VM Heap Reservation
in [Java Programming]
|
Prev: Try to install JSIM
Next: Arkanoid / Breakout
From: DaveR on 17 Aug 2006 13:55 On Sun's 1.5 VM running on Windows, it looks like the VM is trying to allocate (possibly contiguously) the entire virtual address space specified by -mx. This seems to be a change in behavior from previous VM's. Does anyone know if this is the case? I'm trying to wrap my head around what exactly -mx means now if this is the true. Do we need to ensure there's enough swap configured on the machine before booting the VM with a particular value for -mx?
From: Steve W. Jackson on 17 Aug 2006 14:35 In article <1155837348.179944.283670(a)m73g2000cwd.googlegroups.com>, "DaveR" <drubin99(a)yahoo.com> wrote: > On Sun's 1.5 VM running on Windows, it looks like the VM is trying to > allocate (possibly contiguously) the entire virtual address space > specified by -mx. This seems to be a change in behavior from previous > VM's. Does anyone know if this is the case? I'm trying to wrap my > head around what exactly -mx means now if this is the true. Do we need > to ensure there's enough swap configured on the machine before booting > the VM with a particular value for -mx? I've seen no evidence of a change in the behavior of -Xmx since switching to 1.5 on XP. -Xms is the initial heap size and -Xmx is the maximum. So over the life of an app, the heap should never exceed that size. You're sure to suffer ill effects if your combination of real RAM and swap space is tight (and I've never considered Windows good at managing its virtual memory), but it should still work if the -Xmx setting doesn't exceed what the VM can support. = Steve = -- Steve W. Jackson Montgomery, Alabama
From: DaveR on 17 Aug 2006 14:52 Understood regarding -Xms. I've create a single test class with a main containing only a System.out.println and have tried launching the 1.5 vm using -Xmx1024m (note there is no -Xms specified) and I get the following: Error occurred during initialization of VM could not reserve enough space for object heap Which makes absolutele no sense to me. If I lower the -Xmx to something like 750m, the VM launches successfully. The NT kernel is purely demand paged, so if you're Java code exhibits good page locality, you can get away with a small amount of physical adresses space while having adequate swap to cover the needs of your heap. Dave Steve W. Jackson wrote: > In article <1155837348.179944.283670(a)m73g2000cwd.googlegroups.com>, > "DaveR" <drubin99(a)yahoo.com> wrote: > > > On Sun's 1.5 VM running on Windows, it looks like the VM is trying to > > allocate (possibly contiguously) the entire virtual address space > > specified by -mx. This seems to be a change in behavior from previous > > VM's. Does anyone know if this is the case? I'm trying to wrap my > > head around what exactly -mx means now if this is the true. Do we need > > to ensure there's enough swap configured on the machine before booting > > the VM with a particular value for -mx? > > I've seen no evidence of a change in the behavior of -Xmx since > switching to 1.5 on XP. > > -Xms is the initial heap size and -Xmx is the maximum. So over the life > of an app, the heap should never exceed that size. You're sure to > suffer ill effects if your combination of real RAM and swap space is > tight (and I've never considered Windows good at managing its virtual > memory), but it should still work if the -Xmx setting doesn't exceed > what the VM can support. > > = Steve = > -- > Steve W. Jackson > Montgomery, Alabama
From: Steve W. Jackson on 18 Aug 2006 14:16 In article <1155840761.883917.45180(a)h48g2000cwc.googlegroups.com>, "DaveR" <drubin99(a)yahoo.com> wrote: > Understood regarding -Xms. I've create a single test class with a main > containing only a System.out.println and have tried launching the 1.5 > vm using -Xmx1024m (note there is no -Xms specified) and I get the > following: > > Error occurred during initialization of VM could not reserve enough > space for object heap > > Which makes absolutele no sense to me. If I lower the -Xmx to > something like 750m, the VM launches successfully. > > The NT kernel is purely demand paged, so if you're Java code exhibits > good page locality, you can get away with a small amount of physical > adresses space while having adequate swap to cover the needs of your > heap. > > Dave Not being able to use -Xmx1024m there is a serious problem, I'd say... I've been using that with our application for years now, even before we later added the -Xms part. I just pulled out a small class I use for quick testing that has a main with just a println and tested it using only the -Xmx1024m and it worked fine -- Java 1.5.0_06 on XP SP2. We stopped supporting NT 4 and earlier, so I haven't got access to any of those to see what I'd get. = Steve = -- Steve W. Jackson Montgomery, Alabama
From: nicolas.lehuen on 21 Aug 2006 12:40
> Not being able to use -Xmx1024m there is a serious problem, I'd say... > I've been using that with our application for years now, even before we > later added the -Xms part. I just pulled out a small class I use for > quick testing that has a main with just a println and tested it using > only the -Xmx1024m and it worked fine -- Java 1.5.0_06 on XP SP2. We > stopped supporting NT 4 and earlier, so I haven't got access to any of > those to see what I'd get. > > = Steve = > -- > Steve W. Jackson > Montgomery, Alabama I'm also encountering problems under both Windows XP and Windows 2003 Server with 2 GB of RAM. Up until a few weeks ago, I could use -Xmx1600M without any problem. Then during my vacation a colleague called me to tell me that the JVM wouldn't start anymore on our development server (the one under Windows 2003 Server) with this error message : Error occurred during initialization of VM Could not reserve enough space for object heap Could not create the Java virtual machine. Back then, I tried on my development machine (the one under Windows XP) and everything worked properly at -Xmx1600M. Now, coming back at work, my Windows has been updated up to the latest Patch Thursday (we run Windows Update Server at work so my PC wasn't updated until I returned to work). And now, -Xmx1600M doesn't work anymore on my dev PC AND on the dev server. This leads me to think about a negative interaction between the latest patches from Windows Update and the JVM. But I'm really not sure, so if anyone has a similar problem, I'd be glad to hear about it. Regards, Nicolas |