this number will change based on many factors, but in general, when I write data to a hard-drive (e.g. copy a file), how long does it take for that data to actually be written to the platter after Windows says the copy is done?
Could anyone point me in the right direction to discover more on this topic?
If you are looking for a hard number, that is pretty much unknowable. Generally it is the order of a tens to a few hundred milliseconds for the data to start reaching the disk platters, but can be as high as several seconds in a large server disk array with RAID and de-duplication.
The flow of events goes something like this.
The application calls a function like fwrite().
This call is handled by the filesystem layer in your Operating System, which has to figure out what specific disk sectors are to be manipulated.
The SATA/IDE driver in your OS will talk to the hard drive controller hardware. On a modern PC, it typically uses DMA to feed the data to the disk.
The data sits in a write cache inside the hard disk (RAM).
When the physical platters and heads have made it into position, it will begin to transfer the contents of cache onto the platters.
Steps 3-6 may repeat several times depending on how much data is to be written, where on the disk it is to be written. Additionally, there is usually filesystem metadata that must be updated (e.g. free space counters), which will trigger more writes to the disk.
The time it takes from steps 1-3 can be unpredictable in a general purpose OS like Windows due to task scheduling, background threads, and your disk write is probably queued up with a few dozen other processes. I'd say it is usually on the order of 10-100msec on a typical PC. If you go to the Windows Resource Monitor and click the Disk tab, you can get an idea of the average disk queue length. You can use the Performance Monitor to produce more finely-controlled graphs.
Steps 3-4 are largely controlled by the disk controller and disk interface (SATA, SAS, etc). In the server world, you can be talking about a SAN with FC or iSCSI network switches, which impose their own latencies.
Step 5 will be controlled by they physical performance of the disk. Many consumer-grade HDD manufacturers do not post average seek times anymore, but 10-20msec is common.
Interesting detail about Step 5: Some HDDs lie about flushing their write cache to get better benchmark scores.
Step 6 will depend on your filesystem and how much data you are writing.
You are right that there can be a delay between Windows indicating that data writing is finished and the last data actually written. Things to consider are:
Device Manager, Disk Drive, Properties, Policies - Options for disabling Write Caching.
You might be better off using Direct I/O so that Windows does not save it temporarily in File Cache.
If your program writes the data, you can log what has been copied.
If you are sending the data over a network, you are likely to have no control of when the remote system has finished.
To see what is happening, you can set up Perfmon logging. One of my examples of monitoring:
http://www.roylongbottom.org.uk/monitor1.htm#anchor2
I'm currently learning Pig and I'm executing my scripts inside Hortonworks Sandbox. What is bugging me from the very beginning is that the minimum execution time for a Pig script seems to be at least 30-40 seconds. Is that because I'm using the Hortonworks Sandbox or is a normal for Pig scripts? Is there a way to reduce the execution time, because this is really slowing my learning progress? If this execution time is normal can you explain me what is going on and why is that?
PS
I've allocated 2GB RAM for the Hortonworks virtual machine. And just to mention I'm currently executing just simple scripts on small data sets.
If you execute pig in local mode (pig -x local) then it'll run a lot faster but it won't do map-reduce and won't access hdfs - it's good for learning though!
Yes, 30-40 seconds is absolutely normal for Pig, because it has a big overhead for compiling the job, launching JVMs, etc.
As stated in the other answer - you can try to run in local mode. It usually takes me about 15 seconds for a simple job with input containing just a few lines of data. My Cloudera VM is allocated with 4G of RAM, btw.
Lately I have been thinking about investing in a worthy USB pen drive (something along the lines of this), and install Operating Systems on Virtual Machines and start developing on them.
What I have in mind is that I want to be able to carry my development boxes, being a Windows Distribution for .Net development and a Linux Distribution for stuff like RoR, Perl and whatnot, so that I would be able to carry them around where need be...be it work, school, different computers at home etc...
I am thinking of doing this also for backup purposes...ie to backup my almost-single VM file to an external hd, instead of doing routinely updates to my normal Windows Box. I am also thinking about maybe even committing the VM boxes under Source Control (is that even feasible?)
So, am I on the right track with this ? Do you suggest that I try to implement this out?
How feasible is it to have your development box on Virtual Machine that runs from a USB Pen-Drive ?
I absolutely agree with where you are heading. I wish to do this myself.
But if you don't already know, it's not just about drive size, believe it or not USB Flash drives can be much slower than your spinning disk drives!
This can be a big problem if you plan to actually run the VMs directly from the USB drive!
I've tried running a 4GB Windows XP VM on a 32GB Corsair Survivor and the VM was virtually unusuable! Also copying my 4GB VM off and back onto the drive was also quite slow - about 10 minutes to copy it onto the drive.
If you have an esata port I'd highly recommend looking at high-speed ESata options like this Kanguru 32GB ESata/USB Flash drive OR this 32GB one by OCZ.
The read and write speeds of these drives are much higher over ESata than other USB drives. And you can still use them as USB if you don't have an ESata port. Though if you don't have an ESata port you can buy PCI to ESata cards online and even ESata ExpressCards for your laptop.
EDIT: A side note, you'll find the USB flash drives use FAT instead of NTFS. You don't want to use NTFS because it makes a lot more reads & writes on the disk and your drive will only have a limited number of reads & writes before it dies. But by using FAT you'll be limited to max 2GB file size which might be a problem with your VM. If this is the case, you can split your VM disks into 2GB chunks. Also make sure you backup your VM daily incase your drive does reach it's maximum number of writes. :)
This article on USB thumbdrives states,
Never run disk-intensive applications
directly against files stored on the
thumb drive.
USB thumbdrives utilize flash memory and these have a maximum number of writes before going bad and corruption occurs. The author of the previously linked article found it to be in the range of 10,000 - 100,000 writes but if you are using a disk intensive application this could be an issue.
So if you do this, have an aggressive backup policy to backup your work. Similarly, if when you run your development suite, if it could write to the local hard drive as a temporary workspace this would be ideal.
Hopefully you are talking about interpreted language projects. I couldn't imagine compiling a C/C++ of any size on a VM, let alone a VM running off of a USB drive.
I do it quite frequently with Xen, but also include a bare metal bootable kernel on the drive. This is particularly useful when working on something from which a live CD will be based.
The bad side is the bloat on the VM image to keep it bootable across many machines .. so where you would normally build a very lean and mean paravirtualized kernel only .. you have to also include one that has everything including the kitchen sink (up to what you want, i.e. do you need Audio, or token ring, etc?)
I usually carry two sticks, one has Xen + a patched Linux 2.6.26, the other has my various guest images which are ready to boot either way. A debootstrapped copy of Debian or Ubuntu makes a great starting point to create the former.
If nothing else, its fun to tinker with. Sorry to be a bit GNU/Linux centric, but that's what I use exclusively :) I started messing around with this when I had to find an odd path to upgrading my distro, which was two years behind the current one. So, I strapped a guest, installed what I wanted and pointed GRUB at the new LV for my root file system. Inside, I just mounted my old /home LV and away I went.
Check out MojoPac:
http://www.mojopac.com/
Hard-core gamers use it to take world of warcraft with them on the go -- it should work fine for your development needs, at least on Windows. Use cygwin with it for your unix-dev needs.
I used to do this, and found that compiling was so deathly slow, it wasn't worth it.
Keep in mind that USB flash drives are extremely slow (maybe 10 to 100 times slower) compared to hard drives at random write performance (writing lots of small files to a partition which already has lots of files).
A typical build process using GNU tools will create lots of small files - a simple configure script creates thousands of small files and deletes them again just to test the environment before you even start compiling. You could be waiting a long time.
I've got a machine I'm going to be using for development, and it has two 7200 RPM 160 GB SATA HDs in it.
The information I've found on the net so far seems to be a bit conflicted about which things (OS, Swap files, Programs, Solution/Source code/Other data) I should be installing on how many partitions on which drives to get the most benefit from this situation.
Some people suggest having a separate partition for the OS and/or Swap, some don't bother. Some people say the programs should be on the same physical drive as the OS with the data on the other, some the other way around. Same with the Swap and the OS.
I'm going to be installing Vista 64 bit as my OS and regularly using Visual Studio 2008, VMWare Workstation, SQL Server management studio, etc (pretty standard dev tools).
So I'm asking you--how would you do it?
If the drives support RAID configurations in your BIOS, you should do one of the following:
RAID 1 (Mirror) - Since this is a dev machine this will give you the fault tolerance and peace of mind that your code is safe (and the environment since they are such a pain to put together). You get better performance on reads because it can read from both/either drive. You don't get any performance boost on writes though.
RAID 0 - No fault tolerance here, but this is the fastest configuration because you read and write off both drives. Great if you just want as fast as possible performance and you know your code is safe elsewhere (source control) anyway.
Don't worry about mutiple partitions or OS/Data configs because on a dev machine you sort of need it all anyway and you shouldn't be running heavy multi-user databases or anything anyway (like a server).
If your BIOS doesn't support RAID configurations, however, then you might consider doing the OS/Data split over the two drives just to balance out their use (but as you mentioned, keep the programs on the system drive because it will help with caching). Up to you where to put the swap file (OS will give you dump files, but the data drive is probably less utilized).
If they're both going through the same disk controller, there's not going to be much difference performance-wise no matter which way you do it; if you're going to be doing lots of VM's, I would split one drive for OS and swap / Programs and Data, then keep all the VM's on the other drive.
Having all the VM's on an independant drive would let you move that drive to another machine seamlessly if the host fails, or if you upgrade.
Mark one drive as being your warehouse, put all of your source code, data, assets, etc. on there and back it up regularly. You'll want this to be stable and easy to recover. You can even switch My Documents to live here if wanted.
The other drive should contain the OS, drivers, and all applications. This makes it easy and secure to wipe the drive and reinstall the OS every 18-24 months as you tend to have to do with Windows.
If you want to improve performance, some say put the swap on the warehouse drive. This will increase OS performance, but will decrease the life of the drive.
In reality it all depends on your goals. If you need more performance then you even out the activity level. If you need more security then you use RAID and mirror it. My mix provides for easy maintenance with a reasonable level of data security and minimal bit rot problems.
Your most active files will be the registry, page file, and running applications. If you're doing lots of data crunching then those files will be very active as well.
I would suggest if 160gb total capacity will cover your needs (plenty of space for OS, Applications and source code, just depends on what else you plan to put on it), then you should mirror the drives in a RAID 1 unless you will have a server that data is backed up to, an external hard drive, an online backup solution, or some other means of keeping a copy of data on more then one physical drive.
If you need to use all of the drive capacity, I would suggest using the first drive for OS and Applications and second drive for data. Purely for the fact of, if you change computers at some point, the OS on the first drive doesn't do you much good and most Applications would have to be reinstalled, but you could take the entire data drive with you.
As for dividing off the OS, a big downfall of this is not giving the partition enough space and eventually you may need to use partitioning software to steal some space from the other partition on the drive. It never seems to fail that you allocate a certain amount of space for the OS partition, right after install you have several gigs free space so you think you are fine, but as time goes by, things build up on that partition and you run out of space.
With that in mind, I still typically do use an OS partition as it is useful when reloading a system, you can format that partition blowing away the OS but keep the rest of your data. Ways to keep the space build up from happening too fast is change the location of your my documents folder, change environment variables for items such as temp and tmp. However, there are some things that just refuse to put their data anywhere besides on the system partition. I used to use 10gb, these days I go for 20gb.
Dividing your swap space can be useful for keeping drive fragmentation down when letting your swap file grow and shrink as needed. Again this is an issue though of guessing how much swap you need. This will depend a lot on the amount of memory you have and how much stuff you will be running at one time.
For the posters suggesting RAID - it's probably OK at 160GB, but I'd hesitate for anything larger. Soft errors in the drives reduce the overall reliability of the RAID. See these articles for the details:
http://alumnit.ca/~apenwarr/log/?m=200809#08
http://permabit.wordpress.com/2008/08/20/are-fibre-channel-and-scsi-drives-more-reliable/
You can't believe everything you read on the internet, but the reasoning makes sense to me.
Sorry I wasn't actually able to answer your question.
I usually run a box with two drives. One for the OS, swap, typical programs and applications, and one for VMs, "big" apps (e.g., Adobe CS suite, anything that hits the disk a lot on startup, basically).
But I also run a cheap fileserver (just an old machine with a coupla hundred gigs of disk space in RAID1), that I use to store anything related to my various projects. I find this is a much nicer solution than storing everything on my main dev box, doesn't cost much, gives me somewhere to run a webserver, my personal version control, etc.
Although I admit, it really isn't doing much I couldn't do on my machine. I find it's a nice solution as it helps prevent me from spreading stuff around my workstation's filesystem at random by forcing me to keep all my work in one place where it can be easily backed up, copied elsewhere, etc. I can leave it on all night without huge power bills (it uses <50W under load) so it can back itself up to a remote site with a little script, I can connect to it from outside via SSH (so I can always SCP anything I need).
But really the most important benefit is that I store nothing of any value on my workstation box (at least nothing that isn't also on the server). That means if it breaks, or if I want to use my laptop, etc. everything is always accessible.
I would put the OS and all the applications on the first disk (1 partition). Then, put the data from the SQL server (and any other overflow data) on the second disk (1 partition). This is how I'd set up a machine without any other details about what you're building. Also make sure you have a backup so you don't lose work. It might even be worth it to mirror the two drives (if you have RAID capability) so you don't lose any progress if/when one of them fails. Also, backup to an external disk daily. The RAID won't save you when you accidentally delete the wrong thing.
In general I'd try to split up things that are going to be doing a lot of I/O (such as if you have autosave on VS going off fairly frequently) Think of it as sort of I/O multithreading
I've observed significant speedups by putting my virtual machines on a separate disk. Whenever Windows is doing something stupid in the VM (e.g., indexing yet again), it doesn't thrash my Mac's disk quite so badly.
Another issue is that many tools (Visual Studio comes to mind) break in frustrating ways when bits of them are on the non-primary disk.
Use your second disk for big random things.